Many of the most dramatic and memorable moments of the recent ½-nancial crisis involved the failures ornear-failures of some of the nation’sbiggest ½nancial institutions: BearStearns, Lehman Brothers, and aig, to name a few. Much of the subsequentpolicy response has been shaped by adesire either to avert such failures inthe future–for example, by imposinghigher capital requirements on system-ically important ½nancial ½rms–or tolessen the adverse consequences of fail-ures if they do occur–for example, byimproving the methods available to re-solve large institutions in distress.

Yet from the perspective of credit creation and impact on the rest of theeconomy, one of the most damaging as-pects of the crisis was not just the prob-lems of these big ½rms, but also the col-lapse of an entire market, namely the mar-ket for asset-backed securities (abs). For example, the market for so-calledtraditional or consumer abs–thosebased on credit-card, auto, and studentloans–averaged between $50 and $70billion of new issuance per quarter in the years leading up to the crisis. (Thetotal issuance for calendar year 2007 was $238 billion.) However, in the last

quarter of 2008, following the bankrupt-cy of Lehman, total issues in this catego-ry fell to slightly more than $2 billion.Given that banks were suffering theirown problems and were not easily ableto step into the breach, the disappear-ance of this market represented a majorcontraction in the supply of credit toconsumers, and may well have played acentral role in the steep drop in aggre-gate consumption that occurred at thistime. The traditional abs market onlybegan to rebound in mid-2009, with theimplementation of the Federal Reserve’sTerm Asset-Backed Lending Facility, ortalf, which made tens of billions of dol-lars of Federal Reserve loans available onattractive terms to investors seeking tobuy newly issued consumer abs.

In what follows, I explore the role thatthe abs market plays in the broaderprocess of credit creation, focusing onfour sets of issues. First, I describe howthe market works: how pools of loans(for example, mortgages or credit-cardand auto loans) are packaged and struc-tured into abs and how investors suchas hedge funds, pension funds, and bro-ker-dealer ½rms ½nance the acquisitionof these abs. Second, I outline the eco-nomic forces that drive securitization;these include both an ef½ciency-enhanc-ing element of risk-sharing and a less

Jeremy C. Stein

Securitization, shadow banking & ½nancial fragility

© 2010 by Jeremy C. Stein

Dædalus Fall 2010 41

desirable element of banks trying tocircumvent regulatory capital require-ments. Third, I discuss the factors thatappear to have contributed to the mar-ket’s fragility during the crisis period.Finally, I suggest some policy reformsthat might help mitigate this fragilitygoing forward.

Suppose you buy a new car and takeout a $10,000 loan to ½nance the pur-chase. This loan could come from a bank or from the ½nancing arm of anauto company. After the loan is made,the lender has two options: it can holdon to the loan it has originated, or it can securitize it and thereby sell it off. In the ½rst step in the securitization pro-cess, known as pooling, the lender gath-ers your loan with tens of thousands ofother loans like it–from other cars soldaround the country at about the sametime–and assembles these loans togeth-er in a trust. The cash payments comingfrom all the designated loans from thatpoint forward go into the trust.

The second step in the process is tranch-ing. Tranching involves designating dif-ferent classes of claimants on the cashflows to the trust, some of whom are giv-en higher priority than others. Said dif-ferently, tranching spells out who losesmoney and under what conditions ifsome of the loans that make up the poolgo bad. Consider this simpli½ed exam-ple: a pool of ten loans of $10,000 eachall come due at the same time. Each bor-rower can either pay off his loan in itsentirety, or default and pay nothing. One possible structure would divide thepool into ten layers, or tranches, each ofwhich is owed $10,000. The most junioror lowest priority tranche–call it T1–would recoup its money only if all tenloans were repaid; if even a single loanwent bad, T1 would see its investmentwiped out. Thus, T1 stands at the bottom

of the pecking order and is a very riskysecurity. The second most junior tranche,T2, regains its money if nine or more ofthe loans are repaid; equivalently, it getswiped out if there are two or more de-faults. At the top of the hierarchy standsT10, the most senior tranche, which isvery well protected and loses only if allten of the loans in the trust go bad.

As the example suggests, the senior-most tranches of securitizations arelikely to have high credit quality. That is, they experience losses only under rare circumstances, when a large frac-tion of the loans in the underlying poolis hit with defaults. One reflection ofthis tendency is the aaa rating typical-ly assigned to these senior-most tranch-es by the rating agencies. This rating inturn makes them attractive investmentsfor institutional investors that are look-ing for safe places to put their money butare either unable or unwilling to expendthe resources required to do loan-leveldue diligence. Such institutions may nothave the expertise to evaluate individualapplicants for auto loans, but given thereduction in credit risk associated withthe process of pooling and tranching,may be comfortable buying the seniortranches of auto-loan securitizationseven after relatively little investigativeeffort.

While the subprime crisis has calledinto question the whole idea of turningrisky loans into apparently near-risklessaaa-rated abs–a process many havedubbed “alchemy”–the flaw is not somuch with the basic concept of securi-tization, which has been around for along time, but rather with the recklessand excessively complex way in which itwas applied to subprime mortgage loans.The problem in the subprime sphere wasnot that the most senior T10-like tranch-es of subprime pools were rated aaa; itwas that many of the less well-protect-

42 Dædalus Fall 2010

Jeremy C.Steinon the½nancialcrisis &economicpolicy

ed tranches (say the T3s) were as well,though only after a series of machina-tions that were less than transparent to most market participants.1

Figure 1 provides some perspectiveon this point. It plots quarterly issuanceof abs over the period 2000 to 2009,broken into two categories. The ½rstcategory, “traditional” abs, comprisessecuritizations based on consumer cred-it: auto, credit-card, and student loans.The striking characteristic of this seriesis how stable it is in the several years pri-or to the crisis; one gets the sense of amarket that has functioned steadily andunremarkably, with only modest trendgrowth, for a long period of time. Thesecond category, “nontraditional” abs,includes securitizations based on sub-prime mortgage loans as well as second-and third-generation resecuritizations, inwhich the collateral going into the trustis not a pool of actual loans, but rather acollection of tranches created from earli-er rounds of securitizations of subprimeloans and other assets.2 This latter, morehighly engineered market shows the signsof a bubble, with the volume of issuancegrowing explosively in the period 2003to 2007, before completely collapsingduring the crisis period.

Beyond pooling and tranching, a ½-nal important element in the securiti-zation process is maturity transforma-tion. Often, the investors that purchaseabs tranches rely heavily on borrowedmoney, with much of this borrowingbeing very short-term in nature. In theperiod leading up to the crisis, entitiesknown as “structured investment vehi-cles” (sivs) or “conduits” were promi-nent examples of this behavior. Theseentities, which in many cases were af-½liated with sponsoring commercialbanks, held abs tranches and ½nancedthose holdings by issuing commercialpaper, which typically had a maturity

of only days or weeks and thereforehad to be rolled over frequently. Anoth-er example came from the hedge fundsand the broker-dealer ½rms (like BearStearns and Lehman Brothers) that½nanced their holdings of abs withrepurchase agreements (commonly known as “repo”), which are a form of overnight collateralized borrowing.

Collectively, the various investorsthat acquire abs and ½nance them withshort-term borrowing are often referredto as the shadow banking system. Just astraditional commercial banks invest inlong-term loans and ½nance these loansby issuing short-term deposits, the shad-ow banking system invests in securitiesbased on the same sorts of long-termloans (for example, mortgages or autoloans) and ½nances this investment byissuing short-term claims such as com-mercial paper or repo. On the one hand,the shadow banking system performsan economic function that looks muchlike that performed by the traditionalbanking system: it borrows on a short-term basis to fund longer-term loans;this is what is meant by maturity trans-formation.3 On the other hand, it doesnot face the same set of regulations be-cause the institutions involved are gen-erally not banks per se. And it does notbene½t from the same safety net. Forexample, unlike bank deposits, theshort-term ½nancing used by shadowbanks is not federally insured. Nor doshadow-banking players typically havethe right to borrow from the FederalReserve’s discount window.

Not all investors that hold abs ½nancethese holdings with short-term borrow-ing. Anecdotal evidence suggests thatpension funds and insurance companiestend to hold these securities on an unlev-ered basis, that is, without borrowingagainst them. Remarkably, however, lit-tle is known about the relative magni-

Dædalus Fall 2010 43

Securiti-zation,shadowbanking &½nancialfragility

tudes. In particular, as far as I am aware,there are no good empirical estimates to address the following question: whatfraction of the aaa-rated abs in a givenloan category (credit cards, for example)is owned by investors that ½nance theirholdings of these securities primarilywith short-term debt? This is an impor-tant question because, as argued below,the heavy use of short-term debt ½nanc-ing by abs investors contributes to thefragility of the market.

With securitization, loans originated by banks are packaged and sold off to avariety of other end investors. What eco-nomic forces encourage banks to off-load their loans in this way, as opposedto keeping them on their books? Thereare two broad stories that one can tell.

The ½rst, more benign story dependson the principle of risk-sharing. Whenbanks sell their loans into the securitiza-tion market, they distribute the risks as-sociated with these loans across a widerrange of end investors, including pen-sion funds, endowments, insurance com-panies, and hedge funds, rather than tak-

ing on the risks entirely themselves. Thisimproved risk-sharing represents a realeconomic ef½ciency and lowers the ulti-mate cost of making the loans. Moreover,as noted above, the pooling and tranch-ing process, if done properly, makes thesenior tranches of abs relatively easy toevaluate, even for nonspecialized invest-ors that do not have much ability to judgethe credit quality of the individual loansthat underlie these securities. Therefore,it is particularly conducive to expandingthe buyer base.

According to this story, the securitiza-tion of consumer loans and mortgagesclosely parallels the natural transitionthat many growing companies makewhen they reach a certain size and repu-tation and begin to shift their borrowingaway from exclusive reliance on banksand toward the corporate bond market.In either case, if the securities in ques-tion (either abs or corporate bonds) can be easily evaluated by a broader set of investors, it makes sense to tap intothis broader market, as opposed to rely-ing exclusively on the banking sector for ½nancing.

44 Dædalus Fall 2010

Jeremy C.Steinon the½nancialcrisis &economicpolicy

Figure 1Quarterly Issuance of Asset-Backed Securities, 2000–2009

Figure created using data from Thompson sdc.

While this wholesome-sounding sto-ry undoubtedly captures some of whatdrives the securitization market, it isalso incomplete. It has become apparentin recent years that another importantdriver of securitization activity is regula-tory arbitrage, a purposeful attempt bybanks to avoid the rules that dictate howmuch capital they are required to hold.Particular attention in this regard has fo-cused on the bank-sponsored sivs andconduits mentioned above, vehicles thatheld various types of abs and ½nancedthese holdings largely with short-termcommercial paper. What is striking aboutthese shadow-banking vehicles is thatmany of them operated with strong guar-antees from their sponsoring banks. In-deed, when the sivs and conduits gotinto trouble, the banks honored theirguarantees, stepping up and absorbingthe losses.

This outcome runs counter to the spir-it of the risk-sharing story, since ratherthan widely distributing the risks asso-ciated with the abs they created, banksretained them, albeit in an opaque, off-balance-sheet fashion. The most obvi-ous alternative explanation is that banksexploited a regulatory loophole: if theyheld the loans directly on their balancesheets, they faced a regulatory capital re-quirement on these loans; but if they se-curitized the loans and parked them inan off-balance-sheet vehicle (albeit onewith essentially full recourse to the banksin the event of trouble), the regulatorycapital requirement was much reduced.4

While this particular loophole will nodoubt be closed going forward, the moregeneral concern remains. Securitizationand the shadow banking system enablebank-like maturity-transformation activ-ities–speci½cally, the pairing of long-term assets with a short-term fundingstructure–to take place out of the reachof traditional banking regulation. To the

extent that bank regulation is burden-some, it creates a powerful pressure forbanking assets to be securitized and tomigrate out of the formal banking sys-tem. Absent some form of harmoniza-tion that puts shadow banks and tradi-tional banks on more of an equal regu-latory footing, this pressure is likely tointensify as capital requirements onbanks are raised in the wake of the crisis.

Figure 1 illustrates the complete melt-down of the abs market during the ½-nancial crisis, with issuance in both thetraditional consumer (auto, credit-card,and student loans) and nontraditional(subprime) categories falling essentiallyto zero. The nontraditional market start-ed to come apart ½rst, in August 2007, asthe extent of losses on subprime loansbecame more apparent. The traditionalconsumer market held up better for atime, but then also disappeared in thewake of the failure of Lehman Brothersin September 2008.

While less spectacular from a quan-titative perspective, the decline of thetraditional consumer abs market is in many ways the more challenging phenomenon to explain. The demise of the subprime-related abs market represented the deflation of a classicbubble; many of the loans involved were so poorly underwritten that theyshould never have been made in the½rst place. Not surprisingly, when mar-ket participants ½nally began to under-stand this point, the issuance of newsubprime loans dried up.

It is more surprising that abs issu-ance related to auto, credit-card, andstudent loans was hit so hard. There ismuch less reason to believe that thesewere bad loans to begin with: again,the super½cial evidence in Figure 1 does not suggest a bubble in this part of the market. Indeed, overall lending in

Dædalus Fall 2010 45

Securiti-zation,shadowbanking &½nancialfragility

these categories did not completely van-ish in the same way that it did in the sub-prime area; rather, some fraction of thislending reverted back to being done ina nonsecuritized fashion by the banks,which suggests that they still viewed theloans as worth making. However, giventhe limited capacity of the banks, whosecapital by this point was badly impaired,the inability to securitize and therebyoff-load some of their loans no doubtcontributed to a sharp contraction in the overall supply of credit available to consumers.

If the underlying auto, credit-card, andstudent loans were still worth making,what caused the market for abs basedon these loans to contract so sharply? Aprominent emerging hypothesis is that,effectively, there was the analogue of awidespread bank run on the shadowbanking system.5 Recall that many absinvestors ½nance their positions withshort-term borrowing, either in the form of commercial paper or overnightrepurchase agreements. In this sense,they are very much like banks, which½nance long-term loans with short-term deposits. But unlike bank depos-its, the short-term ½nancing that sup-ports the abs market is not insured by the government. This differencemakes the shadow banking system vul-nerable to something that looks like aclassic bank run from the days beforedeposit insurance: as short-term lend-ers lose con½dence and refuse to rollover their loans, investors in abs areforced to liquidate some of their hold-ings to come up with cash. The liquida-tions in turn depress the price of theseabs via a “½re sale” effect. Moreover,short-term lenders view abs as lessattractive collateral as their prices falland become more volatile. The lendersthen pull back even further, leading toanother round of liquidations and price

declines. Once under way, this viciouscycle is very dif½cult to arrest.

One concrete manifestation of thedramatic withdrawal of short-termlending to the abs market comes fromthe behavior of what are called “hair-cuts” in repurchase agreements. When an investor borrows from the repo mar-ket to ½nance its holdings of abs, it isrequired to post a margin, or down pay-ment; this is the haircut. Haircuts onabs were extremely low prior to the cri-sis, on the order of 2 percent. What thismeans is that if, say, a hedge fund want-ed to acquire $1 billion of auto-linkedabs, it only needed to put up $20 mil-lion of its own capital as a down pay-ment. The other $980 million could beborrowed on an overnight basis in therepo market; in many cases, the ulti-mate lenders of this short-term mon-ey were money-market mutual fundslooking to ½nd slightly higher yieldingshort-term investments than, for exam-ple, Treasury bills.

In the midst of the crisis, haircuts sky-rocketed. Even haircuts on traditionalconsumer abs–those not linked in anydirect way to the housing and subprimeproblems–rose to more than 50 percent.6From the perspective of the hedge fundholding $1 billion of auto-linked abs,suddenly it could borrow only $500 mil-lion, and instead of having to post a $20million down payment, it now had to post$500 million. If it did not have the cash onhand, it was forced to liquidate its hold-ings. These forced liquidations, and thepowerful impact they had on both thelevel and volatility of abs prices, in turnjusti½ed the increased skittishness of thelenders in the repo market, because theirprotection was entirely predicated on thecollateral value of the assets they werelending against.

The bank-run analogy offers what feelslike a compelling account of the fragility

46 Dædalus Fall 2010

Jeremy C.Steinon the½nancialcrisis &economicpolicy

of the securitization market. However, itwould be premature to call it a fully em-pirically validated explanation for whythe market dried up so dramatically dur-ing the crisis. For one, as emphasizedabove, it is not known what fraction ofabs was held by investors that ½nancedthemselves in a vulnerable bank-like way–that is, largely with short-term debt. If the fraction turns out to have been,say, 40 percent instead of 80 percent,this ½nding would temper the force of the theory.

There is an alternative, more behav-ioral hypothesis for the fragility of thesecuritization market that does not relyon a predominance of short-term debt½nancing.7 This alternative hypothesisbegins with the observation that a largeproportion of abs tranches–in boththe traditional and subprime sectors–was rated aaa. The aaa rating may haveencouraged investors such as pensionfunds or insurance companies to thinkof these securities as essentially risklessand therefore to treat them as equivalentto Treasury bonds when constructingtheir portfolios. When the problems inthe subprime area became apparent, thispremise was utterly destroyed, and in-vestors that were determined to allocatea fraction of their portfolios to safe as-sets realized that they had to dump theirholdings of aaa-rated abs and buy ac-tual Treasuries instead. Thus, instead of a short-term-debt-driven bank run,we have what might be called a wide-spread buyer’s strike. In this account,the rating agencies’ failures with re-spect to the subprime market under-mined their credibility more generally,so that any aaa-rated tranche of an abs, be it linked to subprime or creditcards, was no longer considered a vir-tually riskless asset.

Of course, the two theories are notmutually exclusive, and may interact in

interesting ways. For example, what be-gins as a simple strike on the part of un-levered buyers may evolve into a run-likephenomenon since the buyer’s strikeputs downward pressure on abs prices,making the position of short-term lend-ers more precarious and thereby encour-aging these lenders to withdraw fromthe market.

To frame the policy issues with respectto securitization and the shadow bank-ing system, it is useful to begin by em-phasizing three key points. First, we arealmost certainly heading in the directionof imposing signi½cantly higher capitalrequirements on large banks. Second,while this is undoubtedly a valuable andmuch-needed reform, and one that holdsthe promise of making the banking sec-tor itself more robust in future episodesof ½nancial volatility, it will also have theeffect of encouraging more credit-cre-ation activity to migrate away from thebanks and toward the shadow-bankingsector, in an effort to evade the burdensassociated with more stringent regula-tion. Third, we have seen that the shad-ow-banking sector can be a powerfulsource of fragility in its own right, onethat can lead to damaging disruptions in the flow of credit to households andbusinesses. Thus it would be a mistaketo pursue a set of policies that focusesheavily on insulating our large banksbut that pays insuf½cient attention topotential vulnerabilities in the rest of the ½nancial system. Rather, the goalshould be a balanced approach that ad-dresses all elements of the system inan integrated fashion.

What concrete steps might be taken inthis regard? I present three speci½c ideasfor regulating the securitization and shad-ow-banking markets. To be transparentabout my own prejudices, I label theideas the good, the bad, and the maybe.

Dædalus Fall 2010 47

Securiti-zation,shadowbanking &½nancialfragility

The good: regulation of haircuts in the abs market. To mitigate the incentivesfor regulatory evasion, and to help re-duce the fragility of credit creation nomatter where it takes place, a systemat-ic effort must be made to impose simi-lar capital standards on a given type ofcredit exposure, irrespective of wheth-er it is a bank, a broker-dealer ½rm, ahedge fund, or any other entity that ends up holding the exposure. This is not an easy task, but one tool thatwould help is broad-based regulation of haircuts (that is, minimum marginrequirements) on abs.8

Consider the case of a consumer loan.If this loan is made by a bank, it will besubject to a capital requirement; thatis, the bank will have to put up someamount of equity against the loan, rath-er than borrow all the money. Now sup-pose instead that the loan is securitizedby the bank and becomes part of a con-sumer abs whose tranches are distrib-uted to various types of investors. Theregulation I have in mind here wouldstipulate that whoever holds a tranche of the abs would be required to post aminimum down payment against thattranche–with the value of the haircutdepending on the seniority of the tranche,the quality of the underlying collateral,and so forth.

For example, before the current crisis,market-determined haircuts on aaa-rated consumer abs tranches were verylow, in the neighborhood of 2 percent.With no further regulation, they are like-ly to return to these levels as markets re-normalize. However, the new regulationmight instead impose a minimum hair-cut requirement on aaa-rated consum-er abs of at least 10 percent, indepen-dent of market conditions. That is, anyinvestor in such a security would be re-quired to post and subsequently main-tain a 10 percent margin at all times.

Such a requirement is nothing conceptu-ally new and should not be dif½cult toenforce; indeed, it is closely analogousto the initial and maintenance marginrequirements that are currently applica-ble to investors in common stocks.

If well structured, these haircut re-quirements have two important bene-½ts. First, they go a long way towardachieving harmonization across organi-zational forms, in that there would nolonger be an obvious regulation-avoid-ance motive for moving the consumerloan off the balance sheet of the bankand into the shadow-banking sector.This bene½t is especially important aswe move toward signi½cant increases in the capital requirements imposedon banks. The goals of these higherbank-capital requirements are likely to be partially frustrated if they drivesigni½cant amounts of activity outsidethe banking system.

Second, for that portion of credit-creation activity that does end up re-siding in the shadow-banking sector,haircut regulation can help dampenthe bank-run-like crisis dynamics de-scribed above. The problem is that ifhaircuts begin at 2 percent before the crisis, and then jump to more than 50percent during the crisis, this increasecreates a powerful forced-selling pres-sure on the owners of abs. If the hair-cuts are instead set at a more prudentvalue before the crisis–again, say 10 percent–so that investors are requiredto put up more of their own cash at theoutset, this forced-selling mechanism,and the vicious spiral it unleashes, might be substantially attenuated.

The bad: extension of the federal safety net to shadow banks. Some observers havetaken the analogy between the tradition-al commercial banking sector and theshadow banking system one step further,arguing that in order to prevent run-like

48 Dædalus Fall 2010

Jeremy C.Steinon the½nancialcrisis &economicpolicy

panics in the latter, it should be coveredby the same federal safety net as the for-mer. This coverage would entail givingshadow-bank entities access to the Fed-eral Reserve’s discount window, as wellas possibly insuring some of their short-term debts. Thus, when a specialized in-vestment vehicle is set up to buy a port-folio of abs ½nanced largely with short-term commercial paper borrowing, thecommercial paper issued by the invest-ment vehicle might be explicitly feder-ally insured, much as some bank depos-its are today. Instead of trying to leanagainst the private market’s propensityto ½nance abs with large amounts ofshort-term debt–as the haircut regula-tion described above would do–this al-ternative approach amounts to embracingthe use of such short-term ½nancing andattempting to use government insuranceto make the world safer in its presence.

What makes this policy unattractive is the moral hazard that it invites, as pri-vate actors seek to exploit government-provided insurance by using it to ½nanceriskier-than-expected activities. This con-cern is particularly acute when the insur-ance is attached to the kinds of highlyengineered ½nancial products that wereheld by some shadow-banking investorsprior to the current crisis–products forwhich risks are often not easily under-stood or accurately measured ahead oftime. For example, one can imagine agovernment insurer trying to devise aformula for risk-based pricing of the in-surance it provides to a specialized in-vestment vehicle, in an effort to deterexcessive risk-taking. But should we ex-pect any such formula to do better thanthose of the rating agencies, which sospectacularly misjudged the risks em-bedded in complex abs based on sub-prime mortgages? Indeed, one can arguethat the mind-bending complexity ofsome of these structures emerged pre-

cisely as a means of gaming the rating-agency formulas. Thus, although a gov-ernment insurance agency would notface the same overt conflicts of interestas the rating agencies, it seems reason-able to worry about how it would farewhen pitted against Wall Street’s best½nancial engineers.

With this bit of pessimism in mind, I would argue that in order to enter-tain the idea of expanding the safety net, one would have to believe that theshort-term debt claims created by theshadow-banking sector are of substan-tial social value–so much so that sus-taining them with moral-hazard-proneinsurance, rather than trying to con-strain them with haircut regulation, is a ½rst-order imperative. I don’t thinkthat we have nearly enough empiricalevidence to meet this burden of proof.Hence I would be strongly inclined tosteer clear of any expansion of the safe-ty net.

The maybe: limiting the creation of “pseu-do-riskless” securities. As discussed above,an alternative theory for the fragility ofthe abs market during the crisis is that,even absent short-term debt ½nancingof abs positions, the proliferation of somany “pseudo-riskless” securities is in-herently dangerous. By pseudo-riskless, I mean aaa-rated securities that appearso safe in good times that investors arelulled into a sense of complacency where-by they treat these securities as equiva-lent to truly riskless Treasuries. Only in acrisis do investors discover that this was a false equivalence, which leads them topanic and dump their holdings of theaaa-rated securities.9

If one takes this point of view, it istempting to think about ways to con-strain the production of those abstranches that can be represented to in-vestors as being near-riskless. One waymight be to require the credit-rating

Dædalus Fall 2010 49

Securiti-zation,shadowbanking &½nancialfragility

agencies to use a coarser set of ratingswhen evaluating abs than when theyevaluate corporate bond issues. For ex-ample, instead of a ½nely tuned scalethat goes from aaa to aa+, aa, aa-, a+, and all the way down to ccc, the ratings for abs might be restricted toone of three broad buckets: a, b, or c.10

While this idea admittedly has a bit ofthe feel of deploying the language police,it might prevent any abs tranche frombeing thought of as near-riskless, sinceeven the highest rating category wouldnow encompass securities with a widerange of credit qualities.

An alternative approach would be toleave the current ratings categories inplace but to impose on the creators ofany abs an upper limit on the amount of highly rated securities that they couldmanufacture from any given underly-ing pool of loans. For example, one rulemight be that only a maximum of 50percent of the dollar value of tranchescoming from any pool of consumerloans could ever seek a rating of aa orhigher; all other subordinate trancheswould have to be targeted at lower rat-ings categories.

I put this last set of ideas in the “maybe”category because I view them as interest-ing and worthy of further thought, but Iam not at this point con½dent that theirvirtues outweigh their potential for un-intended consequences. On the one

hand, they highlight the logical impli-cations of taking a more behavioral per-spective on the abs market’s fragility–of positing a world in which investorsare overly prone to seek out pseudo-risk-less investments and in which ½nancialinnovators actively try to exploit thistendency. On the other hand, the spe-ci½c proposals I have sketched raisesome fairly obvious flags as well. Forexample, restricting the vocabularyavailable to the rating agencies mayhave meaningful effects in the short run, but over time it is easy to imagineindustry conventions evolving so asto work around any such restrictions.If so, it would be a mistake to placemuch long-term faith in this approach.

The overarching goal of ½nancialreform must not be only to fortify a set of large institutions, but rather toreduce the fragility of our entire systemof credit creation. This system involves a complicated interplay between banksand non-banks and between tradition-al forms of lending and securitization.Thus far, more effort has been devotedto the banking side of the equation. Thisimbalance is perhaps not surprising giv-en the accumulated expertise of manyof the regulators involved in the reformprocess. But the dif½cult issues associat-ed with securitization and the shadowbanking system demand equal attention.

50 Dædalus Fall 2010

Jeremy C.Steinon the½nancialcrisis &economicpolicy

endnotes1 More precisely, the riskier lower-rated tranches of subprime securitizations were them-

selves used as the raw material (in place of the original mortgage loans) to create second-and third-generation resecuritizations. Many of the biggest problems in the crisis arosefrom the fact that large fractions of these resecuritized vehicles were also rated aaa, inspite of the dubious collateral supporting them. This is where the most extreme alchemycan be said to have taken place. For a discussion, see Joshua Coval, Jakub Jurek, and ErikStafford, “The Economics of Structured Finance,” Journal of Economic Perspectives 23(2009): 3–25.

Dædalus Fall 2010 51

Securiti-zation,shadowbanking &½nancialfragility

2 The data in the ½gure come from Thompson sdc. While the “nontraditional” categoryincludes securitizations based on subprime mortgage loans, it does not include securitiza-tions based on prime mortgage loans, such as mortgage-backed securities guaranteed bythe government-sponsored enterprises Fannie Mae and Freddie Mac. I am grateful to Sam Hanson, who put together the data and shared it with me.

3 See Gary Gorton and Andrew Metrick, “Securitized Banking and the Run on Repo,” nber working paper 15223 (National Bureau of Economic Research, 2009).

4 For a detailed study of this phenomenon, see Viral Acharya, Phillip Schnabl, and GustavoSuarez, “Securitization without Risk Transfer,” nber working paper 15730 (National Bu-reau of Economic Research, 2010).

5 The bank-run hypothesis is articulated in Gorton and Metrick, “Securitized Banking andthe Run on Repo,” as well as in Daniel Covitz, Nellie Liang, and Gustavo Suarez, “TheAnatomy of a Financial Crisis: The Evolution of Panic-Driven Runs in the Asset-BackedCommercial Paper Market,” working paper (Board of Governors of the Federal Reserve,2009).

6 For more detail on the evolution of repo-market haircuts during the crisis period, see Gorton and Metrick, “Securitized Banking and the Run on Repo.”

7 A version of this hypothesis is presented in Nicola Gennaioli, Andrei Shleifer, and RobertVishny, “Financial Innovation and Financial Fragility,” working paper (Harvard Universi-ty, 2010).

8 Such haircut regulation is discussed in John Geanakoplos, “The Leverage Cycle,” discus-sion paper 1715R (Cowles Foundation, 2010); and in Jeremy Stein, “Monetary Policy asFinancial-Stability Regulation,” working paper (Harvard University, 2010).

9 Again, see Gennaioli, Shleifer, and Vishny, “Financial Innovation and Financial Fragility,”for elaboration on this argument.

10 This scale is used by both Standard and Poor’s and Fitch. The other major rating agency,Moody’s, has a somewhat different alphanumeric convention, albeit with similarly ½ne-grained categories: Aaa, Aa1, Aa2, Aa3, A1, A2, etc.