Screening for Losers Trade Institutions andInformation

Jason S Davis University of Pennsylvania

September 30 2019

Abstract

Trade law scholars have often argued that international institutions can serve a usefuldomestic political role by providing a constraint against domestic demands for protec-tion In this paper I identify a new way in which such institutions and their particularfeatures can be valuable to governments namely that they can provide useful infor-mation about domestic political groups While governments are responsible for the ad-ministration of most legal trade-related actions the information that governments needto determine which actions to pursue is often the private information of the firms andinterest groups that are lobbying for these actions and there are significant incentivesfor such groups to misrepresent this information This paper uses a formal model todemonstrate that governments can use the multitude of legal options available to themto screen between domestic groups for those with the strongest cases a selection processwhich can help to explain amongst other things why trade remedies tend to be struc-tured around meeting criteria instead of as ldquoefficient breachesrdquo requiring compensationwhy disputes pursued via the WTO have such a high rate of success (approximately90) and the conditions under which one would expect investor-state dispute settle-ment (ISDS) to be subject to abuses by private firms

Index termsmdash trade international institutions game theory trade remedies informationpolitics interest groups screening WTO

IntroductionIn determining which industries to protect from trade competition governments are facedwith difficult tradeoffs in an environment of incomplete information While governmentswould like to protect politically important groups who are threatened by trade competitiondoing so imposes aggregate welfare costs that may be politically costly and can impose sig-nificant concentrated costs on downstream industries that are also politically important (forinstance automobile manufacturers are affected by steel tariffs) As a consequence gov-ernments are often best served by protecting only a subset of groups that are harmed bytrade competition such as those that are harmed the most severely or those that are facing

1

ldquounfairrdquo competition from firms abroad However whether or not this is true for any groupis often the private information of that group and they do not have an incentive to honestlyreveal that information given that they benefit unconditionally from more protection

Similarly firms often have a clear sense of how well the laws and regulations of other coun-tries conform with international legal commitments given that these firms have an incen-tive to monitor such things and often experience the consequences directly However firmsbenefit unconditionally from any changes to the policy that will boost their goodsrsquo compet-itiveness and thus have an incentive to misrepresent the legality of any policies that theyfind unfavorable when lobbying the government Governments however may only want toinvest political capital and financial resources in the cases which involve truly unfair prac-tices on the part of foreign actors Thus a question arises are there mechanisms by whichinterest groups can credibly signal these characteristics of their situation in order to allowthe government to overcome the selection problem

This paper presents a formal model in which the selection of the forum andor instrumentthrough which an interest group pursues government support can perform this functionTrade institutions often provide many options for supporting domestic industries from sev-eral kinds of ldquotrade remediesrdquo which allow for temporary protection of industries that aresuffering injury as a result of trade competition (see for instance safeguard measures un-der Article XIX of the GATT or antidumping actions under Article VI) to procedures forchallenging the illegal practices of other countries However these institutions will oftenindependently assess the validity of such claims when faced with a dispute (eg DisputeSettlement proceedings in the WTO) As such given that institutions and instruments dif-fer in their standards for granting or challenging claims an interest grouprsquos selection of oneover another can credibly reveal information about the grouprsquos type

This is possible because with a high enough probability of a claim being rejected or over-turned the expected value of pursuing that claim for an interest group or firm can be lowerthan the fixed costs they would need to pay for lobbying and legal costs Thus institutionscan structure the choice environment such that it is only incentive compatible for the firmswith the strongest cases to pursue certain forms of government action

Consequently this paperrsquos model can improve our understanding of the use of trade institu-tions by states in a number of ways First it can help to explain why states and firms chooseparticular institutions or legal mechanisms over others when pursuing actions that couldfeasibly fall under several umbrellas Second it can help to explain the particularities ofhow institutions and trade remedies are structured - including why we observe a system inwhich states must meet specific criteria to invoke flexibility privileges instead of a systemof ldquocompensationrdquo to aggrieved parties Third it provides a new domestic-based explana-tion for how international institutions can be useful to governments namely that they canuse the threat of international arbitration to sort through a situation otherwise marred by

2

information asymmetries to better select which legal cases to pursue and which actors tosupport with protection

Overview of the LiteratureGiven the importance of trade institutions - especially the WTO - in the international le-gal landscape political scientists economists and legal scholars have devoted significantattention to unpacking the use of dispute settlement and trade remedies by governmentsAn important strand of this literature is the subset that addresses forum-shopping in inter-national trade (Busch 2007 Davis 2009) Broadly this work argues that when states arefiling a dispute they will select the forum that gives them the highest probability of successexcept in such cases where the informal precedent set by a ruling may bind on other sub-sets of countries If the precedent set by a case is likely to be featured in future disputes acountry selecting an institution may trade off the likelihood of success in a particular caseagainst how important the precedent is likely to be in future disputes with third parties(Busch 2007) This provides an explanation for why states might not always choose theinstitution that gives them the highest likelihood of success namely that they may alsoconsider whether they want to set a regional or global precedent

There remain outstanding legal and empirical questions about the degree to which prece-dent exists in international trade law while courts may rely on previous rulings when de-veloping their reasoning they are not bound by stare decisis and it has often been the USposition in disputes that previous Appellate Body rulings should have no impact on futurerulings1 In practice however empirical evidence from political scientists has been gener-ally supportive of the impact of a kind of ldquoinformalrdquo precedent (Kucik and Pelc 2014 Pelc2014) Without drawing any definitive conclusions on this open area of inquiry this paperprovides a different explanation for why states directed by interest groups might select in-stitutions that give them a lower likelihood of success In particular interest groups mightrequest pursuit of a dispute at an institution with rigorous evaluative standards as a wayof signaling credibly the strength of their legal case

Other important research on institutions discusses the inclusion of flexibility provisions ininternational trade agreements This work suggests that flexibility provisions are a meansof making trade agreements more ldquostablerdquo by allowing parties to the agreement a ldquosafetyvalverdquo in cases of heightened domestic political pressure for protection (Rosendorff and Mil-ner 2001 Koremenos 2001 Rosendorff 2005) These arguments outline an ldquoefficient breachrdquostory whereby states are required to pay some cost when making use of flexibility measuresin order to ensure such measures are only used in circumstances where the benefits of tem-porarily leaving the agreement exceed the costs to other parties (Rosendorff and Milner2001 p 831)

1See Panel Report US-Zeroing (Korea)

3

However this literature does not provide a clear sense of why an institution would includemultiple trade remedies indeed Rosendorff (2005) simply states that the different meansldquohave the same effect of allowing temporary relief when the local industry comes under pres-sure from foreign competitorsrdquo (Rosendorff 2005 p396) In contrast this paper suggests thathaving different instruments for temporary protection with differing probabilities of beingoverturned and differing benefits allows governments to match higher levels of protectionto those with the strongest cases

Moreover the efficient breach story of this literature is not especially consistent with theway that flexibility provisions are actually administered As Pelc notes ldquocompensation fol-lowing escape was only widespread in the 1950srdquo (Pelc 2011 p349) Instead trade remedieshave come to be governed by ldquoappeals to exceptionrdquo in which states need to justify theirtemporary departure from prior legal commitments by arguing that they meet certain crite-ria such as ldquosevere injuryrdquo in the face of ldquounforeseen circumstancesrdquo (Pelc 2011 p350) Pelcalso posits an informational explanation for this shift in governance arguing that ldquomembershave a strategic incentive to portray any instance where they face some domestic pressurefor protection as constituting true exigency arising from severe and unforeseeable circum-stancesrdquo suggesting that a criteria-based system with monitoring and enforcement can pro-vide incentives for states to honestly reveal information (Pelc 2011 p355)

This paperrsquos argument is in the same spirit but differs in a number of important respects(1) it suggests that private firms and interest groups not states are the actors that possessthis private information about the degree to which criteria are met (2) it argues that gov-ernments may also want to constrain their use of escape clauses to instances that meet theoutlined criteria but might struggle to do so because of the information asymmetries thatexist with these domestic actors Thus the model discusses intrastate information revela-tion instead of the interstate information revelation discussed by Pelc

More broadly the idea that plaintiffs might have private information about the strengthof their claims that is revealed in the legal process is not a new one to the law and eco-nomics literature (Bebchuck 1984 Reingaum and Wilde 1986) and the concept has evenbeen applied to the study of international courts (Gilligan et al 2010) However this pa-perrsquos particular focus on the private information of interest groups in trade politics and theinformation asymmetries with state-level governments that are the primary agents respon-sible for the pursuit of trade-related cases differentiates it from this other work

This paper also speaks to a relatively nascent literature on informational lobbying with re-spect to trade liberalization This literature outlines how firms can use costly lobbying tosignal the strength and value of potential cases (Brutger 2017 Betz 2018) Like Brutger(2017) I note that firms have private information about the strength of their cases that canbe difficult to reveal credibly but I posit a different (and complementary) mechanism bywhich groups can overcome this credibility problem

4

Lastly this paper speaks to a broader literature on how international institutions and inter-national law can be helpful to national governments (Keohane 1984 Koremenos et al 2001Koremenos 2001 Kucik and Pelc 2016 Pelc 2016) Amongst other things this literature hasdiscussed the role that institutions can play in ldquotying the handsrdquo of governments to policiesthat might be resisted by domestic groups and has identified ldquoinformationrdquo as a potentialbenefit of such institutions However these discussions of information and institutions havefocused almost exclusively on how institutions can help to provide information about otherstates by for instance monitoring compliance

This paper posits a new reason why international institutions can be helpful namely thatthey allow governments to obtain private information about interest groups so as to betterallocate protection and other forms of political support Like the ldquotying handsrdquo argumentthis suggests that governments may use agreements with foreign countries as much to ad-dress their internal politics as their external relations Unlike these previous argumentsthis paper focuses on how institutions can be leveraged to make information transfer fromdomestic groups more credible rather than used as a device to make the governmentrsquos com-mitments to unpopular policies more credible

Reviewing the Trade Law

The Institutional LandscapeWhile many initially hoped that the World Trade Organization (WTO) - established in 1995- would become a single forum through which trade liberalization could be negotiated on abroadly multilateral basis this hope has largely failed to come to fruition from 1990-2010the number of Preferential Trade Agreements (PTAs) rose from about 70 to approximately300 (Baccini et al 2011) This explosion in the number of PTAs has led to the peculiar situ-ation where trade between countries is often governed by several overlapping institutionssuch that complainants may have to choose between forums when lodging a complaint

Moreover these forums can differ significantly in the way they resolve disputes For in-stance while both the North American Free Trade Agreement (NAFTA) and the WTO havedispute resolution proceedings (NAFTA Chapter 19 amp 20 and the DSU respectively) theformer tends to draw panels of those with domestic judicial expertise (this was developedunder pressure from the US Congress to ensure that panelists would have an in-depth fa-miliarity with US Administrative law) while the latter is weighted more towards those withexpertise in international trade law and economics (Howse 1998) This can lead to differ-ences in how similar cases would be adjudicated given that trade law cases often rest ontechnical determinations of ldquomaterial injuryrdquo and ldquodumpingrdquo that are necessarily made byeconomists and statisticians For instance in the somewhat unusual case of the US-CanadaSoftwood Lumber dispute where aspects of the dispute were brought to both the WTO and

5

NAFTA a WTO panel ruled that a threat of material injury to US industries was presentwhile a NAFTA panel ruled precisely the opposite (Pauwelyn 2006)

If forums differ in how they adjudicate disputes and if states have a choice about whichforum to use when pursuing a claim then ldquoforum-shoppingrdquo becomes inevitable wherebystates and firms choose between different forums strategically Furthermore the issue ofoverlapping institutions is likely to continue to increase given that the rapidly rising num-ber of PTAs increases the likelihood that any given potential claim will be governed by morethan one institution

Busch (2007) identifies several reasons why a state might prefer one forum over anotherincluding timeliness of dispute resolution available remedies etc and (as discussed earlier)the value of any informal precedent set (Busch 2007) This paper provides another firmsand interest groups may lobby states to file disputes at particular forums over others asa way of credibly signaling the strength of their legal cases For most of these disputeresolution forums country-level governments are the only actors with standing to pursuedisputes which sets up the informational problem that is this paperrsquos primary subject ofinquiry as states seek to determine which disputes to pursue but private firms and interestgroups possess the information required to best make those decisions Thus this paperrsquosmodel suggests that forum choice can be one way of resolving this information problem

Flexibility ProvisionsIn addition to providing means for governments to claim violations of an agreement tradeagreements generally allow states to temporarily suspend certain obligations under specificcircumstances These have been referred to by Koremenos et al as ldquoflexibility provisionsrdquoie those provisions that allow members to ldquoescaperdquo temporarily from obligations under theagreement usually in response to some unanticipated shock (Koremenos et al 2001) ldquoSafe-guard measuresrdquo as they are often called exist under both NAFTA and the WTO NAFTAhas measures under Chapter 8 and the WTO has measures under GATT Article XIX (Tre-bilcock and Howse 2005) Moreover there are certain measures that have been interpretedas being de facto flexibility provisions even if they appear in principle to target unfair tradepractices for instance it has been argued that anti-dumping (AD) laws are a form of flex-ibility provision (Kucik and Reinhardt 2008) Countervailing duties (CVDs) are the mainother form of trade remedy and are formally a way of responding to prohibited or actionablesubsidies of other countries but they may be used instead as a flexibility provision (partic-ularly given that states have the option to challenge subsidies via the DSU)

Worth noting is that these trade remedies also differ in the degree to which they targetspecific countries or are applied more broadly CVDs apply to the countries accused of im-plementing subsidies and anti-dumping duties apply to the countries accused of dumpingboth are therefore explicitly discriminatory Safeguard measures in contrast are requiredto apply in a nondiscriminatory manner to all countries However Bown (2013) suggests

6

that in many cases appropriately targeted antidumping duties will have economic effectsthat are very similar to those of a broader safeguard measure

For all forms of trade remedies firms and interest groups representing them are likely tohave private information about the strength of their case All of these provisions usuallyrequire evidence of ldquoinjuryrdquo an empirical claim that necessitates showing both that thefirm is under economic duress and that this duress is causally related to competition frominternational trade Anti-dumping provisions also require evidence that imported goodsare being sold in the complainantrsquos country at a lower price than they are being sold inthe exporterrsquos country or that the goods are being sold below cost (Trebilcock and Howse2005) Similarly CVDs require evidence that the exporting country is subsidizing theirexport industries

The ArgumentFirms naturally have information about material injury and prices of competing goods aspart of their regular business practices while governments will often either not have accessto it (as with internal research conducted by the firm) or will be unwilling to expend theresources required to collect it Firms will also often have private information about thecompliance of other countries with their international legal obligations given that thesefirms have an incentive to invest more resources in monitoring other countries for viola-tions and will experience the consequences of violations directly when it affects their bal-ance sheet Indeed Brutger (2017) conducted interviews with a number of trade officialsand trade lawyers who had been involved in prominent trade disputes and found that thesetrade experts believed that the firms involved possessed this kind of private information(Brutger 2017 p 20)

This claim is reinforced by the fact that the agencies tasked with administering trade laware financially constrained for instance the Office of United States Trade Representative(USTR) which is tasked with pursuing any claims via the DSU has a budget of approxi-mately $50 million much of which is allocated towards funding trade negotiations insteadof monitoring domestic firm performance (USTR Budget Report 2015) Similarly the UnitedStates International Trade Commission (USITC) which is tasked with investigating injuryin AD CVD and safeguards cases has an annual budget of approximately $85 million ofwhich approximately $25 million is spent on the administration of trade remedies (USITCBudget Report 2016) The Commerce Department which is responsible for making deter-minations of dumping or unfair subsidies has a budget of approximately $80 million forldquoenforcement and compliancerdquo of ADCVD and has in fact had to request an additional $8million in the 2018 budget in order ldquoto develop factual information and legal justificationto self-initiate US antidumping duty (AD) and countervailing duty (CVD) investigationsrdquosuggesting that financial constraints have limited their capacity to do this in the past (Ross2018 p42)

7

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

ldquounfairrdquo competition from firms abroad However whether or not this is true for any groupis often the private information of that group and they do not have an incentive to honestlyreveal that information given that they benefit unconditionally from more protection

Similarly firms often have a clear sense of how well the laws and regulations of other coun-tries conform with international legal commitments given that these firms have an incen-tive to monitor such things and often experience the consequences directly However firmsbenefit unconditionally from any changes to the policy that will boost their goodsrsquo compet-itiveness and thus have an incentive to misrepresent the legality of any policies that theyfind unfavorable when lobbying the government Governments however may only want toinvest political capital and financial resources in the cases which involve truly unfair prac-tices on the part of foreign actors Thus a question arises are there mechanisms by whichinterest groups can credibly signal these characteristics of their situation in order to allowthe government to overcome the selection problem

This paper presents a formal model in which the selection of the forum andor instrumentthrough which an interest group pursues government support can perform this functionTrade institutions often provide many options for supporting domestic industries from sev-eral kinds of ldquotrade remediesrdquo which allow for temporary protection of industries that aresuffering injury as a result of trade competition (see for instance safeguard measures un-der Article XIX of the GATT or antidumping actions under Article VI) to procedures forchallenging the illegal practices of other countries However these institutions will oftenindependently assess the validity of such claims when faced with a dispute (eg DisputeSettlement proceedings in the WTO) As such given that institutions and instruments dif-fer in their standards for granting or challenging claims an interest grouprsquos selection of oneover another can credibly reveal information about the grouprsquos type

This is possible because with a high enough probability of a claim being rejected or over-turned the expected value of pursuing that claim for an interest group or firm can be lowerthan the fixed costs they would need to pay for lobbying and legal costs Thus institutionscan structure the choice environment such that it is only incentive compatible for the firmswith the strongest cases to pursue certain forms of government action

Consequently this paperrsquos model can improve our understanding of the use of trade institu-tions by states in a number of ways First it can help to explain why states and firms chooseparticular institutions or legal mechanisms over others when pursuing actions that couldfeasibly fall under several umbrellas Second it can help to explain the particularities ofhow institutions and trade remedies are structured - including why we observe a system inwhich states must meet specific criteria to invoke flexibility privileges instead of a systemof ldquocompensationrdquo to aggrieved parties Third it provides a new domestic-based explana-tion for how international institutions can be useful to governments namely that they canuse the threat of international arbitration to sort through a situation otherwise marred by

2

information asymmetries to better select which legal cases to pursue and which actors tosupport with protection

Overview of the LiteratureGiven the importance of trade institutions - especially the WTO - in the international le-gal landscape political scientists economists and legal scholars have devoted significantattention to unpacking the use of dispute settlement and trade remedies by governmentsAn important strand of this literature is the subset that addresses forum-shopping in inter-national trade (Busch 2007 Davis 2009) Broadly this work argues that when states arefiling a dispute they will select the forum that gives them the highest probability of successexcept in such cases where the informal precedent set by a ruling may bind on other sub-sets of countries If the precedent set by a case is likely to be featured in future disputes acountry selecting an institution may trade off the likelihood of success in a particular caseagainst how important the precedent is likely to be in future disputes with third parties(Busch 2007) This provides an explanation for why states might not always choose theinstitution that gives them the highest likelihood of success namely that they may alsoconsider whether they want to set a regional or global precedent

There remain outstanding legal and empirical questions about the degree to which prece-dent exists in international trade law while courts may rely on previous rulings when de-veloping their reasoning they are not bound by stare decisis and it has often been the USposition in disputes that previous Appellate Body rulings should have no impact on futurerulings1 In practice however empirical evidence from political scientists has been gener-ally supportive of the impact of a kind of ldquoinformalrdquo precedent (Kucik and Pelc 2014 Pelc2014) Without drawing any definitive conclusions on this open area of inquiry this paperprovides a different explanation for why states directed by interest groups might select in-stitutions that give them a lower likelihood of success In particular interest groups mightrequest pursuit of a dispute at an institution with rigorous evaluative standards as a wayof signaling credibly the strength of their legal case

Other important research on institutions discusses the inclusion of flexibility provisions ininternational trade agreements This work suggests that flexibility provisions are a meansof making trade agreements more ldquostablerdquo by allowing parties to the agreement a ldquosafetyvalverdquo in cases of heightened domestic political pressure for protection (Rosendorff and Mil-ner 2001 Koremenos 2001 Rosendorff 2005) These arguments outline an ldquoefficient breachrdquostory whereby states are required to pay some cost when making use of flexibility measuresin order to ensure such measures are only used in circumstances where the benefits of tem-porarily leaving the agreement exceed the costs to other parties (Rosendorff and Milner2001 p 831)

1See Panel Report US-Zeroing (Korea)

3

However this literature does not provide a clear sense of why an institution would includemultiple trade remedies indeed Rosendorff (2005) simply states that the different meansldquohave the same effect of allowing temporary relief when the local industry comes under pres-sure from foreign competitorsrdquo (Rosendorff 2005 p396) In contrast this paper suggests thathaving different instruments for temporary protection with differing probabilities of beingoverturned and differing benefits allows governments to match higher levels of protectionto those with the strongest cases

Moreover the efficient breach story of this literature is not especially consistent with theway that flexibility provisions are actually administered As Pelc notes ldquocompensation fol-lowing escape was only widespread in the 1950srdquo (Pelc 2011 p349) Instead trade remedieshave come to be governed by ldquoappeals to exceptionrdquo in which states need to justify theirtemporary departure from prior legal commitments by arguing that they meet certain crite-ria such as ldquosevere injuryrdquo in the face of ldquounforeseen circumstancesrdquo (Pelc 2011 p350) Pelcalso posits an informational explanation for this shift in governance arguing that ldquomembershave a strategic incentive to portray any instance where they face some domestic pressurefor protection as constituting true exigency arising from severe and unforeseeable circum-stancesrdquo suggesting that a criteria-based system with monitoring and enforcement can pro-vide incentives for states to honestly reveal information (Pelc 2011 p355)

This paperrsquos argument is in the same spirit but differs in a number of important respects(1) it suggests that private firms and interest groups not states are the actors that possessthis private information about the degree to which criteria are met (2) it argues that gov-ernments may also want to constrain their use of escape clauses to instances that meet theoutlined criteria but might struggle to do so because of the information asymmetries thatexist with these domestic actors Thus the model discusses intrastate information revela-tion instead of the interstate information revelation discussed by Pelc

More broadly the idea that plaintiffs might have private information about the strengthof their claims that is revealed in the legal process is not a new one to the law and eco-nomics literature (Bebchuck 1984 Reingaum and Wilde 1986) and the concept has evenbeen applied to the study of international courts (Gilligan et al 2010) However this pa-perrsquos particular focus on the private information of interest groups in trade politics and theinformation asymmetries with state-level governments that are the primary agents respon-sible for the pursuit of trade-related cases differentiates it from this other work

This paper also speaks to a relatively nascent literature on informational lobbying with re-spect to trade liberalization This literature outlines how firms can use costly lobbying tosignal the strength and value of potential cases (Brutger 2017 Betz 2018) Like Brutger(2017) I note that firms have private information about the strength of their cases that canbe difficult to reveal credibly but I posit a different (and complementary) mechanism bywhich groups can overcome this credibility problem

4

Lastly this paper speaks to a broader literature on how international institutions and inter-national law can be helpful to national governments (Keohane 1984 Koremenos et al 2001Koremenos 2001 Kucik and Pelc 2016 Pelc 2016) Amongst other things this literature hasdiscussed the role that institutions can play in ldquotying the handsrdquo of governments to policiesthat might be resisted by domestic groups and has identified ldquoinformationrdquo as a potentialbenefit of such institutions However these discussions of information and institutions havefocused almost exclusively on how institutions can help to provide information about otherstates by for instance monitoring compliance

This paper posits a new reason why international institutions can be helpful namely thatthey allow governments to obtain private information about interest groups so as to betterallocate protection and other forms of political support Like the ldquotying handsrdquo argumentthis suggests that governments may use agreements with foreign countries as much to ad-dress their internal politics as their external relations Unlike these previous argumentsthis paper focuses on how institutions can be leveraged to make information transfer fromdomestic groups more credible rather than used as a device to make the governmentrsquos com-mitments to unpopular policies more credible

Reviewing the Trade Law

The Institutional LandscapeWhile many initially hoped that the World Trade Organization (WTO) - established in 1995- would become a single forum through which trade liberalization could be negotiated on abroadly multilateral basis this hope has largely failed to come to fruition from 1990-2010the number of Preferential Trade Agreements (PTAs) rose from about 70 to approximately300 (Baccini et al 2011) This explosion in the number of PTAs has led to the peculiar situ-ation where trade between countries is often governed by several overlapping institutionssuch that complainants may have to choose between forums when lodging a complaint

Moreover these forums can differ significantly in the way they resolve disputes For in-stance while both the North American Free Trade Agreement (NAFTA) and the WTO havedispute resolution proceedings (NAFTA Chapter 19 amp 20 and the DSU respectively) theformer tends to draw panels of those with domestic judicial expertise (this was developedunder pressure from the US Congress to ensure that panelists would have an in-depth fa-miliarity with US Administrative law) while the latter is weighted more towards those withexpertise in international trade law and economics (Howse 1998) This can lead to differ-ences in how similar cases would be adjudicated given that trade law cases often rest ontechnical determinations of ldquomaterial injuryrdquo and ldquodumpingrdquo that are necessarily made byeconomists and statisticians For instance in the somewhat unusual case of the US-CanadaSoftwood Lumber dispute where aspects of the dispute were brought to both the WTO and

5

NAFTA a WTO panel ruled that a threat of material injury to US industries was presentwhile a NAFTA panel ruled precisely the opposite (Pauwelyn 2006)

If forums differ in how they adjudicate disputes and if states have a choice about whichforum to use when pursuing a claim then ldquoforum-shoppingrdquo becomes inevitable wherebystates and firms choose between different forums strategically Furthermore the issue ofoverlapping institutions is likely to continue to increase given that the rapidly rising num-ber of PTAs increases the likelihood that any given potential claim will be governed by morethan one institution

Busch (2007) identifies several reasons why a state might prefer one forum over anotherincluding timeliness of dispute resolution available remedies etc and (as discussed earlier)the value of any informal precedent set (Busch 2007) This paper provides another firmsand interest groups may lobby states to file disputes at particular forums over others asa way of credibly signaling the strength of their legal cases For most of these disputeresolution forums country-level governments are the only actors with standing to pursuedisputes which sets up the informational problem that is this paperrsquos primary subject ofinquiry as states seek to determine which disputes to pursue but private firms and interestgroups possess the information required to best make those decisions Thus this paperrsquosmodel suggests that forum choice can be one way of resolving this information problem

Flexibility ProvisionsIn addition to providing means for governments to claim violations of an agreement tradeagreements generally allow states to temporarily suspend certain obligations under specificcircumstances These have been referred to by Koremenos et al as ldquoflexibility provisionsrdquoie those provisions that allow members to ldquoescaperdquo temporarily from obligations under theagreement usually in response to some unanticipated shock (Koremenos et al 2001) ldquoSafe-guard measuresrdquo as they are often called exist under both NAFTA and the WTO NAFTAhas measures under Chapter 8 and the WTO has measures under GATT Article XIX (Tre-bilcock and Howse 2005) Moreover there are certain measures that have been interpretedas being de facto flexibility provisions even if they appear in principle to target unfair tradepractices for instance it has been argued that anti-dumping (AD) laws are a form of flex-ibility provision (Kucik and Reinhardt 2008) Countervailing duties (CVDs) are the mainother form of trade remedy and are formally a way of responding to prohibited or actionablesubsidies of other countries but they may be used instead as a flexibility provision (partic-ularly given that states have the option to challenge subsidies via the DSU)

Worth noting is that these trade remedies also differ in the degree to which they targetspecific countries or are applied more broadly CVDs apply to the countries accused of im-plementing subsidies and anti-dumping duties apply to the countries accused of dumpingboth are therefore explicitly discriminatory Safeguard measures in contrast are requiredto apply in a nondiscriminatory manner to all countries However Bown (2013) suggests

6

that in many cases appropriately targeted antidumping duties will have economic effectsthat are very similar to those of a broader safeguard measure

For all forms of trade remedies firms and interest groups representing them are likely tohave private information about the strength of their case All of these provisions usuallyrequire evidence of ldquoinjuryrdquo an empirical claim that necessitates showing both that thefirm is under economic duress and that this duress is causally related to competition frominternational trade Anti-dumping provisions also require evidence that imported goodsare being sold in the complainantrsquos country at a lower price than they are being sold inthe exporterrsquos country or that the goods are being sold below cost (Trebilcock and Howse2005) Similarly CVDs require evidence that the exporting country is subsidizing theirexport industries

The ArgumentFirms naturally have information about material injury and prices of competing goods aspart of their regular business practices while governments will often either not have accessto it (as with internal research conducted by the firm) or will be unwilling to expend theresources required to collect it Firms will also often have private information about thecompliance of other countries with their international legal obligations given that thesefirms have an incentive to invest more resources in monitoring other countries for viola-tions and will experience the consequences of violations directly when it affects their bal-ance sheet Indeed Brutger (2017) conducted interviews with a number of trade officialsand trade lawyers who had been involved in prominent trade disputes and found that thesetrade experts believed that the firms involved possessed this kind of private information(Brutger 2017 p 20)

This claim is reinforced by the fact that the agencies tasked with administering trade laware financially constrained for instance the Office of United States Trade Representative(USTR) which is tasked with pursuing any claims via the DSU has a budget of approxi-mately $50 million much of which is allocated towards funding trade negotiations insteadof monitoring domestic firm performance (USTR Budget Report 2015) Similarly the UnitedStates International Trade Commission (USITC) which is tasked with investigating injuryin AD CVD and safeguards cases has an annual budget of approximately $85 million ofwhich approximately $25 million is spent on the administration of trade remedies (USITCBudget Report 2016) The Commerce Department which is responsible for making deter-minations of dumping or unfair subsidies has a budget of approximately $80 million forldquoenforcement and compliancerdquo of ADCVD and has in fact had to request an additional $8million in the 2018 budget in order ldquoto develop factual information and legal justificationto self-initiate US antidumping duty (AD) and countervailing duty (CVD) investigationsrdquosuggesting that financial constraints have limited their capacity to do this in the past (Ross2018 p42)

7

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

information asymmetries to better select which legal cases to pursue and which actors tosupport with protection

Overview of the LiteratureGiven the importance of trade institutions - especially the WTO - in the international le-gal landscape political scientists economists and legal scholars have devoted significantattention to unpacking the use of dispute settlement and trade remedies by governmentsAn important strand of this literature is the subset that addresses forum-shopping in inter-national trade (Busch 2007 Davis 2009) Broadly this work argues that when states arefiling a dispute they will select the forum that gives them the highest probability of successexcept in such cases where the informal precedent set by a ruling may bind on other sub-sets of countries If the precedent set by a case is likely to be featured in future disputes acountry selecting an institution may trade off the likelihood of success in a particular caseagainst how important the precedent is likely to be in future disputes with third parties(Busch 2007) This provides an explanation for why states might not always choose theinstitution that gives them the highest likelihood of success namely that they may alsoconsider whether they want to set a regional or global precedent

There remain outstanding legal and empirical questions about the degree to which prece-dent exists in international trade law while courts may rely on previous rulings when de-veloping their reasoning they are not bound by stare decisis and it has often been the USposition in disputes that previous Appellate Body rulings should have no impact on futurerulings1 In practice however empirical evidence from political scientists has been gener-ally supportive of the impact of a kind of ldquoinformalrdquo precedent (Kucik and Pelc 2014 Pelc2014) Without drawing any definitive conclusions on this open area of inquiry this paperprovides a different explanation for why states directed by interest groups might select in-stitutions that give them a lower likelihood of success In particular interest groups mightrequest pursuit of a dispute at an institution with rigorous evaluative standards as a wayof signaling credibly the strength of their legal case

Other important research on institutions discusses the inclusion of flexibility provisions ininternational trade agreements This work suggests that flexibility provisions are a meansof making trade agreements more ldquostablerdquo by allowing parties to the agreement a ldquosafetyvalverdquo in cases of heightened domestic political pressure for protection (Rosendorff and Mil-ner 2001 Koremenos 2001 Rosendorff 2005) These arguments outline an ldquoefficient breachrdquostory whereby states are required to pay some cost when making use of flexibility measuresin order to ensure such measures are only used in circumstances where the benefits of tem-porarily leaving the agreement exceed the costs to other parties (Rosendorff and Milner2001 p 831)

1See Panel Report US-Zeroing (Korea)

3

However this literature does not provide a clear sense of why an institution would includemultiple trade remedies indeed Rosendorff (2005) simply states that the different meansldquohave the same effect of allowing temporary relief when the local industry comes under pres-sure from foreign competitorsrdquo (Rosendorff 2005 p396) In contrast this paper suggests thathaving different instruments for temporary protection with differing probabilities of beingoverturned and differing benefits allows governments to match higher levels of protectionto those with the strongest cases

Moreover the efficient breach story of this literature is not especially consistent with theway that flexibility provisions are actually administered As Pelc notes ldquocompensation fol-lowing escape was only widespread in the 1950srdquo (Pelc 2011 p349) Instead trade remedieshave come to be governed by ldquoappeals to exceptionrdquo in which states need to justify theirtemporary departure from prior legal commitments by arguing that they meet certain crite-ria such as ldquosevere injuryrdquo in the face of ldquounforeseen circumstancesrdquo (Pelc 2011 p350) Pelcalso posits an informational explanation for this shift in governance arguing that ldquomembershave a strategic incentive to portray any instance where they face some domestic pressurefor protection as constituting true exigency arising from severe and unforeseeable circum-stancesrdquo suggesting that a criteria-based system with monitoring and enforcement can pro-vide incentives for states to honestly reveal information (Pelc 2011 p355)

This paperrsquos argument is in the same spirit but differs in a number of important respects(1) it suggests that private firms and interest groups not states are the actors that possessthis private information about the degree to which criteria are met (2) it argues that gov-ernments may also want to constrain their use of escape clauses to instances that meet theoutlined criteria but might struggle to do so because of the information asymmetries thatexist with these domestic actors Thus the model discusses intrastate information revela-tion instead of the interstate information revelation discussed by Pelc

More broadly the idea that plaintiffs might have private information about the strengthof their claims that is revealed in the legal process is not a new one to the law and eco-nomics literature (Bebchuck 1984 Reingaum and Wilde 1986) and the concept has evenbeen applied to the study of international courts (Gilligan et al 2010) However this pa-perrsquos particular focus on the private information of interest groups in trade politics and theinformation asymmetries with state-level governments that are the primary agents respon-sible for the pursuit of trade-related cases differentiates it from this other work

This paper also speaks to a relatively nascent literature on informational lobbying with re-spect to trade liberalization This literature outlines how firms can use costly lobbying tosignal the strength and value of potential cases (Brutger 2017 Betz 2018) Like Brutger(2017) I note that firms have private information about the strength of their cases that canbe difficult to reveal credibly but I posit a different (and complementary) mechanism bywhich groups can overcome this credibility problem

4

Lastly this paper speaks to a broader literature on how international institutions and inter-national law can be helpful to national governments (Keohane 1984 Koremenos et al 2001Koremenos 2001 Kucik and Pelc 2016 Pelc 2016) Amongst other things this literature hasdiscussed the role that institutions can play in ldquotying the handsrdquo of governments to policiesthat might be resisted by domestic groups and has identified ldquoinformationrdquo as a potentialbenefit of such institutions However these discussions of information and institutions havefocused almost exclusively on how institutions can help to provide information about otherstates by for instance monitoring compliance

This paper posits a new reason why international institutions can be helpful namely thatthey allow governments to obtain private information about interest groups so as to betterallocate protection and other forms of political support Like the ldquotying handsrdquo argumentthis suggests that governments may use agreements with foreign countries as much to ad-dress their internal politics as their external relations Unlike these previous argumentsthis paper focuses on how institutions can be leveraged to make information transfer fromdomestic groups more credible rather than used as a device to make the governmentrsquos com-mitments to unpopular policies more credible

Reviewing the Trade Law

The Institutional LandscapeWhile many initially hoped that the World Trade Organization (WTO) - established in 1995- would become a single forum through which trade liberalization could be negotiated on abroadly multilateral basis this hope has largely failed to come to fruition from 1990-2010the number of Preferential Trade Agreements (PTAs) rose from about 70 to approximately300 (Baccini et al 2011) This explosion in the number of PTAs has led to the peculiar situ-ation where trade between countries is often governed by several overlapping institutionssuch that complainants may have to choose between forums when lodging a complaint

Moreover these forums can differ significantly in the way they resolve disputes For in-stance while both the North American Free Trade Agreement (NAFTA) and the WTO havedispute resolution proceedings (NAFTA Chapter 19 amp 20 and the DSU respectively) theformer tends to draw panels of those with domestic judicial expertise (this was developedunder pressure from the US Congress to ensure that panelists would have an in-depth fa-miliarity with US Administrative law) while the latter is weighted more towards those withexpertise in international trade law and economics (Howse 1998) This can lead to differ-ences in how similar cases would be adjudicated given that trade law cases often rest ontechnical determinations of ldquomaterial injuryrdquo and ldquodumpingrdquo that are necessarily made byeconomists and statisticians For instance in the somewhat unusual case of the US-CanadaSoftwood Lumber dispute where aspects of the dispute were brought to both the WTO and

5

NAFTA a WTO panel ruled that a threat of material injury to US industries was presentwhile a NAFTA panel ruled precisely the opposite (Pauwelyn 2006)

If forums differ in how they adjudicate disputes and if states have a choice about whichforum to use when pursuing a claim then ldquoforum-shoppingrdquo becomes inevitable wherebystates and firms choose between different forums strategically Furthermore the issue ofoverlapping institutions is likely to continue to increase given that the rapidly rising num-ber of PTAs increases the likelihood that any given potential claim will be governed by morethan one institution

Busch (2007) identifies several reasons why a state might prefer one forum over anotherincluding timeliness of dispute resolution available remedies etc and (as discussed earlier)the value of any informal precedent set (Busch 2007) This paper provides another firmsand interest groups may lobby states to file disputes at particular forums over others asa way of credibly signaling the strength of their legal cases For most of these disputeresolution forums country-level governments are the only actors with standing to pursuedisputes which sets up the informational problem that is this paperrsquos primary subject ofinquiry as states seek to determine which disputes to pursue but private firms and interestgroups possess the information required to best make those decisions Thus this paperrsquosmodel suggests that forum choice can be one way of resolving this information problem

Flexibility ProvisionsIn addition to providing means for governments to claim violations of an agreement tradeagreements generally allow states to temporarily suspend certain obligations under specificcircumstances These have been referred to by Koremenos et al as ldquoflexibility provisionsrdquoie those provisions that allow members to ldquoescaperdquo temporarily from obligations under theagreement usually in response to some unanticipated shock (Koremenos et al 2001) ldquoSafe-guard measuresrdquo as they are often called exist under both NAFTA and the WTO NAFTAhas measures under Chapter 8 and the WTO has measures under GATT Article XIX (Tre-bilcock and Howse 2005) Moreover there are certain measures that have been interpretedas being de facto flexibility provisions even if they appear in principle to target unfair tradepractices for instance it has been argued that anti-dumping (AD) laws are a form of flex-ibility provision (Kucik and Reinhardt 2008) Countervailing duties (CVDs) are the mainother form of trade remedy and are formally a way of responding to prohibited or actionablesubsidies of other countries but they may be used instead as a flexibility provision (partic-ularly given that states have the option to challenge subsidies via the DSU)

Worth noting is that these trade remedies also differ in the degree to which they targetspecific countries or are applied more broadly CVDs apply to the countries accused of im-plementing subsidies and anti-dumping duties apply to the countries accused of dumpingboth are therefore explicitly discriminatory Safeguard measures in contrast are requiredto apply in a nondiscriminatory manner to all countries However Bown (2013) suggests

6

that in many cases appropriately targeted antidumping duties will have economic effectsthat are very similar to those of a broader safeguard measure

For all forms of trade remedies firms and interest groups representing them are likely tohave private information about the strength of their case All of these provisions usuallyrequire evidence of ldquoinjuryrdquo an empirical claim that necessitates showing both that thefirm is under economic duress and that this duress is causally related to competition frominternational trade Anti-dumping provisions also require evidence that imported goodsare being sold in the complainantrsquos country at a lower price than they are being sold inthe exporterrsquos country or that the goods are being sold below cost (Trebilcock and Howse2005) Similarly CVDs require evidence that the exporting country is subsidizing theirexport industries

The ArgumentFirms naturally have information about material injury and prices of competing goods aspart of their regular business practices while governments will often either not have accessto it (as with internal research conducted by the firm) or will be unwilling to expend theresources required to collect it Firms will also often have private information about thecompliance of other countries with their international legal obligations given that thesefirms have an incentive to invest more resources in monitoring other countries for viola-tions and will experience the consequences of violations directly when it affects their bal-ance sheet Indeed Brutger (2017) conducted interviews with a number of trade officialsand trade lawyers who had been involved in prominent trade disputes and found that thesetrade experts believed that the firms involved possessed this kind of private information(Brutger 2017 p 20)

This claim is reinforced by the fact that the agencies tasked with administering trade laware financially constrained for instance the Office of United States Trade Representative(USTR) which is tasked with pursuing any claims via the DSU has a budget of approxi-mately $50 million much of which is allocated towards funding trade negotiations insteadof monitoring domestic firm performance (USTR Budget Report 2015) Similarly the UnitedStates International Trade Commission (USITC) which is tasked with investigating injuryin AD CVD and safeguards cases has an annual budget of approximately $85 million ofwhich approximately $25 million is spent on the administration of trade remedies (USITCBudget Report 2016) The Commerce Department which is responsible for making deter-minations of dumping or unfair subsidies has a budget of approximately $80 million forldquoenforcement and compliancerdquo of ADCVD and has in fact had to request an additional $8million in the 2018 budget in order ldquoto develop factual information and legal justificationto self-initiate US antidumping duty (AD) and countervailing duty (CVD) investigationsrdquosuggesting that financial constraints have limited their capacity to do this in the past (Ross2018 p42)

7

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

However this literature does not provide a clear sense of why an institution would includemultiple trade remedies indeed Rosendorff (2005) simply states that the different meansldquohave the same effect of allowing temporary relief when the local industry comes under pres-sure from foreign competitorsrdquo (Rosendorff 2005 p396) In contrast this paper suggests thathaving different instruments for temporary protection with differing probabilities of beingoverturned and differing benefits allows governments to match higher levels of protectionto those with the strongest cases

Moreover the efficient breach story of this literature is not especially consistent with theway that flexibility provisions are actually administered As Pelc notes ldquocompensation fol-lowing escape was only widespread in the 1950srdquo (Pelc 2011 p349) Instead trade remedieshave come to be governed by ldquoappeals to exceptionrdquo in which states need to justify theirtemporary departure from prior legal commitments by arguing that they meet certain crite-ria such as ldquosevere injuryrdquo in the face of ldquounforeseen circumstancesrdquo (Pelc 2011 p350) Pelcalso posits an informational explanation for this shift in governance arguing that ldquomembershave a strategic incentive to portray any instance where they face some domestic pressurefor protection as constituting true exigency arising from severe and unforeseeable circum-stancesrdquo suggesting that a criteria-based system with monitoring and enforcement can pro-vide incentives for states to honestly reveal information (Pelc 2011 p355)

This paperrsquos argument is in the same spirit but differs in a number of important respects(1) it suggests that private firms and interest groups not states are the actors that possessthis private information about the degree to which criteria are met (2) it argues that gov-ernments may also want to constrain their use of escape clauses to instances that meet theoutlined criteria but might struggle to do so because of the information asymmetries thatexist with these domestic actors Thus the model discusses intrastate information revela-tion instead of the interstate information revelation discussed by Pelc

More broadly the idea that plaintiffs might have private information about the strengthof their claims that is revealed in the legal process is not a new one to the law and eco-nomics literature (Bebchuck 1984 Reingaum and Wilde 1986) and the concept has evenbeen applied to the study of international courts (Gilligan et al 2010) However this pa-perrsquos particular focus on the private information of interest groups in trade politics and theinformation asymmetries with state-level governments that are the primary agents respon-sible for the pursuit of trade-related cases differentiates it from this other work

This paper also speaks to a relatively nascent literature on informational lobbying with re-spect to trade liberalization This literature outlines how firms can use costly lobbying tosignal the strength and value of potential cases (Brutger 2017 Betz 2018) Like Brutger(2017) I note that firms have private information about the strength of their cases that canbe difficult to reveal credibly but I posit a different (and complementary) mechanism bywhich groups can overcome this credibility problem

4

Lastly this paper speaks to a broader literature on how international institutions and inter-national law can be helpful to national governments (Keohane 1984 Koremenos et al 2001Koremenos 2001 Kucik and Pelc 2016 Pelc 2016) Amongst other things this literature hasdiscussed the role that institutions can play in ldquotying the handsrdquo of governments to policiesthat might be resisted by domestic groups and has identified ldquoinformationrdquo as a potentialbenefit of such institutions However these discussions of information and institutions havefocused almost exclusively on how institutions can help to provide information about otherstates by for instance monitoring compliance

This paper posits a new reason why international institutions can be helpful namely thatthey allow governments to obtain private information about interest groups so as to betterallocate protection and other forms of political support Like the ldquotying handsrdquo argumentthis suggests that governments may use agreements with foreign countries as much to ad-dress their internal politics as their external relations Unlike these previous argumentsthis paper focuses on how institutions can be leveraged to make information transfer fromdomestic groups more credible rather than used as a device to make the governmentrsquos com-mitments to unpopular policies more credible

Reviewing the Trade Law

The Institutional LandscapeWhile many initially hoped that the World Trade Organization (WTO) - established in 1995- would become a single forum through which trade liberalization could be negotiated on abroadly multilateral basis this hope has largely failed to come to fruition from 1990-2010the number of Preferential Trade Agreements (PTAs) rose from about 70 to approximately300 (Baccini et al 2011) This explosion in the number of PTAs has led to the peculiar situ-ation where trade between countries is often governed by several overlapping institutionssuch that complainants may have to choose between forums when lodging a complaint

Moreover these forums can differ significantly in the way they resolve disputes For in-stance while both the North American Free Trade Agreement (NAFTA) and the WTO havedispute resolution proceedings (NAFTA Chapter 19 amp 20 and the DSU respectively) theformer tends to draw panels of those with domestic judicial expertise (this was developedunder pressure from the US Congress to ensure that panelists would have an in-depth fa-miliarity with US Administrative law) while the latter is weighted more towards those withexpertise in international trade law and economics (Howse 1998) This can lead to differ-ences in how similar cases would be adjudicated given that trade law cases often rest ontechnical determinations of ldquomaterial injuryrdquo and ldquodumpingrdquo that are necessarily made byeconomists and statisticians For instance in the somewhat unusual case of the US-CanadaSoftwood Lumber dispute where aspects of the dispute were brought to both the WTO and

5

NAFTA a WTO panel ruled that a threat of material injury to US industries was presentwhile a NAFTA panel ruled precisely the opposite (Pauwelyn 2006)

If forums differ in how they adjudicate disputes and if states have a choice about whichforum to use when pursuing a claim then ldquoforum-shoppingrdquo becomes inevitable wherebystates and firms choose between different forums strategically Furthermore the issue ofoverlapping institutions is likely to continue to increase given that the rapidly rising num-ber of PTAs increases the likelihood that any given potential claim will be governed by morethan one institution

Busch (2007) identifies several reasons why a state might prefer one forum over anotherincluding timeliness of dispute resolution available remedies etc and (as discussed earlier)the value of any informal precedent set (Busch 2007) This paper provides another firmsand interest groups may lobby states to file disputes at particular forums over others asa way of credibly signaling the strength of their legal cases For most of these disputeresolution forums country-level governments are the only actors with standing to pursuedisputes which sets up the informational problem that is this paperrsquos primary subject ofinquiry as states seek to determine which disputes to pursue but private firms and interestgroups possess the information required to best make those decisions Thus this paperrsquosmodel suggests that forum choice can be one way of resolving this information problem

Flexibility ProvisionsIn addition to providing means for governments to claim violations of an agreement tradeagreements generally allow states to temporarily suspend certain obligations under specificcircumstances These have been referred to by Koremenos et al as ldquoflexibility provisionsrdquoie those provisions that allow members to ldquoescaperdquo temporarily from obligations under theagreement usually in response to some unanticipated shock (Koremenos et al 2001) ldquoSafe-guard measuresrdquo as they are often called exist under both NAFTA and the WTO NAFTAhas measures under Chapter 8 and the WTO has measures under GATT Article XIX (Tre-bilcock and Howse 2005) Moreover there are certain measures that have been interpretedas being de facto flexibility provisions even if they appear in principle to target unfair tradepractices for instance it has been argued that anti-dumping (AD) laws are a form of flex-ibility provision (Kucik and Reinhardt 2008) Countervailing duties (CVDs) are the mainother form of trade remedy and are formally a way of responding to prohibited or actionablesubsidies of other countries but they may be used instead as a flexibility provision (partic-ularly given that states have the option to challenge subsidies via the DSU)

Worth noting is that these trade remedies also differ in the degree to which they targetspecific countries or are applied more broadly CVDs apply to the countries accused of im-plementing subsidies and anti-dumping duties apply to the countries accused of dumpingboth are therefore explicitly discriminatory Safeguard measures in contrast are requiredto apply in a nondiscriminatory manner to all countries However Bown (2013) suggests

6

that in many cases appropriately targeted antidumping duties will have economic effectsthat are very similar to those of a broader safeguard measure

For all forms of trade remedies firms and interest groups representing them are likely tohave private information about the strength of their case All of these provisions usuallyrequire evidence of ldquoinjuryrdquo an empirical claim that necessitates showing both that thefirm is under economic duress and that this duress is causally related to competition frominternational trade Anti-dumping provisions also require evidence that imported goodsare being sold in the complainantrsquos country at a lower price than they are being sold inthe exporterrsquos country or that the goods are being sold below cost (Trebilcock and Howse2005) Similarly CVDs require evidence that the exporting country is subsidizing theirexport industries

The ArgumentFirms naturally have information about material injury and prices of competing goods aspart of their regular business practices while governments will often either not have accessto it (as with internal research conducted by the firm) or will be unwilling to expend theresources required to collect it Firms will also often have private information about thecompliance of other countries with their international legal obligations given that thesefirms have an incentive to invest more resources in monitoring other countries for viola-tions and will experience the consequences of violations directly when it affects their bal-ance sheet Indeed Brutger (2017) conducted interviews with a number of trade officialsand trade lawyers who had been involved in prominent trade disputes and found that thesetrade experts believed that the firms involved possessed this kind of private information(Brutger 2017 p 20)

This claim is reinforced by the fact that the agencies tasked with administering trade laware financially constrained for instance the Office of United States Trade Representative(USTR) which is tasked with pursuing any claims via the DSU has a budget of approxi-mately $50 million much of which is allocated towards funding trade negotiations insteadof monitoring domestic firm performance (USTR Budget Report 2015) Similarly the UnitedStates International Trade Commission (USITC) which is tasked with investigating injuryin AD CVD and safeguards cases has an annual budget of approximately $85 million ofwhich approximately $25 million is spent on the administration of trade remedies (USITCBudget Report 2016) The Commerce Department which is responsible for making deter-minations of dumping or unfair subsidies has a budget of approximately $80 million forldquoenforcement and compliancerdquo of ADCVD and has in fact had to request an additional $8million in the 2018 budget in order ldquoto develop factual information and legal justificationto self-initiate US antidumping duty (AD) and countervailing duty (CVD) investigationsrdquosuggesting that financial constraints have limited their capacity to do this in the past (Ross2018 p42)

7

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

Lastly this paper speaks to a broader literature on how international institutions and inter-national law can be helpful to national governments (Keohane 1984 Koremenos et al 2001Koremenos 2001 Kucik and Pelc 2016 Pelc 2016) Amongst other things this literature hasdiscussed the role that institutions can play in ldquotying the handsrdquo of governments to policiesthat might be resisted by domestic groups and has identified ldquoinformationrdquo as a potentialbenefit of such institutions However these discussions of information and institutions havefocused almost exclusively on how institutions can help to provide information about otherstates by for instance monitoring compliance

This paper posits a new reason why international institutions can be helpful namely thatthey allow governments to obtain private information about interest groups so as to betterallocate protection and other forms of political support Like the ldquotying handsrdquo argumentthis suggests that governments may use agreements with foreign countries as much to ad-dress their internal politics as their external relations Unlike these previous argumentsthis paper focuses on how institutions can be leveraged to make information transfer fromdomestic groups more credible rather than used as a device to make the governmentrsquos com-mitments to unpopular policies more credible

Reviewing the Trade Law

The Institutional LandscapeWhile many initially hoped that the World Trade Organization (WTO) - established in 1995- would become a single forum through which trade liberalization could be negotiated on abroadly multilateral basis this hope has largely failed to come to fruition from 1990-2010the number of Preferential Trade Agreements (PTAs) rose from about 70 to approximately300 (Baccini et al 2011) This explosion in the number of PTAs has led to the peculiar situ-ation where trade between countries is often governed by several overlapping institutionssuch that complainants may have to choose between forums when lodging a complaint

Moreover these forums can differ significantly in the way they resolve disputes For in-stance while both the North American Free Trade Agreement (NAFTA) and the WTO havedispute resolution proceedings (NAFTA Chapter 19 amp 20 and the DSU respectively) theformer tends to draw panels of those with domestic judicial expertise (this was developedunder pressure from the US Congress to ensure that panelists would have an in-depth fa-miliarity with US Administrative law) while the latter is weighted more towards those withexpertise in international trade law and economics (Howse 1998) This can lead to differ-ences in how similar cases would be adjudicated given that trade law cases often rest ontechnical determinations of ldquomaterial injuryrdquo and ldquodumpingrdquo that are necessarily made byeconomists and statisticians For instance in the somewhat unusual case of the US-CanadaSoftwood Lumber dispute where aspects of the dispute were brought to both the WTO and

5

NAFTA a WTO panel ruled that a threat of material injury to US industries was presentwhile a NAFTA panel ruled precisely the opposite (Pauwelyn 2006)

If forums differ in how they adjudicate disputes and if states have a choice about whichforum to use when pursuing a claim then ldquoforum-shoppingrdquo becomes inevitable wherebystates and firms choose between different forums strategically Furthermore the issue ofoverlapping institutions is likely to continue to increase given that the rapidly rising num-ber of PTAs increases the likelihood that any given potential claim will be governed by morethan one institution

Busch (2007) identifies several reasons why a state might prefer one forum over anotherincluding timeliness of dispute resolution available remedies etc and (as discussed earlier)the value of any informal precedent set (Busch 2007) This paper provides another firmsand interest groups may lobby states to file disputes at particular forums over others asa way of credibly signaling the strength of their legal cases For most of these disputeresolution forums country-level governments are the only actors with standing to pursuedisputes which sets up the informational problem that is this paperrsquos primary subject ofinquiry as states seek to determine which disputes to pursue but private firms and interestgroups possess the information required to best make those decisions Thus this paperrsquosmodel suggests that forum choice can be one way of resolving this information problem

Flexibility ProvisionsIn addition to providing means for governments to claim violations of an agreement tradeagreements generally allow states to temporarily suspend certain obligations under specificcircumstances These have been referred to by Koremenos et al as ldquoflexibility provisionsrdquoie those provisions that allow members to ldquoescaperdquo temporarily from obligations under theagreement usually in response to some unanticipated shock (Koremenos et al 2001) ldquoSafe-guard measuresrdquo as they are often called exist under both NAFTA and the WTO NAFTAhas measures under Chapter 8 and the WTO has measures under GATT Article XIX (Tre-bilcock and Howse 2005) Moreover there are certain measures that have been interpretedas being de facto flexibility provisions even if they appear in principle to target unfair tradepractices for instance it has been argued that anti-dumping (AD) laws are a form of flex-ibility provision (Kucik and Reinhardt 2008) Countervailing duties (CVDs) are the mainother form of trade remedy and are formally a way of responding to prohibited or actionablesubsidies of other countries but they may be used instead as a flexibility provision (partic-ularly given that states have the option to challenge subsidies via the DSU)

Worth noting is that these trade remedies also differ in the degree to which they targetspecific countries or are applied more broadly CVDs apply to the countries accused of im-plementing subsidies and anti-dumping duties apply to the countries accused of dumpingboth are therefore explicitly discriminatory Safeguard measures in contrast are requiredto apply in a nondiscriminatory manner to all countries However Bown (2013) suggests

6

that in many cases appropriately targeted antidumping duties will have economic effectsthat are very similar to those of a broader safeguard measure

For all forms of trade remedies firms and interest groups representing them are likely tohave private information about the strength of their case All of these provisions usuallyrequire evidence of ldquoinjuryrdquo an empirical claim that necessitates showing both that thefirm is under economic duress and that this duress is causally related to competition frominternational trade Anti-dumping provisions also require evidence that imported goodsare being sold in the complainantrsquos country at a lower price than they are being sold inthe exporterrsquos country or that the goods are being sold below cost (Trebilcock and Howse2005) Similarly CVDs require evidence that the exporting country is subsidizing theirexport industries

The ArgumentFirms naturally have information about material injury and prices of competing goods aspart of their regular business practices while governments will often either not have accessto it (as with internal research conducted by the firm) or will be unwilling to expend theresources required to collect it Firms will also often have private information about thecompliance of other countries with their international legal obligations given that thesefirms have an incentive to invest more resources in monitoring other countries for viola-tions and will experience the consequences of violations directly when it affects their bal-ance sheet Indeed Brutger (2017) conducted interviews with a number of trade officialsand trade lawyers who had been involved in prominent trade disputes and found that thesetrade experts believed that the firms involved possessed this kind of private information(Brutger 2017 p 20)

This claim is reinforced by the fact that the agencies tasked with administering trade laware financially constrained for instance the Office of United States Trade Representative(USTR) which is tasked with pursuing any claims via the DSU has a budget of approxi-mately $50 million much of which is allocated towards funding trade negotiations insteadof monitoring domestic firm performance (USTR Budget Report 2015) Similarly the UnitedStates International Trade Commission (USITC) which is tasked with investigating injuryin AD CVD and safeguards cases has an annual budget of approximately $85 million ofwhich approximately $25 million is spent on the administration of trade remedies (USITCBudget Report 2016) The Commerce Department which is responsible for making deter-minations of dumping or unfair subsidies has a budget of approximately $80 million forldquoenforcement and compliancerdquo of ADCVD and has in fact had to request an additional $8million in the 2018 budget in order ldquoto develop factual information and legal justificationto self-initiate US antidumping duty (AD) and countervailing duty (CVD) investigationsrdquosuggesting that financial constraints have limited their capacity to do this in the past (Ross2018 p42)

7

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

NAFTA a WTO panel ruled that a threat of material injury to US industries was presentwhile a NAFTA panel ruled precisely the opposite (Pauwelyn 2006)

If forums differ in how they adjudicate disputes and if states have a choice about whichforum to use when pursuing a claim then ldquoforum-shoppingrdquo becomes inevitable wherebystates and firms choose between different forums strategically Furthermore the issue ofoverlapping institutions is likely to continue to increase given that the rapidly rising num-ber of PTAs increases the likelihood that any given potential claim will be governed by morethan one institution

Busch (2007) identifies several reasons why a state might prefer one forum over anotherincluding timeliness of dispute resolution available remedies etc and (as discussed earlier)the value of any informal precedent set (Busch 2007) This paper provides another firmsand interest groups may lobby states to file disputes at particular forums over others asa way of credibly signaling the strength of their legal cases For most of these disputeresolution forums country-level governments are the only actors with standing to pursuedisputes which sets up the informational problem that is this paperrsquos primary subject ofinquiry as states seek to determine which disputes to pursue but private firms and interestgroups possess the information required to best make those decisions Thus this paperrsquosmodel suggests that forum choice can be one way of resolving this information problem

Flexibility ProvisionsIn addition to providing means for governments to claim violations of an agreement tradeagreements generally allow states to temporarily suspend certain obligations under specificcircumstances These have been referred to by Koremenos et al as ldquoflexibility provisionsrdquoie those provisions that allow members to ldquoescaperdquo temporarily from obligations under theagreement usually in response to some unanticipated shock (Koremenos et al 2001) ldquoSafe-guard measuresrdquo as they are often called exist under both NAFTA and the WTO NAFTAhas measures under Chapter 8 and the WTO has measures under GATT Article XIX (Tre-bilcock and Howse 2005) Moreover there are certain measures that have been interpretedas being de facto flexibility provisions even if they appear in principle to target unfair tradepractices for instance it has been argued that anti-dumping (AD) laws are a form of flex-ibility provision (Kucik and Reinhardt 2008) Countervailing duties (CVDs) are the mainother form of trade remedy and are formally a way of responding to prohibited or actionablesubsidies of other countries but they may be used instead as a flexibility provision (partic-ularly given that states have the option to challenge subsidies via the DSU)

Worth noting is that these trade remedies also differ in the degree to which they targetspecific countries or are applied more broadly CVDs apply to the countries accused of im-plementing subsidies and anti-dumping duties apply to the countries accused of dumpingboth are therefore explicitly discriminatory Safeguard measures in contrast are requiredto apply in a nondiscriminatory manner to all countries However Bown (2013) suggests

6

that in many cases appropriately targeted antidumping duties will have economic effectsthat are very similar to those of a broader safeguard measure

For all forms of trade remedies firms and interest groups representing them are likely tohave private information about the strength of their case All of these provisions usuallyrequire evidence of ldquoinjuryrdquo an empirical claim that necessitates showing both that thefirm is under economic duress and that this duress is causally related to competition frominternational trade Anti-dumping provisions also require evidence that imported goodsare being sold in the complainantrsquos country at a lower price than they are being sold inthe exporterrsquos country or that the goods are being sold below cost (Trebilcock and Howse2005) Similarly CVDs require evidence that the exporting country is subsidizing theirexport industries

The ArgumentFirms naturally have information about material injury and prices of competing goods aspart of their regular business practices while governments will often either not have accessto it (as with internal research conducted by the firm) or will be unwilling to expend theresources required to collect it Firms will also often have private information about thecompliance of other countries with their international legal obligations given that thesefirms have an incentive to invest more resources in monitoring other countries for viola-tions and will experience the consequences of violations directly when it affects their bal-ance sheet Indeed Brutger (2017) conducted interviews with a number of trade officialsand trade lawyers who had been involved in prominent trade disputes and found that thesetrade experts believed that the firms involved possessed this kind of private information(Brutger 2017 p 20)

This claim is reinforced by the fact that the agencies tasked with administering trade laware financially constrained for instance the Office of United States Trade Representative(USTR) which is tasked with pursuing any claims via the DSU has a budget of approxi-mately $50 million much of which is allocated towards funding trade negotiations insteadof monitoring domestic firm performance (USTR Budget Report 2015) Similarly the UnitedStates International Trade Commission (USITC) which is tasked with investigating injuryin AD CVD and safeguards cases has an annual budget of approximately $85 million ofwhich approximately $25 million is spent on the administration of trade remedies (USITCBudget Report 2016) The Commerce Department which is responsible for making deter-minations of dumping or unfair subsidies has a budget of approximately $80 million forldquoenforcement and compliancerdquo of ADCVD and has in fact had to request an additional $8million in the 2018 budget in order ldquoto develop factual information and legal justificationto self-initiate US antidumping duty (AD) and countervailing duty (CVD) investigationsrdquosuggesting that financial constraints have limited their capacity to do this in the past (Ross2018 p42)

7

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

that in many cases appropriately targeted antidumping duties will have economic effectsthat are very similar to those of a broader safeguard measure

For all forms of trade remedies firms and interest groups representing them are likely tohave private information about the strength of their case All of these provisions usuallyrequire evidence of ldquoinjuryrdquo an empirical claim that necessitates showing both that thefirm is under economic duress and that this duress is causally related to competition frominternational trade Anti-dumping provisions also require evidence that imported goodsare being sold in the complainantrsquos country at a lower price than they are being sold inthe exporterrsquos country or that the goods are being sold below cost (Trebilcock and Howse2005) Similarly CVDs require evidence that the exporting country is subsidizing theirexport industries

The ArgumentFirms naturally have information about material injury and prices of competing goods aspart of their regular business practices while governments will often either not have accessto it (as with internal research conducted by the firm) or will be unwilling to expend theresources required to collect it Firms will also often have private information about thecompliance of other countries with their international legal obligations given that thesefirms have an incentive to invest more resources in monitoring other countries for viola-tions and will experience the consequences of violations directly when it affects their bal-ance sheet Indeed Brutger (2017) conducted interviews with a number of trade officialsand trade lawyers who had been involved in prominent trade disputes and found that thesetrade experts believed that the firms involved possessed this kind of private information(Brutger 2017 p 20)

This claim is reinforced by the fact that the agencies tasked with administering trade laware financially constrained for instance the Office of United States Trade Representative(USTR) which is tasked with pursuing any claims via the DSU has a budget of approxi-mately $50 million much of which is allocated towards funding trade negotiations insteadof monitoring domestic firm performance (USTR Budget Report 2015) Similarly the UnitedStates International Trade Commission (USITC) which is tasked with investigating injuryin AD CVD and safeguards cases has an annual budget of approximately $85 million ofwhich approximately $25 million is spent on the administration of trade remedies (USITCBudget Report 2016) The Commerce Department which is responsible for making deter-minations of dumping or unfair subsidies has a budget of approximately $80 million forldquoenforcement and compliancerdquo of ADCVD and has in fact had to request an additional $8million in the 2018 budget in order ldquoto develop factual information and legal justificationto self-initiate US antidumping duty (AD) and countervailing duty (CVD) investigationsrdquosuggesting that financial constraints have limited their capacity to do this in the past (Ross2018 p42)

7

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

As a consequence firms need to signal their private information to governments if govern-ments are to be able to select between cases when choosing which disputes to pursue orwhom to grant protection via a trade remedy However as firms benefit unconditionallyfrom these government actions they always have an incentive to say that their injury ishigh and that their case is otherwise strong While firms can (and generally do) disclosedocumentation to support their case there is no way to ensure that this information is notbeing presented in a biased fashion with certain information being withheld or presentedin a manner so as to paint the firmrsquos case in the most positive light

Governments likely do not want to invest reputational capital and financial resources inpursuing cases that have a high probability of failure Moreover governments would oftenprefer not to protect those firms which have not experienced material injury given that pro-tecting such ldquoweakly injuredrdquo groups provides fewer political gains while imposing costs onpotentially influential downstream competitors as well as aggregate welfare costs that maybe politically important Put differently distributing trade protection is a ldquonegative sumgoodrdquo giving trade protection to some groups reduces the total amount of resources that canbe distributed to all parties Thus absent credible information transfer from firms to thegovernment governments may choose not to protect any groups at all or not to pursue anytrade disputes

The argument of this paper is that firms may choose to pursue support via institutions ormechanisms that give them a lower probability of success or that are more costly if thisallows them to credibly signal private information about the strength of their cases In ef-fect a firm may choose a ldquohard testrdquo of the strength of their case if it allows governments todistinguish between that group and others that would be unwilling to select the hard testdue to their even lower likelihood of ldquopassingrdquo it

This works if the dispute settlement process can reveal information about a firmrsquos typewhich I argue is the case in practice for two reasons First if a dispute is initiated (either bya domestic firm or by a foreign firm challenging a trade remedy) the case will often reduce totechnical determinations of fact by economists and statisticians often with further factualdisclosures to the panel by firms on the opposite side of the case If this is an independentsignal that is positively correlated with the strength of the case this can reveal informa-tion about a firmrsquos type Second for trade remedies if we assume that governments willonly initiate disputes in which they think they have a chance of succeeding then disputeinitiation reflects the aggregated information of both foreign states and foreign firms Thusweak cases are both more likely to be challenged and (not independently) more likely to beoverturned when challenged The combination of these two things reflects the ldquoprobabilityof successrdquo when a government implements a trade remedy in a way that is directly relatedto a firmrsquos type

8

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

Not only does this screening mechanism help with legal case selection (helping to explain theextremely high success rate of WTO complainants) but it allows governments to separatethe groups that they would want to protect from those that they would not even if a legalinstitution regulating trade did not exist Thus this paper provides another argument forwhy governments may want to have independent institutions that regulate and enforcetrade policy namely because these institutions may help them obtain useful informationthat then allows them to better satisfy their domestic political objective function

Firms and Interest Groups as the Primary ActorsThis paper treats firms and interest groups as the primary actors involved in pursuing traderemedies and initiating disputes with governments playing the more passive role of simplyaccepting or rejecting the overtures by these private groups It is thus worth consideringthe degree to which this characterization accords with the empirical evidence

For trade remedies there is a significant amount of evidence supporting this descriptionof the process In the United States the vast majority of trade remedy cases have beeninitiated by workers firms or industry associations 2003 out of 2022 since 1980 to be pre-cise or approximately 9912 In these trade remedy cases governments either formallydetermine whether to pursue a case once a determination has been made by the USITC(as with safeguards) or have a significant degree of influence over the process by way ofthe Commerce Department ruling on dumping or prohibitedactionable subsidies by foreigncountries Thus the process looks exactly as described firms and interest groups petitionfor protection via the trade remedy of their choice and governments choose whether or notto accept their request subject to a plausibility check of injury by the USITC

The issue is a little more subtle when it comes to dispute settlement given that usuallystates are the only actors with standing to initiate disputes However even in these casesthe evidence suggests that governments generally only initiate disputes in response to lob-bying from domestic firms and interest groups Ryu and Stone (2018) provide statisticalevidence that this is the case demonstrating that political contributions by affected firmsincrease substantially prior to the initiation of disputes Betz (2018) also provides examplesof interest group driven disputes (Betz 2018 p 641-642)

Model

Set-UpIn this model there are two players Government (G) and an interest group (I) The inter-est group is lobbying for trade protection and has private information about their degree ofinjury θ with interest group types distributed θ sim f (θ) with support on Θ= [0w] Assume

2Chad Bown in The Washington Post April 21 2018

9

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

for simplicity that f (θ) is absolutely continuous Governments want to protect the moststrongly injured groups (ie those with high θ) but not weakly injured groups (those withlow θ) Thus interest groups always have an incentive to tell the government that they areseverely injured in order to try to obtain the most protection This can be generalized bythinking of θ as more broadly representing the strength of a grouprsquos legal case

Now suppose that I can choose between two institutional mechanisms when pursuing pro-tection and one of those mechanisms gives a strictly lower probability of success than an-other For expository purposes we can call the more stringent mechanism DS and the lessstringent mechanism CVD this reflects a relatively clean example of screening whereinan interest group facing subsidies can either lobby the government to pursue a legal claimagainst another country via dispute settlement (DS) or file for CVDs via the USITC andDepartment of Commerce However there are many instruments and institutions that in-terest groups can choose from so this is only one example of such a selection process otherswill be elaborated upon later in the paper The government then observes which request ismade and decides whether to grant it

If protection is pursued via DS then a panel rules in favor of I with probability πD(θ) Ifprotection is pursued via CVD then the CVD stands (ie is either not challenged or is chal-lenged and upheld) with probability πC(θ) Both partπD

partθgt 0 and partπC

partθgt 0 which captures the

idea that higher types (ie those with stronger cases) have a higher chance of success viaeither legal mechanism but πD(θ)ltπC(θ) forallθ isinΘ to reflect the fact that dispute settlementis the more ldquostringentrdquo of the two mechanisms3

Government obtains g(θ) from protecting an interest group with partgpartθ

gt 0 representing thatthey prefer to provide protection to higher types Note that g(θ) can be either positive ornegative Government obtains a payoff of zero from either rejecting an interest grouprsquos re-quest or if the panel rejects a claim or the CVD is overturned (leading to no protection)

This approach abstracts from any other characteristics - private benefits of lobbying contri-butions political geography etc - that might matter to Government but are not related toan interest grouprsquos type A straightforward extension would include some other parameterin this function that could shift g(θ) up or down depending on these characteristics Thiswould have the effect of creating different θ cutoffs for interest groups who differ on thesecharacteristics but would not fundamentally change the results this model should be takenas providing an analysis of what happens holding these characteristics fixed

Interest groups obtain payoffs of v(θ) if they receive protection with partvpartθ

gt 0 reflecting thefact that protection is more important to them if they are strongly injured They experiencefixed costs cD gt 0 if Government pursues their claim via DS and fixed costs cC gt 0 if Gov-

3Note that the existence of derivatives of the probability functions implies those functions are continuousacross their domains in θ In this paper all functions of θ will be continuous

10

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

12v(θprimeprime)πD(θprimeprime)minus cD g(θprimeprime)πD(θprimeprime) Accept

00 Re ject

DS 2

00Re ject

v(θprimeprime)πC(θprimeprime)minus cC g(θprimeprime)πC(θprimeprime)AcceptCV D

θprimeprime

12

v(θprime)πD(θprime)minus cD g(θprime)πD(θprime) Accept

00 Re jectDS 2

00Re ject

v(θprime)πC(θprime)minus cC g(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 1 Institutional Selection Model

ernment approves their claim via CVD So pursuing a claim is costly for the group and mayonly be profitable if the chance of succeeding is sufficiently high The model also assumesthat cD ge cC for simplicity Industries obtain a payoff of zero if their request is denied inany case This leads to the game tree found in Figure 1

As can be seen in Figure 1 the firm observes their type as drawn from the support Θand then choose whether to pursue DS or CV D θprimeθprimeprime are arbitrary draws from the typedistribution f (θ) Government does not observe type directly but observes whether or notDS or CV D was pursued before choosing whether or not to Accept or Reject the claimIf they Reject in any case the payoff is zero to both parties If they Accept payoffs aredetermined as outlined above as a function of I rsquos type

Equilibrium Analysis PoolingThroughout the equilibrium analysis I assume that pursuing a claim for protection viaCVD is profitable for all types (ie v(θ)minus cC gt 0 forallθ isin Θ) A simple extension to the modelwould allow a group to choose not to pursue a claim at all and receive a payoff of 0 or topursue unilateral protection outside of any international legal institution and avoid any le-gal costs (unilateral protection can be treated as a degenerate case where πC(θ) = 1 forallθ isinΘand cC = 0) However this modelrsquos primary goal is to address the selection process betweenmechanisms when at least one of them is profitable to both types One can simply keep inmind that unilateral protection or not pursuing protection is always an option and note thatthis model applies to instances in which all groups can gain by having their claim pursuedvia at least one mechanism

I also assume for simplicity that there existθ isinΘ such that v(θ)minus cD ge 0 In words there existat least some interest groups that would prefer to have a claim via DS pursued than to havethat claim rejected and receive a payoff of zero Absent this DS would be strictly dominated

11

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

by CVD for all interest group types in which case the model could not be expected to providemuch insight

Several pooling equilibria are possible in this model For instance it is possible to sustainequilibria where G protects all groups and all groups pursue protection via CVD irrespectiveof type This is described in the following proposition

Proposition 1 Pooling equilibria exist where all interest groups pursue protection via CVDand Government accepts all claims (call this conjecture ldquoALLCVD) if the following conditionholds

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

Proof in appendix Intuitively the above condition means that the percentage of interestgroups that are strongly injured is high enough that it is still better for Government topursue protection even if there is no way to distinguish between groups of different typesThis allows for equlibria in which all group types pursue the least stringent institution andGovernment chooses to pursue all claims As this is I rsquos best outcome (regardless of type)these equlibria can be sustained by any distribution of off-equilibrium path beliefs

It is also possible to have pooling where all types pursue protection via DS under a morerestrictive set of conditions This is described in the following proposition

Proposition 2 A pooling equilibrium exists where all interest groups pursue protectionvia DS and Government accepts all claims (call this conjecture ldquoALLDS) if the followingconditions hold

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0 (1)

v(θ)minus cD gt 0forallθ isinΘ (2)

and Government must have off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0 (3)

Proof in appendix Taking these conditions in turn (1) simply replicates the pooling condi-tion from Proposition 1 but for DS ensuring that that the type distribution is sufficientlypositively weighted to make accepting all claims better than rejecting all of them (2) en-sures that all interest group types prefer to pursue a claim via DS if it will be acceptedrather than choosing DS and having their claim rejected (resulting in a payoff of 0) (3)ensures that upon observing a group pursuing CVD Government believes that it is likelyenough that the deviating type is weakly injured that they do not wish to protect that de-viator (3) is needed to ensure that interest groups are not tempted to switch to the lessstringent institution

12

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

One can also obtain pooling equilibria where all types choose either CVD or DS and Govern-ment never protects This is described in the following proposition

Proposition 3 Pooling equilibria exist where either (i) all interest groups pursue protectionvia DS and are rejected by Government

E[g(θ)|ALLCV D]=int w

0g(θ)πC(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πD(q)p(q)dq le 0

Or where (ii) all interest groups pursue protection via CVD and are rejected by Governmentif

E[g(θ)|ALLDS]=int w

0g(θ)πD(θ) f (θ)dθ le 0

with off-equilibrium path beliefs q sim p(q) such thatint w

0g(q)πC(q)p(q)dq le 0

Proof in appendix This proposition reverses the relevant pooling conditions from Proposi-tions 1 and 2 to ensure that Government does not have an incentive to Accept upon observinga claim then sets off the equilibrium path beliefs such that government also prefers not toprotect deviators who pursue a claim via the other institution Thus in these equilibria theinstitution is entirely unused and Government does not allocate protection to any groups

All of these pooling equilibria share the characteristic that they are uninformative in thesense that if these equilibria apply the institutions provide no new information to Govern-ment about the types of petitioning interest groups While these equilibria are importantand likely apply in certain cases this paperrsquos added value is primarily in describing the sit-uations in which institutions can be leveraged by Government to extract information fromgroups that would otherwise not be revealed

Separating EquilibriaGiven this orientation this paperrsquos primary focus will be on separating equilibria ie thoseinstances where interest groupsrsquo selection of various institutions can help to reveal informa-tion about their type To find these we proceed as follows

In any separating equilibrium we can establish a cutpoint w above which a group selectsDS and below which they select CVD The only pure strategy that governments can adoptthat would induce separation is

σG = (Accept|DSRe ject|CV D)=

Accept if I chooses DSRe ject if I chooses CVD

13

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

Otherwise CVD weakly dominates for the interest group (given that πD(θ)ltπC(θ)) Statedintuitively if Government adopts any strategy other than accepting only the claims pursuedvia the more stringent institution groups will never have an incentive to choose the ldquohardertestrdquo because Government would not be providing them any benefits over choosing the ldquoeasytestrdquo

Assuming Government adopts this strategy we can now consider the interest grouprsquos incen-tives conditional on this denoting the grouprsquos expected utility function UI We derive thecutpoint w by determining where UI(DS) = UI(CV D) with UI(CV D) = 0 (given that theirclaim will simply be rejected by the government) and UI(DS)= v(θ)πD(θ)minuscD Setting theseequal to each other we get

UI(DS)=UI(CV D)= v(θ)πD(θ)minus cD = 0harrπD(θ)v(θ)= cD

This implicitly defines a cutpoint w if such a cutpoint exists To ensure its existence simplyassume that existθ isinΘ such that UI(DS)ltUI(CV D) and existθ isinΘ such that UI(DS)gtUI(CV D)This implies that there are some types that are low enough that it is not incentive compati-ble for them to choose DS and some high enough that it is incentive compatible a necessarycondition for the analysis of this paper to have any traction

Given this we can define Governmentrsquos utility function as UG(Re ject)= 0 and

UG(Accept|DS)=int w

w[g(θ)πD(θ)+ (1minusπD(θ))(0)] f (θ)dθ =

int w

wg(θ)πD(θ) f (θ)dθ

UG(Accept|CV D)=int w

0[g(θ)πC(θ)+ (1minusπC(θ))(0)] f (θ)dθ =

int w

0g(θ)πC(θ) f (θ)dθ

This allows us to construct a separating equilibrium where the strongest types select intopursuing claims via the more stringent institution and the weaker types select into pursu-ing claims via the less stringent institution with Government accepting all claims via themore stringent institution and rejecting all claims via the less stringent institution This issummarized in the following proposition

Proposition 4 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

σG =

Accept if I pursues DSRe ject if I pursues CVD

So long asUG(Accept|DS)ge 0geUG(Accept|CV D)

14

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

Proof follows immediately from preceding discussion Consequently in this equilibrium theldquohard testrdquo of the DS allows the Government to screen between strongly and weakly injuredgroups This holds so long as Government prefer to Accept the claims by types screened tothe harder test but prefer to Reject the claims by types screened to the weaker test

However it is worth noting that depending on the position of the cutpoint it is likely thatGovernment will end up accepting some claims that it would prefer to reject or rejectingsome claims that it would prefer to accept This is summarized in the following corollary

Corollary 1 With the exception of the case where g(w) = 0 in any separating equilibriumGovernment will either protect some types that they would prefer not to (ie there will besome subset ΘL sub [ww] such that forallθL isin ΘL g(θL) lt 0) OR Government will fail to protectsome types that they would prefer to (there will be some subset ΘH sub [0 w] such that forallθH isinΘH g(θH)gt 0)

Proof in appendix Thus while institutions allow for some degree of strategic informationtransfer in the case of a separating equilibrium there will almost always be some informa-tion that Government would prefer to have that will remain unrevealed

A Degenerate Case Considering Institutional DesignAs mentioned earlier a degenerate case of the model is where an interest group is choosingbetween pursuing a claim via an international institution or via unilateral protection fromthe government which in this model would be where

πC(θ)= 1 forallθ isinΘ and cC = 0

The mathematics of this situation are simply what is obtained when one applies these valuesto the model but the substantive interpretation is different In this case we are not talkingabout selecting between institutional mechanisms but about why a firm might choose topursue a claim via an international institution at all given that there is a chance that thisclaim might be rejected The model suggests that even if governments would most preferto protect all strongly injured firms without having to deal with the uncertainty of goingthrough an international institution groups may still pursue protection via these institu-tions if it can help screen between groups Thus this degenerate case focuses onersquos attentionon a potentially important reason why governments might agree to institutions with exter-nal arbitration namely that such arbitration (or the threat of it) may provide governmentswith information that allows them to better select groups to protect in order to satisfy theirdomestic objective function

This can also help to provide insight into the important question of institutional designFirst we can consider Governmentrsquos welfare under different equilibria in the degeneratecase A pooling equilibrium on unilateral protection is what you would have if there was

15

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

no international institution through which to screen cases and the outcome in this scenariowould hinge on whether you have int w

0g(θ) f (θ)dθ ge 0

With this condition following from Proposition 1 If this condition holds all groups wouldapply for protection and receive it If it does not all groups would apply for protectionand would be rejected The utility to Government in a world without institutions that canindependently evaluate and overturn claims is thus

maxint w

0g(θ) f (θ)dθ0

In contrast under the separating equilibrium groups with higher types self-select into theinstitutionallegal mechanism for pursuing protection while others pursue unilateral pro-tection and are rejected Thus the utility to government isint w

wg(θ)πD(θ) f (θ)dθ

Which is greater than zero by the conditions outlined in Proposition 4 and thus higher thanwhat Government gets under the pooling equilibrium in which no firm is protected Thisleads to the following proposition

Proposition 5 Government is better off in a separating equilibrium induced by an interna-tional institution than they would be if an institution did not exist wheneverint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

Proof follows immediately from preceding discussion We can further unpack the differentcomponents of this condition to get some insight into when having an institution will bepreferred by Government We can rewrite the condition as followsint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ+

int w

wg(θ) f (θ)dθ

harrint w

wg(θ)πD(θ) f (θ)dθ︸ ︷︷ ︸

α

minusint w

wg(θ) f (θ)dθ︸ ︷︷ ︸

β

gtint w

0g(θ) f (θ)dθ︸ ︷︷ ︸

∆

First note that it is not immediately clear whether α or β is greater since it may be thecase that g(θ) is negative for some values of θ isin [ww] (see Corollary 1) This can be thoughtof as an added benefit to Government perhaps there are some cases they end up pursuingeven in a separating equilibrium which they would rather have an international institution

16

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

overturn because such cases were just ldquomixed uprdquo with the good cases in the separatingequilibrium

However for simplicity letrsquos consider a situation where g(θ) gt 0 forallθ isin [ww] In this caseαltβ so αminusβlt 0 However it is also the case that ∆lt 0 by the implications of the construc-tion of the separating equilibrium described in Proposition 4 otherwise Government wouldnot reject upon observing a request for unilateral protection Thus the trade-off here forgovernment becomes the following does having an institution lead to enough ldquogoodrdquo casesbeing overturned that itrsquos worse than the cost of giving protection to more of the ldquobadrdquo types

This gives some rationale for why governments might set up international institutions withbinding arbitration procedures that assess a lot of information to make technical determina-tions of fact but which also tend to skew towards leaving protectionist claims unchallengedThe flexibility to protect certain groups is important to governments (the no protection equi-librium is worse than any screening equilibrium) but it is also usually the case that pro-tecting all groups makes Government worse off than if it was able to extract some degree ofprivate information from those groups

While the model could help explain the specific features that governments choose to in-clude when designing institutions it is also not necessary to assume that this rationale wasat work at the institutional design stage for it to be useful in understanding how the in-stitutions actually function For instance while provisions allowing groups to petition forrelief via trade remedies have existed in US law for many decades their administration hasshifted since the 1950s from a compensation-based efficient breach structure to one of meet-ing criteria in order to invoke such trade remedies (Pelc 2011) We would expect this kindof shift to be favored by governments if they can benefit from leveraging this threat of inter-national arbitration to induce groups to separate even if it means that some of their traderemedies end up getting overturned Furthermore the persistence of trade remedies as atool used by governments and their inclusion in new PTAs suggests that trade remediescontinue to be an equilibrium strategy for governments and this paperrsquos model can help toexplain why

Shifts in Protectionist SentimentSince the election of Donald Trump in 2016 there has been a significant shift in the waythat trade remedies and trade institutions have been used by the US government in a waythat seems largely inconsistent with prior practice and which threatens to upend the sta-bility of international institutions like the WTO4 While at first glance these changes mightseem inconsistent with the screening story outlined in this paper the model can in fact pro-vide significant insight into the situation by treating the changes as the result of an upward

4The use of Section 232 security exemptions to protect steel aluminum and (possibly) automobiles hasbeen singled out by many as a particular threat to the stability of the international trade regime

17

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

shift in protectionist sentiment

Specifically the model would treat a significant upward shift in protectionist sentiment inthe government as a shift upward in g(θ) where this is defined as a move to some g+(θ)such that

forallθ isinΘ g+(θ)ge g(θ)

andexistΘprime subΘ with positive Lebesgue measure such that g+(θprime)gt g(θprime)forallθprime isinΘprime

This just means that there exists some non-measure-zero subset of the type space suchthat g(θ) has increased over that subset intuitively Government gets higher utility forsome potential types of firms under g+(θ) then under g(θ)5 Thus we can consider thecomparative statics exercise of assessing what happens in the model with this upward shiftin g(θ) To start we have the following proposition

Proposition 6 A shift upward in g(θ) makes a pooling outcome where every firm gets pro-tection more likely by increasing the likelihood that the conditions of Proposition 1 or Propo-sition 2 will be met

Proof in appendix Thus the modelrsquos predictions fit well with what we have observed dur-ing the Trump administration ie a situation where virtually every petition by an interestgroup has been pursued by the government after an upwards shift in protectionist senti-ment6

Perhaps even more importantly we can also consider what impact an upward shift in protec-tionist sentiment would have on the value placed by the government on the institution itselfRecall from Proposition 5 that a government prefers institutional regulation to a world ofunilateral protection if int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ

If g(θ) increases the downside risk entailed in πD(θ) where institutions reject cases deter-mined by the government to be worth pursuing becomes more important than the benefitsobtained from screening out low types - particularly given that fewer types (if any) are likelyto fall below the threshold where g(θ)lt 0 This leads to the following proposition

Proposition 7 An upward shift in protectionist sentiment makes Government more likelyto prefer the outcome obtained without an institution to an institution-induced separatingequilibrium

5This is essentially statewise dominance of g+(θ) over g(θ)6As examples historically safeguard measures have been pursued by US governments in less than 50

of the instances in which the USITC gave the government the opportunity to impose them and Section 232has been used only twice out of 14 investigations (Bown and Joseph 2017) As of the time this was writtenthe Trump administration had pursued both safeguards cases that had gone through the USITC (on washingmachines and solar panels) had implemented Section 232 tariffs on steel and aluminum and had launched aSection 232 investigation into autos and auto parts

18

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

12vD(θprimeprime)πD(θprimeprime)minus cD gD(θprimeprime)πD(θprimeprime)

Accept

00 Re ject

DS 2

00Re ject

vC(θprimeprime)πC(θprimeprime)minus cC gC(θprimeprime)πC(θprimeprime)Accept

CV D

θprimeprime

12

vD(θprime)πD(θprime)minus cD gD(θprime)πD(θprime)Accept

00 Re jectDS 2

00Re ject

vC(θprime)πC(θprime)minus cC gC(θprime)πC(θprime)Accept

CV D

θprime

N θ isinΘ

Figure 2 Institutional Selection Model with Varying Benefits

Proof in appendix In other words an upward shift in protectionist sentiment makes thegovernment less likely to value the institution and consequently more likely to risk itscollapse because the revealed information becomes relatively less valuable

Varying Benefits Across InstitutionsMechanismsUp until this point the model has assumed that the benefits to both players from protec-tion are the same across institutions This is useful for considering cases where the mainreason for selecting a more stringent institution is the signaling value but it is also worthconsidering what the impact of allowing these valuations to vary across institutions wouldbe given that this more closely accords with reality

In particular letrsquos consider a case where vD(θ) and vC(θ) are different with vD(θ) gt vC(θ)This says that an interest group values a positive outcome from DS more highly than theyvalue a positive outcome from CVD This could describe a case where an interestrsquos grouprsquospreferred outcome is to have a subsidy overturned via the WTO but they may settle forCVDs because they have a higher likelihood of it being upheld

For government we can assign the opposite preferences ie gD(θ) lt gC(θ) This is to cap-ture the idea that Government may prefer to provide lower types with more limited forms ofprotection like CVDs but would be unwilling to pursue disputes on their behalf via inter-national institutions This way of modeling Government creates additional strategic tensionbetween Government and the interest group given that Government would prefer to inducetypes pursuing DS to choose CVDs instead however since the interest group is the firstmover this has few implications for which equilibria survive sequential rationality The(slightly) modified game tree is shown in Figure 2

In this version of the model pooling equilibria are constructed in a similar fashion as before

19

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

(in Propositions 1-3) with minor adjustments to the pooling conditions and restrictions onoff equilibrium path beliefs to account for the varying benefits across institutions Separat-ing equilibria however can be substantively different

One possible separating equilibrium is constructed almost identically to Proposition 4 ieif Government chooses a strategy of σG = (Accept|DSRe ject|CV D) the analysis is largelythe same though the position of the cutpoint might change This is because despite thefact that gC(θ)gt gD(θ) forallθ isinΘ it is still possible that in a separating equilibrium the typesbeing screened to CVDs are low enough that Government prefers not to pursue their claims7

However a new separating equilibrium also becomes possible with varying benefits onein which Governmentrsquos strategy is σG = (Accept|DS Accept|CV D) and groups separatebased on their types because of the distinct benefits to the different institutions In thiscase the cutpoint is defined implicitly by the type where UI(DS)=UI(CV D) or where

vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC

However this cutpoint may or may not exist depending on the shape of the probability andbenefit functions We can ensure existence with some additional assumptions outlined inthe following lemma

Lemma 1 A cutpoint w exists where vD(θ)πD(θ)minus cD = vC(θ)πC(θ)minus cC if the following con-ditions hold

cD = cC

limθrarrw

πD(θ)= 1

limθrarrw

vC(θ) 6= limθrarrw

vD(θ)

partvD(θ)πD(θ)partθ

gt partvC(θ)πC(θ)partθ

Proof in appendix Substantively this means that the absolute highest types have a nearcertain probability of their claims being upheld the costs of pursuing claims across institu-tions is the same and the expected utility of pursuing DS is increasing faster in θ than theexpected utility of pursuing CV D

Thus under these assumptions it makes sense for at least some group types above a par-ticular threshold to select into DS over CVD We can thus establish a new separating equi-librium

Proposition 8 A separating equilibrium exists where

σI =

DS if θ gt wEither DS or CVD if θ = wCV D if θ lt w

7Formally this is the case ifint w

0 gC(θ)πC(θ) f (θ)dθ le 0

20

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

σG =

Accept if I pursues DSAccept if I pursues CVD

So long as the conditions in Lemma 1 hold8 in addition to the following conditionsint w

0gC(θ)πC(θ) f (θ)dθ ge 0

int w

wgD(θ)πD(θ) f (θ)dθ ge 0

Proof in appendix Intuitively this simply means that Government prefers to Accept ratherthan Reject claims whether or not an interest group is choosing to pursue that claim via CVDor DS This becomes possible because of the varying benefits in this case they are willing toallocate different levels of protection to different groups depending on how strongly injuredthey are

What does all this mean substantively If different institutionsmechanisms provide differ-ent benefits to interest groups and governments it becomes possible to sustain an equilib-rium that reveals information about group types in which no claims via either mechanismare rejected In some cases this may more closely accord with what we observe most claimsfor CVDs are approved by governments9 and requests for dispute settlement are often pur-sued but it is not an identical set of groups that applies for each

Allowing Pursuit of Multiple Mechanisms Simultaneously

We can also consider the case where benefits vary across institutionsmechanisms and in-terest groups can pursue more than one mechanism at the same time This can revealinformation about I rsquos type in much the same way as in the separating equilibria of previousversions of the model but now we obtain different cutpoints for each institution ie theyare implicitly defined by

vC(θ)πC(θ)= cC

vD(θ)πD(θ)= cD

If we call the cutpoint under CVD wC and the cutpoint under DS wD Governmentrsquos bestresponse to this cutpoint strategy is as follows

σG(CV D)=

Accept ifint w

wCgC(θ)πC(θ) f (θ)dθ gt 0

Re ject if otherwise

σG(DS)=

Accept ifint w

wDgD(θ)πD(θ) f (θ)dθ gt 0

Re ject if otherwise

8The conditions of Lemma 1 are sufficient but not necessary for a separating equilibrium A more thoroughaccount of separating equilibria is included with the proof of Proposition 8

9The Commerce Department has ruled in favor of claimants more than 80 of the time (USITC 2010 p 4)

21

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

This logic can easily be generalized to any countable set of institutional mechanisms leadingto the following proposition

Proposition 9 In a variant of the model where there exist an arbitrary set of institutionalmechanisms Γ and where an interest group can pursue claims via any of these mechanismssimultaneously the following separating equilibrium exists

σI(θ)=

Pursue if v j(θ)π j(θ)ge c j forall j isinΓDonrsquot Pursue if otherwise

and

σG( j)=

Accept ifint w

w jg j(θ)π j(θ) f (θ)dθ ge 0forall j isinΓ

Re ject if otherwise

Where each w j is determined by the point where

v j(θ)π j(θ)= c j

Proof follows immediately from prior discussion Thus allowing interest groups to pursuemultiple mechanisms simultaneously allows for an equilibrium in which some groups do ex-actly that with Government accepting all claims via whatever subset of these mechanismsscreen a sufficiently positively weighted distribution of interest group types This accordswell with what we sometimes observe there are some instances where governments pursuemultiple actions at the same time in response to interest group demands10 Note howeverthat this structure changes the cutpoints relative to a situation where government actionsare mutually exclusive

The Role of the Government as a GatekeeperThere is some variation in the degree to which governments serve as gatekeepers to these in-stitutional mechanisms Most prominently investor-state dispute settlement (ISDS) mech-anisms which exist in a number of PTAs allow firms to directly sue governments Beyondthese procedures exist in certain agreements that allow private firms standing to petitionfor dispute resolution as with Chapter 19 in NAFTA Moreover the procedure for imple-menting anti-dumping measures and countervailing duties in the US formally allows norole for the president or politics in the process as the trade remedies are triggered almostautomatically upon affirmative USITC and Commerce Department rulings however theevidence largely suggests that politics can enter into these proceedings via the Commerce

10If however rejection from one mechanism leads to other actions also being overturned - for example if anegative ruling in dispute settlement on subsidies leads to CVDs being removed - then this would more prop-erly be thought of as an instance of the original version of the model where groups choose whether to pursuea more stringent remedy knowing they have a higher probability of losing all remedies as a consequence Inthis case the benefit derived by the group from the more stringent remedy would include the value of the lessstringent one that may be applied simultaneously

22

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

Department determinations (Busch et al 2008 p 6-7)

Given that measures like ISDS have been some of the biggest sources of controversy in tradenegotiations - Elizabeth Warren for instance has written multiple op-eds opposing ISDSand written a letter to Robert Lighthizer urging him to oppose ISDS provisions in NAFTArenegotiations11 - it is worth evaluating what the model suggests the impact of removingthe governmentrsquos role as gatekeeper should be

Perhaps unexpectedly the model suggests that there should be little impact to doing so aslong as the institution is properly calibrated Essentially what drives the selection processin this model is that the probability of success compared to the cost of pursuing a disputeis such that only higher types find it incentive compatible to pursue disputes so as long asthese probability and cost functions are properly designed it should not matter whether thegovernment has the ability to reject claims because they do not have the private informa-tion required to determine which groups have legitimate cases anyways

Where things get more complicated is when there are multiple overlapping institutionseach with varying probabilities of success and costs to pursuing disputes In this environ-ment there only needs to be one institution in the menu of possible options that is poorlycalibrated for things to become problematic from the standpoint of governments if thereexists an institution that is too permissive relative to its costs then low type firms will dis-proportionately use it to pursue weak legal cases

Indeed this seems to accord well with observation Famously (and controversially) PhilipMorris sued Uruguay over measures designed to reduce smoking (such as requiring graphicwarnings on cigarette packages) and did so via ISDS provisions in a bilateral investmenttreaty between Switzerland and Uruguay12 This would appear to be a clear example offorum-shopping for the most permissive resolution process possible and the weakness ofthe case was evident both on the merits (efforts to promote public health are explicitly pro-tected in nearly every agreement of this sort) and its eventual outcome (Philip Morris lost)Statistical evidence also supports the claim that ISDS measures have become havens forweak cases Pelc (2017) finds that investors win about 21 of indirect expropriation law-suits via ISDS which make up about 70 of the cases lodged via ISDS since 1993 (Pelc2017 p 562 p 580)

In this sense government gatekeeping provides a check on the institutional calibration pro-cess If an institution is poorly calibrated governments will simply reject the vast majorityof claims that go through it keeping it from becoming a serious issue If however the insti-tution is correctly calibrated then it should be able to operate more or less autonomously

11See September 19 2017 Senator Elizabeth Warren Press Release12July 8 2016 The Associated Press

23

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

ApplicationsThere are many possible outcomes described by this paper which depend on the particularformulation of the model (eg allowing for varying benefits or not) and which of multipleequilibria occurs However we can broadly separate the outcomes into two broad categoriesinformative or uninformative In informative equilibria the selection of various mechanismsor institutions helps to reveal information about a grouprsquos type via screening It is underthese circumstances that institutions can be useful to governments in the ways describedby this paper Thus it is worth considering what kinds of choices might allow for thiskind of screening I outline several different choices that interest groups can make betweendifferent institutions and instruments each of which provides an opportunity for screening

Different Kinds of Screening

Between International Institutions Forum ShoppingOne way in which screening might occur is between international institutions for instanceif different institutions differ in their likelihoods for approving certain claims then firmsmay lobby the government to pursue a particular dispute forum over another in a way thatsignals information However assessing when this might be the case is difficult due to theoverwhelmingly large number of PTAs with different legal standards that are not alwayseasily placed in a ldquostringency hierarchyrdquo as well as difficulties in measuring ldquosuccessrdquo orldquofailurerdquo in any legal case (oftentimes a disputant will receive a mixture of favorable andunfavorable rulings) etc

However in the case of United States dispute resolution via the WTO and NAFTA itwould appear that the stylized facts correspond to the screening equilibria of the modelThe NAFTA dispute resolution process was developed under heavy influence from the USCongress and the US is the largest and most powerful country in NAFTA by far so onewould expect NAFTA to be the institution most likely to return a ruling favorable to US in-terests - indeed there are instances in which the US has advocated for dispute resolution tobe moved from the WTO to NAFTA after proceedings had begun as with the Tuna-DolphinII case between the US and Mexico (Pauwelyn 2009) However despite this the US has ahigh success rate for cases via the WTO while their success rate via NAFTA is generallymuch lower (Guzman 2002 Davis 2012 McRae and Siwiec 2010) This accords well withthe predictions of the model while NAFTA is arguably more favorable to the United Statesfirms with the strongest cases select into pursuing their disputes via the WTO

Between Dispute Settlement and Trade RemediesAnother way in which screening might occur is between pursuing claims via dispute set-tlement or pursuing government support via some trade remedy The cleanest example of

24

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

this as mentioned earlier is subsidies in the US an interest group can choose to lobby thegovernment to challenge a foreign countryrsquos laws or they can simply apply for CVDs viathe USITC and Department of Commerce Dispute settlement gives a much lower chanceof success given that the majority of CVDs go unchallenged the worst case scenario of aCVD is that it is challenged by a foreign country in which case it would go to dispute res-olution anyway and the firm would be no worse off than if they had started with disputesettlement A similar selection process could be at work with antidumping a firm couldlobby to have foreign antidumping measures challenged or could engage in ldquoretaliatoryrdquoantidumping whereby they request antidumping measures be put in place for imports thatthey are competing with (this works only if the industry has both imports and exports fromthe country in question)13 The stylized facts on this are largely consistent with the sepa-rating equilibria of the model as mentioned earlier WTO claimants have extremely highrates of success while nearly every trade remedy that has been challenged by the WTO hasbeen ruled inconsistent with the trade law in at least one respect (Bown 2005 p 1)

Between Different Trade RemediesTrade institutions allow firms to pursue temporary protection via multiple instrumentssafeguard measures antidumping provisions or countervailing duties In many ways thesemechanisms are substitutes for each other (see Bown 2013) but they vary in the stringencyof what needs to be demonstrated in order to claim protection For instance both safe-guard measures and antidumping duties require proof of injury to a firm but antidumpingmeasures require additional evidence that goods are being dumped at lower prices in thedomestic market Thus if antidumping provisions have more stringent requirements wemight expect screening to occur whereby firms with the strongest cases select into pursuingthem

The stylized facts once again appear to comport well with the separating equilibrium storyMany antidumping measures are approved (in the US hundreds have been implementedsince 1995) and only a small percentage (less than 10) of these have been challenged viathe WTO DSU (Bown 2015a) In contrast safeguard measures have been employed by theUnited States only six times since 1995 of which four have been challenged and then foundin violation of WTO law (Bown 2015b Bown 2015c)

In a separating equilibrium we would expect that screening would occur in which firmswith stronger cases pursue antidumping while those with weaker cases would choose topursue safeguard measures and would then be rejected What we observe is that very fewfirms formally apply for safeguards I posit that this may be because applications to theUSITC are not made in a vacuum but in the shadow of prior lobbying activity (Hansen1990) Firms already have a sense of the likelihood of their claim being approved by thegovernment prior to filing a claim with the USITC thus the vast majority of such firms

13Blonigen and Bown (2003) provides evidence of this occurring

25

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

with weak cases have already been informally ldquopre-rejectedrdquo for safeguard provisions anddo not bother filing petitions14 Despite this past presidents have still decided not to pursuethe majority of cases that pass the USITCrsquos determination of injury test they rejected 21out of 40 cases total from 1974-2016 (Bown and Joseph 2017)

The remaining cases that are pursued are those for which the United States has an over-whelming political interest in doing so irrespective of the strength of the case as with thefamous US steel tariffs of 2002 In contrast pursuing an antidumping claim can be a wayof signaling injury credibly to the government which may at least partially account for thegreater use of antidumping and relatively low rates of such measures being declined or over-turned

It is worth discussing a few caveats to this characterization Some have argued that safe-guards are more ldquocostlyrdquo to use noting that they formally require ldquoserious injuryrdquo insteadof ldquomaterial injuryrdquo and noting that they used to require that compensation be providedto those countries that faced increased tariffs as a result of their use However safeguardswere reformed with the advent of the WTO and currently do not require compensationfor any measures put in place for three years or less It is also not entirely clear how thestandards for assessing ldquomaterial injuryrdquo are different from those for ldquoserious injuryrdquo giventhat neither Congress nor the GATT precisely defined what makes injury ldquoseriousrdquo withCongress providing only a list of characteristics that might be indicators (Sykes 2003 p 7)However if one believed that antidumping measures are uniformly less costly than safe-guard measures then one could argue the outcome resembles the pooling equilibria of themodel more than the separating equilibria

Between International Institutions and Unilateral ProtectionPerhaps most importantly screening may also occur between firms pursuing protection via apotentially arbitrable mechanism embedded in some international institution versus pursu-ing protection unilaterally from the government Any interest group could bypass the wholeprocess of legal trade remedies and lobby the government directly for protection this wouldhave a 0 probability of being overturned as there would be no opportunity for it to belegally challenged This corresponds to the degenerate case described during the discussionof the model In this case screening occurs when firms subject themselves to the legal costsand possibility of being overturned associated with pursuing protection via means subject tobinding arbitration in a separating equilibrium only those with the strongest cases shouldbe willing to do so and those that try to pursue unilateral protection would usually havetheir efforts denied

This also appears to correspond with reality As recent events have made clear the United

14This pre-filing rejection process as applied to CVDs is discussed openly by Commerce Department officialsin The Washington Post July 13 2003

26

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

States has always had a way of pursuing protection without fear of legal retaliation security-based tariffs under Section 232 which are explicitly allowed for under GATTWTO lawSection 232 claims do not require an assessment of injury by the USITC and were designedwith almost no restrictions on what government could define as ldquosecurityrdquo out of deferenceto state sovereignty It has been described by trade law scholars as a ldquocatch-all clauserdquo thatis ldquoso broad self-judging and ambiguous that it can obviously be abusedrdquo (Jackson 1997 p230 Pelc 2016 p 3) The clause is so broad that many trade lawyers have concluded thattariffs implemented under a security justification are unreviewable by WTO panels Indeedin the Commerce Department ruling advising that steel tariffs be implemented under Sec-tion 232 there is not even an attempt to provide reasoning along traditional security linesinstead security is expanded to mean ldquoeconomic securityrdquo with the Commerce Departmentarguing that establishing tariffs to ensure the long term economic health of the steel indus-try is important on these amorphous grounds (Commerce Department Report 2018)

Moreover Section 232 has almost never been used to provide protection to any groups de-spite the fact that it has such broad legal potential for abuse In fact the last time it wasinvoked by the US was in 1986 by the Reagan administration and that was under veryspecific circumstances involving broader negotiations with the affected parties15 Why thenhave governments historically exercised this high degree of restraint This paper provides acompelling rationale that the only groups that would pursue protection via this mechanismwould be those that could not meet the criteria of any other mechanism and thus usuallythose with the lowest types

As mentioned earlier this kind of screening directs our attention to questions of institu-tional design Why might governments want to establish international institutions thatbind them to using trade protection in only a restricted set of circumstances This paperprovides an important new explanation for these choices that governments may be betterable to allocate protection because of the information obtained through the screening deviceof international institutions This potential benefit is demonstrated in the model and hasnot been explored in the existing literature Furthermore the model also provides insightinto the conditions where governments might cease to value these benefits notably witha strong enough upward shift in protectionist sentiment governments may no longer careto preserve the institutions that allow for this kind of screening as they may believe them-selves to be better off in an equilibrium in which they can protect as many groups as possiblewithout restraint

ConclusionAs discussed in this paper international institutions provide governments with a dizzyingarray of options for supporting domestic industries each with varying costs and benefits

15May 21 1986 The New York Times

27

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

Governments can choose where (ie at which international institution) to file trade dis-putes whether to launch a trade dispute or make use of ldquoflexibility measuresrdquo in a retalia-tory fashion and which flexibility measure (amongst several) to use to impose temporaryprotection This paperrsquos model has demonstrated that governments can use this multitudeof options to screen between domestic groups for those with the strongest legal cases wherethe strength of a case is private information that interest groups would otherwise not havean incentive to honestly reveal

This selection process can help to explain amongst other things why disputes pursued viathe WTO have such a high rate of success (approximately 90) and why claims via theISDS have such a low rate of success Furthermore this paper has provided an explanationfor the ldquocriteriardquo based system of trade remedies and demonstrated another way in whichinternational institutions and their particular features can be valuable to governmentsnamely their potential for providing governments with useful information about domesticpolitical groups Thus the paper contributes to our broad understanding of the areas ofinstitutional selection and design both within the domain of trade politics and plausibly inother domains that share similar features

ReferencesBaccini L Dur A Elsig M amp Milewicz K (2011) The Design of Preferential TradeAgreements In W T Report (Ed) Preferential Trade Agreements and the WTO A NewEra edited by the WTO

Bagwell K amp Staiger R W (2004) The Economics of the World Trading System MITPress

Bebchuk L A (1984) Litigation and Settlement Under Imperfect Information The RANDJournal of Economics 15(3) 404ndash415

Betz T (2018) Domestic Institutions Trade Disputes and the Monitoring and Enforcementof International Law International Interactions 44 04

Blonigen B A amp Bown C P (2003) Antidumping and retaliation threats Journal ofInternational Economics 60(2) 249ndash273

Bown C P (nd-a) Global Antidumping Database (Report) The World Bank Retrievedfrom httpeconworldbankorgttbdgad

Bown C P (nd-b) Global Dispute Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbddsud

Bown C P (nd-c) Global Safeguard Database (Report) The World Bank Retrieved fromhttpeconworldbankorgttbdgsgd

28

Bown C P (2005) Trade Remedies and World Trade Organization Dispute SettlementWhy Are So Few Challenged The Journal of Legal Studies 34(2) 515ndash55

Bown C P (2013) How Different Are Safeguards from Antidumping Evidence From USTrade Policies Towards Steel Review of Industrial Organization v 42 449ndash481

Bown C P amp Joseph J (2017) Solar and Washing Machine Safeguards in Context TheHistory of US Section 201 Use (Report) Peterson Institute for International Economics

Brutger R (2017) Litigation for Sale Private Firms and WTO Dispute Escalation

Busch M L (2007) Overlapping Institutions Forum Shopping and Dispute Settlement inInternational Trade International Organization 309ndash338

Busch M L Raciborski R amp Reinhardt E (2008) Does the Rule of Law Matter TheWTO and US Antidumping Investigations

Davis C L (2009 March) Overlapping Institutions in Trade Policy Perspectives on Politics7(1) 25ndash31

Davis C L (2012) Why Adjudicate Enforcing Trade Rules in the WTO Princeton Univer-sity Press

DOC (2018) The Effect of Imports on the National Security (Report) Department ofCommerce

Gilligan M Johns L amp Rosendorff B P (2010 February) Strengthening InternationalCourts and the Early Settlement of Disputes Journal of Conflict Resolution 54(1) 5ndash38

Guzman A (2002) The Political Economy of Litigation and Settlement at the WTO (Re-search Paper) Public Law and Legal Theory

Hansen W L (1990) The International Trade Commission and the Politics of Protection-ism American Political Science Review 84 01

Howse R (1998) Settling Trade Remedy Disputes When the WTO Forum Is Better than theNAFTA CD Howe Communique

Jackson J H (1997) The World Trading System Law and Policy of International EconomicRelations MIT Press

Keohane R O (1984) After Hegemony Cooperation and Discord in the World PoliticalEconomy Princeton University Press

Koremenos B (2001) Loosening the Ties That Bind A Learning Model of AgreementFlexibility International Organization 55 02

29

Koremenos B Lipson C amp Snidal D (2001) The Rational Design of International Insti-tutions International Organization 55 04

Kucik J amp Pelc K J (2014) Can International Legal Rulings Deter Financial MarketsSay Yes

Kucik J amp Pelc K J (2016) Over-commitment and Backsliding in International TradeEuropean Journal of Political Research 55 02

Kucik J amp Reinhardt E (2008) Does Flexibility Promote Cooperation An Application tothe Global Trade Regime International Organization 62 03

McRae D amp Siwiec J (2010) NAFTA Dispute Settlement Success or Failure In America(Ed) Arturo Oropeza Garcia Mexico City Corporacion Industrial Grafica

Pauwelyn J (2006) Editorial Comment Adding Sweeteners to Softwood Lumber TheWTO-NAFTA rsquoSpaghetti Bowlrsquo Is Cooking Journal of International Economic Law 9(1)197ndash206

Pauwelyn J (2009 April) Mexico Secures WTO Tuna-Dolphin Panel US Seeks Move ToNAFTA Inside US Trade 24 2009

Pelc K (2011) Seeking Escape The Use of Escape Clauses in International Trade Agree-ments International Studies Quarterly 53 02

Pelc K (2014) The Politics of Precedent in International Law A Social Network Applica-tion American Political Science Review 108 03

Pelc K (2016) Making and Bending Rules The Design of Exceptions and Escape Clausesin Trade Law Cambridge University Press

Pelc K (2017) What Explains the Low Success Rate of Investor-State Disputes Interna-tional Organization 71 03

Reinganum J F amp Wilde L L (1986) Settlement Litigation and the Allocation ofLitigation Costs The RAND Journal of Economics 17(4) 557ndash566

Rosendorff B P (2005) Stability and Rigidity Politics and Design of the WTOrsquos DisputeSettlement Procedure American Political Science Review 99(03) 389ndash400

Rosendorff B P amp Milner H V (2001) The Optimal Design of International Trade Insti-tutions Uncertainty and Escape International Organization 55 04

Ross W L (2018) Budget in Brief Fiscal Year 2018 (Report) Department of Commerce

Ryu J amp Stone R W (2018) Plaintiffs by Proxy A firm-level approach to WTO disputeresolution The Review of International Organizations 13(2) 273ndash308

30

Sykes A (2003 May) The Safeguards Mess A Critique of WTO Jurisprudence Lawamp Economics Working Papers Retrieved from httpschicagounbounduchicagoedu

law_and_economics571

Trebilcock M J (2011) Understanding Trade Law Cheltenham UK Northampton MAEdward Elgar Pub

Trebilcock M J amp Howse R (2005) The Regulation of International Trade 3rd Edition(3rd ed) London New York Routledge

USITC (2010) Import Injury Investigations Case Statistics FY 1980-2008 (Report) UnitedStates International Trade Commission

USITC (2016) Budget Justification Fiscal Year 2016 (Report) United States InternationalTrade Commission

USTR (2015) Fiscal Year 2015 Budget (Report) Office of the United States Trade Repre-sentation

Appendix

Proof of Proposition 1Proposition 1 encompasses two pooling Perfect Bayesian Nash equilibria (PBNE) each ofwhich is conditional on different off-equilibrium path beliefs Consider first the followingconjecture where Government G and the interest group I pursue the following strategiesσG and σI respectively

σG =

Accept if I pursues CVDRe ject if I pursues DS

σI =CVD forallθ isinΘThis is incentive compatible for all θ types of I by earlier assumption at least one institutionis assumed to be profitable For Government Accept is better than Reject ifint w

0g(θ)πC(θ) f (θ)dθ ge 0

Because in this conjecture all types choose CVD Reject is better than Accept upon observingDS when off-equilibrium path beliefs q sim p(q) are such thatint w

0g(q)πc(q)p(q)dq le 0

However if we reverse this condition Grsquos strategy must change to Accept|DS Howeverthis does not change I rsquos best response since πD(θ) lt πC(θ) forallθ isinΘ and cD gt cC by assump-tion v(θ)πD(θ)minus cD lt v(θ)πC(θ)minus cC which ensures that CV D remains I rsquos best responseThus a new equilibrium holds where

31

Koremenos B Lipson C amp Snidal D (2001) The Rational Design of International Insti-tutions International Organization 55 04

Kucik J amp Pelc K J (2014) Can International Legal Rulings Deter Financial MarketsSay Yes

Kucik J amp Pelc K J (2016) Over-commitment and Backsliding in International TradeEuropean Journal of Political Research 55 02

Kucik J amp Reinhardt E (2008) Does Flexibility Promote Cooperation An Application tothe Global Trade Regime International Organization 62 03

McRae D amp Siwiec J (2010) NAFTA Dispute Settlement Success or Failure In America(Ed) Arturo Oropeza Garcia Mexico City Corporacion Industrial Grafica

Pauwelyn J (2006) Editorial Comment Adding Sweeteners to Softwood Lumber TheWTO-NAFTA rsquoSpaghetti Bowlrsquo Is Cooking Journal of International Economic Law 9(1)197ndash206

Pauwelyn J (2009 April) Mexico Secures WTO Tuna-Dolphin Panel US Seeks Move ToNAFTA Inside US Trade 24 2009

Pelc K (2011) Seeking Escape The Use of Escape Clauses in International Trade Agree-ments International Studies Quarterly 53 02

Pelc K (2014) The Politics of Precedent in International Law A Social Network Applica-tion American Political Science Review 108 03

Pelc K (2016) Making and Bending Rules The Design of Exceptions and Escape Clausesin Trade Law Cambridge University Press

Pelc K (2017) What Explains the Low Success Rate of Investor-State Disputes Interna-tional Organization 71 03

Reinganum J F amp Wilde L L (1986) Settlement Litigation and the Allocation ofLitigation Costs The RAND Journal of Economics 17(4) 557ndash566

Rosendorff B P (2005) Stability and Rigidity Politics and Design of the WTOrsquos DisputeSettlement Procedure American Political Science Review 99(03) 389ndash400

Rosendorff B P amp Milner H V (2001) The Optimal Design of International Trade Insti-tutions Uncertainty and Escape International Organization 55 04

Ross W L (2018) Budget in Brief Fiscal Year 2018 (Report) Department of Commerce

Ryu J amp Stone R W (2018) Plaintiffs by Proxy A firm-level approach to WTO disputeresolution The Review of International Organizations 13(2) 273ndash308

30

Sykes A (2003 May) The Safeguards Mess A Critique of WTO Jurisprudence Lawamp Economics Working Papers Retrieved from httpschicagounbounduchicagoedu

law_and_economics571

Trebilcock M J (2011) Understanding Trade Law Cheltenham UK Northampton MAEdward Elgar Pub

Trebilcock M J amp Howse R (2005) The Regulation of International Trade 3rd Edition(3rd ed) London New York Routledge

USITC (2010) Import Injury Investigations Case Statistics FY 1980-2008 (Report) UnitedStates International Trade Commission

USITC (2016) Budget Justification Fiscal Year 2016 (Report) United States InternationalTrade Commission

USTR (2015) Fiscal Year 2015 Budget (Report) Office of the United States Trade Repre-sentation

Appendix

Proof of Proposition 1Proposition 1 encompasses two pooling Perfect Bayesian Nash equilibria (PBNE) each ofwhich is conditional on different off-equilibrium path beliefs Consider first the followingconjecture where Government G and the interest group I pursue the following strategiesσG and σI respectively

σG =

Accept if I pursues CVDRe ject if I pursues DS

σI =CVD forallθ isinΘThis is incentive compatible for all θ types of I by earlier assumption at least one institutionis assumed to be profitable For Government Accept is better than Reject ifint w

0g(θ)πC(θ) f (θ)dθ ge 0

Because in this conjecture all types choose CVD Reject is better than Accept upon observingDS when off-equilibrium path beliefs q sim p(q) are such thatint w

0g(q)πc(q)p(q)dq le 0

However if we reverse this condition Grsquos strategy must change to Accept|DS Howeverthis does not change I rsquos best response since πD(θ) lt πC(θ) forallθ isinΘ and cD gt cC by assump-tion v(θ)πD(θ)minus cD lt v(θ)πC(θ)minus cC which ensures that CV D remains I rsquos best responseThus a new equilibrium holds where

31

Sykes A (2003 May) The Safeguards Mess A Critique of WTO Jurisprudence Lawamp Economics Working Papers Retrieved from httpschicagounbounduchicagoedu

law_and_economics571

Trebilcock M J (2011) Understanding Trade Law Cheltenham UK Northampton MAEdward Elgar Pub

Trebilcock M J amp Howse R (2005) The Regulation of International Trade 3rd Edition(3rd ed) London New York Routledge

USITC (2010) Import Injury Investigations Case Statistics FY 1980-2008 (Report) UnitedStates International Trade Commission

USITC (2016) Budget Justification Fiscal Year 2016 (Report) United States InternationalTrade Commission

USTR (2015) Fiscal Year 2015 Budget (Report) Office of the United States Trade Repre-sentation

Appendix

Proof of Proposition 1Proposition 1 encompasses two pooling Perfect Bayesian Nash equilibria (PBNE) each ofwhich is conditional on different off-equilibrium path beliefs Consider first the followingconjecture where Government G and the interest group I pursue the following strategiesσG and σI respectively

σG =

Accept if I pursues CVDRe ject if I pursues DS

σI =CVD forallθ isinΘThis is incentive compatible for all θ types of I by earlier assumption at least one institutionis assumed to be profitable For Government Accept is better than Reject ifint w

0g(θ)πC(θ) f (θ)dθ ge 0

Because in this conjecture all types choose CVD Reject is better than Accept upon observingDS when off-equilibrium path beliefs q sim p(q) are such thatint w

0g(q)πc(q)p(q)dq le 0

However if we reverse this condition Grsquos strategy must change to Accept|DS Howeverthis does not change I rsquos best response since πD(θ) lt πC(θ) forallθ isinΘ and cD gt cC by assump-tion v(θ)πD(θ)minus cD lt v(θ)πC(θ)minus cC which ensures that CV D remains I rsquos best responseThus a new equilibrium holds where

31

σG =

Accept if I pursues CVDAccept if I pursues DS

σI =CVD forallθ isinΘ

Proof of Proposition 2Consider the conjecture

σG =

Re ject if I pursues CVDAccept if I pursues DS

σI =DS forallθ isinΘIf I pursues CVD they are rejected and receive a payoff of 0 So they choose DS if v(θ)minuscD ge0 which is true forallθ isinΘ by condition 2 of the proposition

Similar to Proposition 1 Government receives a payoff ofint w

0g(θ)πD(θ) f (θ)dθ

for Accept which is ge 0 (and thus is preferable to rejecting) by condition 1 of the propositionThus all we need to ensure this is a PBNE is to set off equilibrium path beliefs q sim p(q) suchthat Government prefers to reject when they observe CV D which is ensured condition 3 ofthe proposition ie int w

0g(q)πC(q)p(q)dq le 0

Proof of Proposition 3Each of these pooling equilibria rely on Government rejecting any deviations by I to an-other institution other than the one conjectured ensuring this is simply a matter of settingthe correct off-equilibrium path beliefs as outlined in the proposition Under these circum-stances I is indifferent between institutions receiving a payoff of 0 in either case Thenall that remains to ensure a PBNE is for Government to prefer to Reject rather than Acceptgiven that all types are pursuing claims via that institution the conditions ensuring thisare also outlined in the proposition Thus we have two pooling equilibria one under theconditions outlined in Proposition 3 part (i)

σG =

Re ject if I pursues CVDRe ject if I pursues DS

σI =CVD forallθ isinΘ

32

And one under the conditions outlined in Proposition 3 part (ii)

σG =

Re ject if I pursues CVDRe ject if I pursues DS

σI =DS forallθ isinΘ

Proof of Corollary 1Consider any w where g(w) 6= 0 If g(w)lt 0 then since g(θ) is continuous and monotonicallyincreasing and since there must exist θ such that g(θ) gt 0 (otherwise it could not be thatUG(Accept|DS) ge 0) there must exist some w gt w such that g(w) = 0 Furthermore itmust be the case that forallθ isin (w w) g(θ) lt 0 This set ΘL in Corollary 1 represents typesof firms receiving Government protection that a perfectly informed Government would turndown Symmetrically if g(w) gt 0 there must exist w lt w such that g(w) = 0 implying thatforallθ isin (w w) g(θ)gt 0 This set ΘH in Corollary 1 represents types of firms that do not receiveGovernment protection that a perfectly informed Government would protect

Proof of Proposition 6Consider the pooling conditions from Proposition 1 and Proposition 2 respectively

E[g(θ)|AllCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

E[g(θ)|AllDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0

We have defined an upward shift in protectionist sentiment as a shift to g+(θ) such thatg+(θ) ge g(θ)forallθ isin Θ with some positive Lebesgue measure subset Θprime such that g+(θprime) gtg(θprime)forallθprime isinΘprime We can partition Θ into two sets Θprime and Θ

primeprime where g+(θ

primeprime) = g(θ

primeprime)forallθprimeprime isinΘprimeprime

Thus for both pooling conditions we can rewrite the relevant integrals for either institutionz as

E[g(θ)|z]=intΘprime

g(θprime)πz(θprime) f (θprime)dθprime︸ ︷︷ ︸α

+intΘ

primeprime g(θprimeprime)πz(θ

primeprime) f (θ

primeprime)dθ

primeprime

︸ ︷︷ ︸β

Which we can compare to the following after a shift in protectionist sentiment

E[g+(θ)|z]=intΘprime

g+(θprime)πz(θprime) f (θprime)dθprime︸ ︷︷ ︸A

+intΘ

primeprime g+(θprimeprime)πz(θ

primeprime) f (θ

primeprime)dθ

primeprime

︸ ︷︷ ︸B

Because g+(θprimeprime) = g(θprimeprime)forallθprimeprime isin Θprimeprime it must be the case that β = B Similarly since g+(θprime) gtg(θprime)forallθprime isinΘprime it must be the case that A gtα if either is nonzero Furthermore since Θprime haspositive Lebesgue measure at least one must be nonzero Thus we have shown that

E[g+(θ)|z]gt E[g(θ)|z]

33

Which means that the pooling conditions of Proposition 1 and 2 are more likely to be metafter an upward shift in protectionist sentiment

Proof of Proposition 7From the text note that an institution-induced separating equilibrium is preferred to noinstitution whenever

UG(Institution)=int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ =UG(No Institution)

Now in a similar fashion to the proof of Proposition 6 partition Θ into Θprime and Θprimeprime We wantto see which side of the inequality changes more with a shift to g+(θ) Following Proposition6 the integrals will be the same over Θprimeprime thus we can compare

∆UG(Institution)=intΘprime

g+(θprime)πD(θprime) f (θprime)dθprimeminusintΘprime

g(θprime)πD(θprime) f (θprime)dθprime

=intΘprime

[g+(θprime)minus g(θprime)]πD(θprime) f (θprime)dθprime

∆UG(No Institution)=intΘprime

g+(θprime) f (θprime)dθprimeminusintΘprime

g(θprime) f (θprime)dθprime

=intΘprime

[g+(θprime)minus g(θprime)] f (θprime)dθprime

Since πD(θprime) le 1forallθprime isinΘprime and since there must exist θprime isinΘprime such that πD(θprime) lt 1 given thatπD(middot) is strictly monotonically increasing ∆UG(No Institution) gt ∆UG(Institution) Or inwords an upwards shift in protectionist sentiment has a more significant positive impacton Governmentrsquos payoff in the ldquono institutionrdquo equilibrium than in a separating equilibriumobtained under an institution and thus this upwards shift in sentiment makes the institu-tion relatively less attractive

Proof of Lemma 1We are comparing the following two values

UI(DS)= vD(θ)πD(θ)minus cD

UI(CV D)= vC(θ)πC(θ)minus cC

Since πC gt πD and probabilities are bounded above by one l imθrarrwπD(θ) = 1 implies thatl imθrarrwπC(θ)= 1 by squeeze theorem Thus by applying properties of limits we can computethe following

l imθrarrwUI(DS)= (1)l imθrarrwvD(θ)minus cD

l imθrarrwUI(CV D)= (1)l imθrarrwvC(θ)minus cC

34

Since cC and cD are identical this suggests that as θ rarr w UI(DS) gt UI(CV D) sincevD(θ) gt vC(θ) forallθ isin Θ Thus since we have earlier assumed that there exists θ such thatUI(DS|θ)lt 0 which implies that UI(CV D|θ)gtUI(DS|θ) by the continuity in θ of vC and vDand the intermediate value theorem there must exist some θ such that UI(DS)=UI(CV D)

However in order to establish that this point w is a cutpoint we must further establish thatall types above that point prefer DS and all types below that point prefer CV D To ensurethis consider that I prefers DS over CV D whenever

vD(θ)πD(θ)minus cD minusvC(θ)πC(θ)+ cC gt 0

Which if cC = cD

harr vD(θ)πD(θ)minusvC(θ)πC(θ)gt 0

The final condition of Lemma 1 ensures that this expression is strictly monotonically in-creasing in θ consequently the inequality is satisfied for all θ types above w and not satis-fied for all θ types below w Thus w is a cutpoint

Proof of Proposition 8This proof largely follows from the discussion in text and the proof for Lemma 1 Lemma1 establishes that I will adopt the strategy outlined in Proposition 8 in response to Propo-sition 8rsquos conjectured strategy by Government The additional conditions in Proposition 8ensure that Governmentrsquos strategy is a best response to I rsquos cutpoint strategy by statingthat the payoff to Government of accepting claims by the subset of types screened to eitherinstitution is still higher than what is obtained by rejecting all claims to that institution

However some discussion of the role of Lemma 1 is warranted Without the conditions ofLemma 1 a separating equilibrium is still possible but it has a somewhat less clean inter-pretation Consider that without the monotonicity assumption from Lemma 1 there neednot be a cutpoint above which types select into the more stringent institution indeed whilehigher types will have higher valuations from DS and higher probabilities of success thanlower types it would be possible that their probability of success and valuations might in-crease faster for CV D leading to a peculiar situation in which higher types might select intothe less stringent institution In this case instead of a cutpoint strategy we can constructthe following strategy for I

σI =

DS if vD(θ)πD(θ)minus cD ge vC(θ)πC(θ)minus cC

CV D if vD(θ)πD(θ)minus cD lt vC(θ)πC(θ)minus cC

Noting that depending on the shape of the probability and benefit functions these two dis-joint sets could be any union of subsets of the type space If we call the DS subset ΘD and

35

the CV D subset ΘC we can consider the following strategy by Government

σG =

Accept if I pursues DSAccept if I pursues CVD

Which will be a best response whenever the following conditions holdintΘC

gC(θ)πC(θ) f (θ)dθ ge 0

intΘD

gD(θ)πD(θ) f (θ)dθ ge 0

Thus a separating equilibrium can be constructed if the above conditions hold without rely-ing on Lemma 1 This separating equilibrium does reveal information about the type of Ito Government but in a somewhat less easily interpretable fashion Nonetheless the equi-librium described will only hold if a sufficiently positively weighted distribution of typesis screened to DS to overcome its relatively low chances of success compared to CV D andGovernmentrsquos preference for CV D over DS in this variant of the model

36

And one under the conditions outlined in Proposition 3 part (ii)

σG =

Re ject if I pursues CVDRe ject if I pursues DS

σI =DS forallθ isinΘ

Proof of Corollary 1Consider any w where g(w) 6= 0 If g(w)lt 0 then since g(θ) is continuous and monotonicallyincreasing and since there must exist θ such that g(θ) gt 0 (otherwise it could not be thatUG(Accept|DS) ge 0) there must exist some w gt w such that g(w) = 0 Furthermore itmust be the case that forallθ isin (w w) g(θ) lt 0 This set ΘL in Corollary 1 represents typesof firms receiving Government protection that a perfectly informed Government would turndown Symmetrically if g(w) gt 0 there must exist w lt w such that g(w) = 0 implying thatforallθ isin (w w) g(θ)gt 0 This set ΘH in Corollary 1 represents types of firms that do not receiveGovernment protection that a perfectly informed Government would protect

Proof of Proposition 6Consider the pooling conditions from Proposition 1 and Proposition 2 respectively

E[g(θ)|AllCV D]=int w

0g(θ)πC(θ) f (θ)dθ ge 0

E[g(θ)|AllDS]=int w

0g(θ)πD(θ) f (θ)dθ ge 0

We have defined an upward shift in protectionist sentiment as a shift to g+(θ) such thatg+(θ) ge g(θ)forallθ isin Θ with some positive Lebesgue measure subset Θprime such that g+(θprime) gtg(θprime)forallθprime isinΘprime We can partition Θ into two sets Θprime and Θ

primeprime where g+(θ

primeprime) = g(θ

primeprime)forallθprimeprime isinΘprimeprime

Thus for both pooling conditions we can rewrite the relevant integrals for either institutionz as

E[g(θ)|z]=intΘprime

g(θprime)πz(θprime) f (θprime)dθprime︸ ︷︷ ︸α

+intΘ

primeprime g(θprimeprime)πz(θ

primeprime) f (θ

primeprime)dθ

primeprime

︸ ︷︷ ︸β

Which we can compare to the following after a shift in protectionist sentiment

E[g+(θ)|z]=intΘprime

g+(θprime)πz(θprime) f (θprime)dθprime︸ ︷︷ ︸A

+intΘ

primeprime g+(θprimeprime)πz(θ

primeprime) f (θ

primeprime)dθ

primeprime

︸ ︷︷ ︸B

Because g+(θprimeprime) = g(θprimeprime)forallθprimeprime isin Θprimeprime it must be the case that β = B Similarly since g+(θprime) gtg(θprime)forallθprime isinΘprime it must be the case that A gtα if either is nonzero Furthermore since Θprime haspositive Lebesgue measure at least one must be nonzero Thus we have shown that

E[g+(θ)|z]gt E[g(θ)|z]

33

Which means that the pooling conditions of Proposition 1 and 2 are more likely to be metafter an upward shift in protectionist sentiment

Proof of Proposition 7From the text note that an institution-induced separating equilibrium is preferred to noinstitution whenever

UG(Institution)=int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ =UG(No Institution)

Now in a similar fashion to the proof of Proposition 6 partition Θ into Θprime and Θprimeprime We wantto see which side of the inequality changes more with a shift to g+(θ) Following Proposition6 the integrals will be the same over Θprimeprime thus we can compare

∆UG(Institution)=intΘprime

g+(θprime)πD(θprime) f (θprime)dθprimeminusintΘprime

g(θprime)πD(θprime) f (θprime)dθprime

=intΘprime

[g+(θprime)minus g(θprime)]πD(θprime) f (θprime)dθprime

∆UG(No Institution)=intΘprime

g+(θprime) f (θprime)dθprimeminusintΘprime

g(θprime) f (θprime)dθprime

=intΘprime

[g+(θprime)minus g(θprime)] f (θprime)dθprime

Since πD(θprime) le 1forallθprime isinΘprime and since there must exist θprime isinΘprime such that πD(θprime) lt 1 given thatπD(middot) is strictly monotonically increasing ∆UG(No Institution) gt ∆UG(Institution) Or inwords an upwards shift in protectionist sentiment has a more significant positive impacton Governmentrsquos payoff in the ldquono institutionrdquo equilibrium than in a separating equilibriumobtained under an institution and thus this upwards shift in sentiment makes the institu-tion relatively less attractive

Proof of Lemma 1We are comparing the following two values

UI(DS)= vD(θ)πD(θ)minus cD

UI(CV D)= vC(θ)πC(θ)minus cC

Since πC gt πD and probabilities are bounded above by one l imθrarrwπD(θ) = 1 implies thatl imθrarrwπC(θ)= 1 by squeeze theorem Thus by applying properties of limits we can computethe following

l imθrarrwUI(DS)= (1)l imθrarrwvD(θ)minus cD

l imθrarrwUI(CV D)= (1)l imθrarrwvC(θ)minus cC

34

Since cC and cD are identical this suggests that as θ rarr w UI(DS) gt UI(CV D) sincevD(θ) gt vC(θ) forallθ isin Θ Thus since we have earlier assumed that there exists θ such thatUI(DS|θ)lt 0 which implies that UI(CV D|θ)gtUI(DS|θ) by the continuity in θ of vC and vDand the intermediate value theorem there must exist some θ such that UI(DS)=UI(CV D)

However in order to establish that this point w is a cutpoint we must further establish thatall types above that point prefer DS and all types below that point prefer CV D To ensurethis consider that I prefers DS over CV D whenever

vD(θ)πD(θ)minus cD minusvC(θ)πC(θ)+ cC gt 0

Which if cC = cD

harr vD(θ)πD(θ)minusvC(θ)πC(θ)gt 0

The final condition of Lemma 1 ensures that this expression is strictly monotonically in-creasing in θ consequently the inequality is satisfied for all θ types above w and not satis-fied for all θ types below w Thus w is a cutpoint

Proof of Proposition 8This proof largely follows from the discussion in text and the proof for Lemma 1 Lemma1 establishes that I will adopt the strategy outlined in Proposition 8 in response to Propo-sition 8rsquos conjectured strategy by Government The additional conditions in Proposition 8ensure that Governmentrsquos strategy is a best response to I rsquos cutpoint strategy by statingthat the payoff to Government of accepting claims by the subset of types screened to eitherinstitution is still higher than what is obtained by rejecting all claims to that institution

However some discussion of the role of Lemma 1 is warranted Without the conditions ofLemma 1 a separating equilibrium is still possible but it has a somewhat less clean inter-pretation Consider that without the monotonicity assumption from Lemma 1 there neednot be a cutpoint above which types select into the more stringent institution indeed whilehigher types will have higher valuations from DS and higher probabilities of success thanlower types it would be possible that their probability of success and valuations might in-crease faster for CV D leading to a peculiar situation in which higher types might select intothe less stringent institution In this case instead of a cutpoint strategy we can constructthe following strategy for I

σI =

DS if vD(θ)πD(θ)minus cD ge vC(θ)πC(θ)minus cC

CV D if vD(θ)πD(θ)minus cD lt vC(θ)πC(θ)minus cC

Noting that depending on the shape of the probability and benefit functions these two dis-joint sets could be any union of subsets of the type space If we call the DS subset ΘD and

35

the CV D subset ΘC we can consider the following strategy by Government

σG =

Accept if I pursues DSAccept if I pursues CVD

Which will be a best response whenever the following conditions holdintΘC

gC(θ)πC(θ) f (θ)dθ ge 0

intΘD

gD(θ)πD(θ) f (θ)dθ ge 0

Thus a separating equilibrium can be constructed if the above conditions hold without rely-ing on Lemma 1 This separating equilibrium does reveal information about the type of Ito Government but in a somewhat less easily interpretable fashion Nonetheless the equi-librium described will only hold if a sufficiently positively weighted distribution of typesis screened to DS to overcome its relatively low chances of success compared to CV D andGovernmentrsquos preference for CV D over DS in this variant of the model

36

Which means that the pooling conditions of Proposition 1 and 2 are more likely to be metafter an upward shift in protectionist sentiment

Proof of Proposition 7From the text note that an institution-induced separating equilibrium is preferred to noinstitution whenever

UG(Institution)=int w

wg(θ)πD(θ) f (θ)dθ gt

int w

0g(θ) f (θ)dθ =UG(No Institution)

Now in a similar fashion to the proof of Proposition 6 partition Θ into Θprime and Θprimeprime We wantto see which side of the inequality changes more with a shift to g+(θ) Following Proposition6 the integrals will be the same over Θprimeprime thus we can compare

∆UG(Institution)=intΘprime

g+(θprime)πD(θprime) f (θprime)dθprimeminusintΘprime

g(θprime)πD(θprime) f (θprime)dθprime

=intΘprime

[g+(θprime)minus g(θprime)]πD(θprime) f (θprime)dθprime

∆UG(No Institution)=intΘprime

g+(θprime) f (θprime)dθprimeminusintΘprime

g(θprime) f (θprime)dθprime

=intΘprime

[g+(θprime)minus g(θprime)] f (θprime)dθprime

Since πD(θprime) le 1forallθprime isinΘprime and since there must exist θprime isinΘprime such that πD(θprime) lt 1 given thatπD(middot) is strictly monotonically increasing ∆UG(No Institution) gt ∆UG(Institution) Or inwords an upwards shift in protectionist sentiment has a more significant positive impacton Governmentrsquos payoff in the ldquono institutionrdquo equilibrium than in a separating equilibriumobtained under an institution and thus this upwards shift in sentiment makes the institu-tion relatively less attractive

Proof of Lemma 1We are comparing the following two values

UI(DS)= vD(θ)πD(θ)minus cD

UI(CV D)= vC(θ)πC(θ)minus cC

Since πC gt πD and probabilities are bounded above by one l imθrarrwπD(θ) = 1 implies thatl imθrarrwπC(θ)= 1 by squeeze theorem Thus by applying properties of limits we can computethe following

l imθrarrwUI(DS)= (1)l imθrarrwvD(θ)minus cD

l imθrarrwUI(CV D)= (1)l imθrarrwvC(θ)minus cC

34

Since cC and cD are identical this suggests that as θ rarr w UI(DS) gt UI(CV D) sincevD(θ) gt vC(θ) forallθ isin Θ Thus since we have earlier assumed that there exists θ such thatUI(DS|θ)lt 0 which implies that UI(CV D|θ)gtUI(DS|θ) by the continuity in θ of vC and vDand the intermediate value theorem there must exist some θ such that UI(DS)=UI(CV D)

However in order to establish that this point w is a cutpoint we must further establish thatall types above that point prefer DS and all types below that point prefer CV D To ensurethis consider that I prefers DS over CV D whenever

vD(θ)πD(θ)minus cD minusvC(θ)πC(θ)+ cC gt 0

Which if cC = cD

harr vD(θ)πD(θ)minusvC(θ)πC(θ)gt 0

The final condition of Lemma 1 ensures that this expression is strictly monotonically in-creasing in θ consequently the inequality is satisfied for all θ types above w and not satis-fied for all θ types below w Thus w is a cutpoint

Proof of Proposition 8This proof largely follows from the discussion in text and the proof for Lemma 1 Lemma1 establishes that I will adopt the strategy outlined in Proposition 8 in response to Propo-sition 8rsquos conjectured strategy by Government The additional conditions in Proposition 8ensure that Governmentrsquos strategy is a best response to I rsquos cutpoint strategy by statingthat the payoff to Government of accepting claims by the subset of types screened to eitherinstitution is still higher than what is obtained by rejecting all claims to that institution

However some discussion of the role of Lemma 1 is warranted Without the conditions ofLemma 1 a separating equilibrium is still possible but it has a somewhat less clean inter-pretation Consider that without the monotonicity assumption from Lemma 1 there neednot be a cutpoint above which types select into the more stringent institution indeed whilehigher types will have higher valuations from DS and higher probabilities of success thanlower types it would be possible that their probability of success and valuations might in-crease faster for CV D leading to a peculiar situation in which higher types might select intothe less stringent institution In this case instead of a cutpoint strategy we can constructthe following strategy for I

σI =

DS if vD(θ)πD(θ)minus cD ge vC(θ)πC(θ)minus cC

CV D if vD(θ)πD(θ)minus cD lt vC(θ)πC(θ)minus cC

Noting that depending on the shape of the probability and benefit functions these two dis-joint sets could be any union of subsets of the type space If we call the DS subset ΘD and

35

the CV D subset ΘC we can consider the following strategy by Government

σG =

Accept if I pursues DSAccept if I pursues CVD

Which will be a best response whenever the following conditions holdintΘC

gC(θ)πC(θ) f (θ)dθ ge 0

intΘD

gD(θ)πD(θ) f (θ)dθ ge 0

Thus a separating equilibrium can be constructed if the above conditions hold without rely-ing on Lemma 1 This separating equilibrium does reveal information about the type of Ito Government but in a somewhat less easily interpretable fashion Nonetheless the equi-librium described will only hold if a sufficiently positively weighted distribution of typesis screened to DS to overcome its relatively low chances of success compared to CV D andGovernmentrsquos preference for CV D over DS in this variant of the model

36

Since cC and cD are identical this suggests that as θ rarr w UI(DS) gt UI(CV D) sincevD(θ) gt vC(θ) forallθ isin Θ Thus since we have earlier assumed that there exists θ such thatUI(DS|θ)lt 0 which implies that UI(CV D|θ)gtUI(DS|θ) by the continuity in θ of vC and vDand the intermediate value theorem there must exist some θ such that UI(DS)=UI(CV D)

However in order to establish that this point w is a cutpoint we must further establish thatall types above that point prefer DS and all types below that point prefer CV D To ensurethis consider that I prefers DS over CV D whenever

vD(θ)πD(θ)minus cD minusvC(θ)πC(θ)+ cC gt 0

Which if cC = cD

harr vD(θ)πD(θ)minusvC(θ)πC(θ)gt 0

The final condition of Lemma 1 ensures that this expression is strictly monotonically in-creasing in θ consequently the inequality is satisfied for all θ types above w and not satis-fied for all θ types below w Thus w is a cutpoint

Proof of Proposition 8This proof largely follows from the discussion in text and the proof for Lemma 1 Lemma1 establishes that I will adopt the strategy outlined in Proposition 8 in response to Propo-sition 8rsquos conjectured strategy by Government The additional conditions in Proposition 8ensure that Governmentrsquos strategy is a best response to I rsquos cutpoint strategy by statingthat the payoff to Government of accepting claims by the subset of types screened to eitherinstitution is still higher than what is obtained by rejecting all claims to that institution

However some discussion of the role of Lemma 1 is warranted Without the conditions ofLemma 1 a separating equilibrium is still possible but it has a somewhat less clean inter-pretation Consider that without the monotonicity assumption from Lemma 1 there neednot be a cutpoint above which types select into the more stringent institution indeed whilehigher types will have higher valuations from DS and higher probabilities of success thanlower types it would be possible that their probability of success and valuations might in-crease faster for CV D leading to a peculiar situation in which higher types might select intothe less stringent institution In this case instead of a cutpoint strategy we can constructthe following strategy for I

σI =

DS if vD(θ)πD(θ)minus cD ge vC(θ)πC(θ)minus cC

CV D if vD(θ)πD(θ)minus cD lt vC(θ)πC(θ)minus cC

Noting that depending on the shape of the probability and benefit functions these two dis-joint sets could be any union of subsets of the type space If we call the DS subset ΘD and

35

the CV D subset ΘC we can consider the following strategy by Government

σG =

Accept if I pursues DSAccept if I pursues CVD

Which will be a best response whenever the following conditions holdintΘC

gC(θ)πC(θ) f (θ)dθ ge 0

intΘD

gD(θ)πD(θ) f (θ)dθ ge 0

Thus a separating equilibrium can be constructed if the above conditions hold without rely-ing on Lemma 1 This separating equilibrium does reveal information about the type of Ito Government but in a somewhat less easily interpretable fashion Nonetheless the equi-librium described will only hold if a sufficiently positively weighted distribution of typesis screened to DS to overcome its relatively low chances of success compared to CV D andGovernmentrsquos preference for CV D over DS in this variant of the model

36

the CV D subset ΘC we can consider the following strategy by Government

σG =

Accept if I pursues DSAccept if I pursues CVD

Which will be a best response whenever the following conditions holdintΘC

gC(θ)πC(θ) f (θ)dθ ge 0

intΘD

gD(θ)πD(θ) f (θ)dθ ge 0

Thus a separating equilibrium can be constructed if the above conditions hold without rely-ing on Lemma 1 This separating equilibrium does reveal information about the type of Ito Government but in a somewhat less easily interpretable fashion Nonetheless the equi-librium described will only hold if a sufficiently positively weighted distribution of typesis screened to DS to overcome its relatively low chances of success compared to CV D andGovernmentrsquos preference for CV D over DS in this variant of the model

36