Over-Declaration of Standard Essential Patents andDeterminants of Essentiality∗

Marc Van Audenrode†, Jimmy Royer‡,Robin Stitzing§, Pekka Sääskilahti¶

March 31, 2017

Abstract

Standard Essential Patents (SEPs) cover technologies necessary to meet industrystandards that are established in Standard Setting Organizations (SSOs), but not allpatents declared as SEPs are actually essential to the implementation of the standardor even covered by the standard specifications. SEPs are self-declared by patent-holdingcompanies as essential for the implementation of the standard, and are not subject toany SSO review. Loss of enforceability due to failure to declare and the anticipation ofa stronger bargaining position in SEP license negotiations create incentives for firms toover-declare patents or to at least to err towards declaring when in doubt. SSO IPRpolicies contribute to over-declaration, because patent holders are required to declare allpatents that might be essential. Changes to patent claims and standard specificationsduring the patent application prosecution can also result in that a declared patentapplication no longer matches the standard once granted and thus the declared SEPbecomes non-essential. We analyze patent and company attributes that are associatedwith technical essentiality among SEPs of the 4G LTE cellular standard using a uniquedataset. Our results show that a declaration against a specific technical specificationdocument of the standard is a strong predictor of technical essentiality. While thereis no general link between forward citations and technical essentiality, we find thatcitations from SEPs declared to the same standard predict technical essentiality. Patent

∗The views presented in this paper are those of the authors and do not necessarily reflect those of ouremployers or their clients. We thank PA Consulting for providing the data for this study. We are very gratefulto Mika Anttila, Amitabh Dixit, Heikki Einola, Mark McCarty, Matthew Noble, Tuomas Saukkonen, GarySavitt, Matt Urbanawiz, Harri Valio, and Niklas Östman for helpful comments and advice and Sean Burpoe,Alex Hippel, and Jee-Yeon Lehmann for help with the data. Any errors or omissions are our own.†Analysis Group, Inc.‡Analysis Group, Inc.§Nokia Corporation¶Nokia Corporation. Corresponding Author. Email: pekka.saaskilahti@nokia.com

1

ownership changes before the essentiality declaration do not affect standard essentiality.Our results provide guidance to the policy debate on standards, licensing of SEPs, andpatent system, and call for a recognition of SEP over-declaration in the economicsliterature on standardization and SEPs.

JEL Classifications: K11, K21, L44, L51, L52Key Words: standards; standard essential patents, intellectual property, innovation

2

1 Introduction

The development of technical standards have taken on increasing importance with the rise ofcomplexity and global reach of technology. By establishing rules and specifications for productsand services, standards facilitate interconnection and interoperability of related productsand yield important benefits for the functioning of markets and innovation. In particular,standards can create large, global markets by establishing a common set of characteristicswith which different producers can comply. Consequently, they generate economies of scale inproduction, reduce transaction costs, and foster competition by supporting market recognitionof products offered by new entrants. Moreover, by expanding the potential product andpatent licensing markets, standards encourage investment in the research and development(R&D) of new technologies.

Despite these clear benefits to static and dynamic efficiencies, standard setting requirescoordination by many stakeholders with competing interests, which is often difficult to achievethrough market mechanisms. To resolve such market failures, industry players and, in certaininstances, government entities have created standard setting organizations (SSOs), whoseprimary role is coordinating standards development. A critical aspect of standardization isto maintain a workable balance between the creation of incentives for innovation and thepromotion of competition and wide adoption of standards in the downstream product market.SSOs therefore generally require its members to both self-disclose and license patents that aretechnologically essential to the standard or “standard essential patents” (SEPs) to memberand non-member firms on fair, reasonable, and non-discriminatory (FRAND) terms.

Given their impact on competition, technology adoption, and innovation, the processof standard development and mechanisms for declaration and licensing SEPs have been ofgreat interest to policy makers and economists. Recent studies on standard setting havelargely focused on the impact of SSOs and SEPs on the development, adoption, and/ordiffusion of standards and technological change (e.g. Rysman and Simcoe 2008, Baron andPohlmann 2013b, Pohlmann, Neuhäusler and Blind 2016), the effect of certain SSO rules onstandardization and competition (e.g., Chiao, Lerner, and Tirole, 2006; Lemley and Shapiro,2013; Lerner and Tirole, 2006, 2015), or patent characteristics associated with disclosed SEPs(e.g. Bekkers, Bongard and Nuvolari 2011).

However, a notable but under-recognized shortcoming of these studies is that they take asgiven that patents disclosed as SEPs are truly essential to the standard. While the probalisticnature of patents (Farrell and Shapiro 2008) is accepted in the economic literature, and

3

oppositions and court invalidations have been linked to patent quality (e.g. Lanjouw andSchankerman 2001, Galasso and Schankerman 2015), the corresponding uncertainty related tostandard-essentiality has not attracted attention despite the importance of standards acrossmany industries. The premise of certain essentiality is highly problematic when one considersrecent observations by industry experts (e.g. Pohlmann 2016, Bekkers and Updegrove 2012)and policymakers (e.g. European Commission 2016a, European Commission 2016b) thatcurrent SSO policies create incentives for patent owners to “over-declare” or to “pad” theirpatents as being essential to the standard. These critics point to a number of factors relatedto the lack of clarity in the declaration process, absence of screening mechanisms, and thepositive relationship between the number and share of declared SEPs and licensing royaltyrevenue. For example, the European Commission (2014) suggests that over-declarationrates of SEPs could be as high as 80 percent. Dewatripont and Legros’s (2013) theoreticalanalysis highlights that incentives to over-declare SEPs likely differ across different kind ofinnovators. However, save for Lerner, Tabakovic and Tirole (2016), there is little work onSEP over-declaration despite the growing importance and proliferation of standards, andthere is no empirical analysis taking advantage of the available technical assessments of actualessentiality that can shed light on the over-declaration issue directly.

In this study, we aim to address this critical gap in literature by providing the firstrigorous empirical analysis of patent and owner characteristics associated with SEPs that areevaluated to be actually essential to the declared standard, as determined by independentevaluation of technical essentiality (“judged essentiality”). Although the question of whetheran SEP is truly essential or not is ultimately a legal one that can only be determined ina court of law, judged essentiality can provide a useful proxy for the true essentiality ofSEPs. We focus on SEPs disclosed to the European Telecommunications Standards Institute(ETSI) for the 4th-generation cellular telecommunications standard, LTE.1. We combinedata from ETSI on the characteristics of LTE standard essential patents with independentassessment of their technical essentiality, to assess the prevalence of over-declaration and toanalyze patent characteristics that are associated with true essentiality among LTE SEPs.Our paper complements the recent work by Lerner et al. (2016) who focus on patent holders’declaration strategies. They study under which circumstances declarations are made so thatindividual patents are identified in the declaration, and under which circumstances firms

1ETSI is part of the 3rd Generation Partnership Project (3GPP), a group of regional SSOs, responsiblefor the standardization of the fourth generation cellular telecommunications standard called LTE (Long TermEvolution).

4

make portfolio-wide blanket declarations covering all owned patents.Technical standardization is particularly important for the telecommunications industry

given the industry’s fundamental need for global interconnection and interoperability ofvarious products and services. Moreover, new entrants have displaced traditional playersin the downstream mobile handset market at increasing rates, creating a divide betweeninnovators and downstream implementers. These factors have led to difficult licensingnegotiations and a number of litigations over the validity and the essentiality of patents.2

Hence, it is especially pertinent and valuable to examine the issue of true essentiality in thetelecommunications industry.

Results of our logit regression analysis do not show that SEPs that were either traded ordeclared as essential to the standard by a non-practicing entity (NPE) would be more or lesslikely to be judged essential than other SEPs. However, our results clearly show that SEPsdeclared against a specific technical specification of the LTE standard are more likely to bejudged essential than SEPs declared with a reference to the general standard only. Declarationto a specific technical specification increases the likelihood of judged essentiality by 37.3percent in our preferred regression specification with the largest set of control variables. Thisfinding indicates that firms declaring against a specific standard document understand betterthe content of the standard and their patents and apply more rigorous internal reviews ofdeclarations. On the other hand, granted patents are around 7.7 percent less likely to bejudged essential than patents that have not yet been granted. One explanation for thisfinding is the time consuming patent application prosecution process that can cause claimtext changes so that an initially essential patent application no longer matches the standardonce granted. Likewise, standard specifications can also change during the prosecution ofa patent application. Patent forward citations, which are the most used proxy for patentquality in economic literature, are in general not correlated with essentiality. However, thecount of forward citations from other SEPs declared to the same standard is positivelycorrelated with essentiality, and this correlation is driven by citations from other SEPs judgedessential to the same standard. This finding indicates that once forward citations originatefrom the same standard, they explain technological closeness of the patents. In particular,forward citations from the same standard as a proxy of patent value relate primarily to thelikelihood of infringement of the SEPs by the standard specifications. Citations from other

2For example, Bekkers and West (2009) discuss Nokia vs. InterDigital. Nokia alleged in the UK that 31InterDigital SEPs declared to 3G UMTS were not essential to the standard. Prior to the trial, InterDigitaldropped essentiality claims for 26 of these patents.

5

patents declared and thus likely to be related to the same standard can be used therefore as apredictor of essentiality, even if their essentiality status of these SEPs has not been assessed.We also find that the shorter length of the first independent claim of the SEP predicts higherlikelihood of essentiality. The result is closely related to the finding on forward citations fromthe same standard, because claim length primarily explains the possibility of infringement bythe implementations of the standard specifications. We also investigate how patent validitymight be related to essentiality. We construct a variable that measures the similarity betweenthe first independent claims of US and EP patents belonging to the same patent family. Ifthese two claims are similar, it indicates that the underlying innovation meets legal standardsand the claim text has been well drafted. Such a claim is more likely to be found valid iftested. We find that claim similarity does not explain technical essentiality, which confirmsthe view that technical essentiality is primarily related to infringement possibility than toother factors of patent value.

Our findings are important for understanding the impact of standard setting on patentlicensing, technological change, and competition in the technology industry. First, pol-icymakers and economists have both stressed that information asymmetries in standardsetting, especially in the disclosure and licensing of SEPs, create a number of inefficienciesin the upstream and downstream markets. By providing a better understanding of theprevalence of the over-declaration of SEPs, ownership of truly essential SEPs, and patentcharacteristics associated with true essentiality, our findings can increase transparency andreduce uncertainty related to the licensing of SEPs and adoption of standards. Second, ourfindings can inform standard-setting organizations and policymakers in designing policiesto reduce over-declaration rates. For example, a requirement to make declarations againstspecific technical specifications could be made mandatory by SSOs. Third, our findingscan serve as the basis for a screening model of true essentiality, which will help to increaseefficiency of standardization without imposing the excessively burdensome requirement ofpatent-by-patent technical analysis. In addition, understanding the empirical determinantsof true essentiality can reduce uncertainty about true SEP holdings and thus improve SEPlicensing negotiations.

The remainder of the paper is organized as follows. Section 2 provides brief institutionalbackground on standard setting and intellectual property rights in SSOs, followed by Section3, which provides an overview of related literature on over-declaration and essentiality. Section4 describes the data, and Sections 5 to 7 present our empirical methodology and discuss theresults. Section 8 concludes and proposes avenues for future research.

6

2 Standard Setting Organizations and Standard Essen-tial Patents

2.1 The Role of Standard Setting Organizations

SSOs facilitate several critical functions in technology and market development: discovery,standardization, and self-regulation by establishing a regulatory frame with their IPR policies(Lerner and Tirole 2015). First, SSOs investigate and certify the technical merits of variouscombinations of functionalities. Second, following this evaluation process, SSOs select aparticular technology for a standard. Each SSO has its own set of procedures for introducing,adjusting, approving, publishing and revising candidate standards. However, most SSOsgenerally follow an “open-setting procedure,” in which the SSO provides information tointerested parties on the standard-setting and members attend meetings, vote on the stan-dardization decisions, and contribute technologies to the standard (Baron and Spulber 2015b).To support the adoption of the standard in the market, an SSO’s intellectual property rights(IPR) policy must balance market use of patents covering technologies that are essential tothe implementation of the standard (SEPs) with protection of the property rights of thesestandard essential patents and encouraging continued innovation.

Although an SEP is generally defined as a patent covering an invention essential to a givenindustry standard such that infringing on that patent is unavoidable in implementing thestandard, the definition of what is essential can differ across SSOs (Maskus and Merrill 2013).Most SSOs provide a “technical essentiality” criterion, defining essentiality to mean thereare no technological alternatives to implementing the standard. A few SSOs (e.g., IEEEand VITA), however, include the concept of “commercial essentiality” in their IPR policies,meaning that alternatives that do not infringe on the patent exist but are not commerciallyviable to bring to market. ETSI, the SSO associated with the SEPs analyzed in thispaper, explicitly rules out the concept of commercial essentiality in its IPR policies, definingessentiality solely on a technical basis.

2.2 The European Telecommunications Standards Institute (ETSI)and the 3GPP-LTE Standard

The issue of patent-based standardization is most prominent in the telecommunicationsindustry. At least seven SSOs exist in the telecommunications industry, including the

7

European Telecommunications Standards Institute (ETSI), the organization to which theSEPs analyzed in this study are disclosed. ETSI is an entity focused on information andtelecommunications technologies, responsible for developing the GSM and 3G UMTS/W-CDMA standards, and most recently, the LTE standard. The ETSI IPR policy applies to 4GLTE standard analyzed in this study. Over 800 member organizations across 66 countriesbelong to the ETSI, including many leading companies and R&D organizations. Since 2003,it has collaborated with the European Patent Office to provide public access to an IPRdatabase and library with over 35,000 technical contributions made to ETSI as part of thestandardization process.

The Long Term Evolution (LTE) standard is the 4th generation (4G) standard forhigh-speed wireless communication for mobile phones and data terminals. The standardis developed under the 3rd Generation Partnership Project (3GPP) which is a partnershipbetween seven regional SSOs and it defines and publishes the technical specifications of theLTE standard. The 3GPP Release 8 of the LTE standard was frozen in December 2008, andthe first LTE equipment were build based on this release version.3

2.3 Disclosure and Licensing of SEPs

SSO policies regarding SEPs comprise the rules governing disclosure and licensing commit-ments of its members. Most SSOs require their members to self-disclose patents that areessential to the standard or may become essential to when the final standard is adopted.According to Bekkers and Updegrove (2012), none of the major SSOs require their members toconduct a patent search for the disclosure process or stipulate a formal process for adjudicatingthe essentiality of patents. SSO disclosure requirements differ across several dimensions,including rules for the timing and the specificity of the disclosures. Although most SSOs callfor disclosures as early as possible, the meaning of what is early is oftentimes impreciselydefined. SSO policies also vary with regard to whether disclosures must identify specificpatents that are essential (“specific disclosures”) or companies are allowed to make blanketstatements that they own essential patents without precisely identifying them (“blanketdisclosures”). Some SSOs like ETSI ask for specific patent disclosures while IEEE and otherspermit blanket disclosures. Importantly, ETSI IPR Policy Section 4.1 calls for patentees todeclare even if in doubt:

3More information on 3GPP and LTE can be found on 3GPP webpages www.3gpp.org. See also Bekkersand Martinelli (2012) for a historical narrative of the development of cellular standards.

8

"Subject to Clause 4.2 below, each MEMBER shall use its reasonable endeav-ours, in particular during the development of a STANDARD or TECHNICALSPECIFICATION where it participates, to inform ETSI of ESSENTIAL IPRs ina timely fashion. In particular, a MEMBER submitting a technical proposal fora STANDARD or TECHNICAL SPECIFICATION shall, on a bona fide basis,draw the attention of ETSI to any of that MEMBER’s IPR which might beESSENTIAL if that proposal is adopted." [emphasis added]

Thus, the IPR Policy is itself a source of declaration imprecision. The Policy aims atcapturing all possible SEPs rather than capturing only those patents that are truly essentialto the standard. If even in doubt, the patent holder is obliged to declare4. Failing todeclare has more severe consequences that include loss of enforceability of the patent andpossible accusations of patent ambush. Declaring a non-essential patent has significantlyfewer consequences, if at all. Consequently, patentees interpreting the policy requirement havevarying declaration policies resulting in varying degrees of preciseness in their declarations.Because there is no confirmation nor rejection of the declaration, the accuracy of declarationsrests solely on the declaring company.

SEP disclosures are typically tied to a requirement that the members license their SEPsto standard implementers on “fair, reasonable, and non-discriminatory” (FRAND) terms.However, SSO policies generally do not specify exact licensing terms and remain vague aboutthe meaning of FRAND, which has led to divergent interpretations of FRAND licensingcommitments and reliance on courts in licensing negotiations including that the essentialityof declared patents is determined by the courts.

3 Overview of Literature on Over-Declaration, PatentValue, and Standard Essential Patents

3.1 SEP Over-Declaration

A patent that is declared as an SEP may not be truly essential to the standard for severalreasons (European Commission 2014). First, because patents may have been disclosed beforethe finalization of the standard, the disclosed SEP may no longer be essential to the finalstandard specification. Second, the patent application of the SEP may have been rejected

4See also Bekkers and Updegrove (2012).

9

and therefore the patent is no longer valid or the scope of the application may have beennarrowed following the disclosure and is no longer essential to the standard. Third, the patentmay not be infringed by particular implementations of the standard, and therefore, is nottruly indispensable. In addition to these three cases in which the declaration may have beenmade appropriately based on information available at the time of the disclosure and in linewith the (ETSI) IPR Policy obliging to disclose all patents that might be essential, therecan also be “untruthful” disclosures in which parties knowingly declare non-essential patentsto be SEPs (European Commission 2014). These false declarations are primarily driven byrent-seeking patent owners who may be interested in influencing the direction of the standarddevelopment and/or superficially bolstering the size of their SEP portfolio to increase theirbargaining power in licensing negotiations5.

In view of these cases of non-essentiality among disclosed patents, industry professionalsand policy makers have pointed to a number of factors that can contribute to suspectedhigh over-declaration rate. First, the self-declarations of SEPs by patent holders are subjectto minimal or no screening by the SSOs. SSO policies typically only prescribe rules forthe availability of patent licenses and their terms and do not outline systematic policiesor evaluation mechanisms for assessing whether the patents declared to be SEPs are trulyessential to the standard. Second, the process of verifying the technical essentiality of patentsis increasingly difficult in light of the rapid evolution and growing complexity of technology.Third, standards and evaluation of patents may change over time, yet SSO policies do notformulate systematic rules for updating declarations of essentiality. Fourth, because failureto disclose SEPs may prevent the patent owner from commercially exploiting its essentialpatents later on, firms may find it in their best interest to over-disclose. Fifth, patent holdersaim to have their technologies incorporated into standards for the revenue stream that thisgenerates through royalties as well as to remain competitive in the market through theirknowledge of the technology. In summary, over-declaration of SEPs results from the lack oftransparent and systematic screening and updating mechanisms for essentiality, the difficultyand the high cost of verifying essentiality, and large incentives to own larger shares of theSEPs yielding bargaining power in license negotiations.

Recent findings from technical assessments of SEPs and outcomes from patent litigationsindicate over-declaration to be a highly prevalent phenomenon. For example, a series of tech-

5A larger SEP portfolio may be perceived more valuable and the SEP holder can thus extract higherlicense fees for it, but a larger SEP portfolio can also be helpful in diluting other SEP holder’s demands oflicense fees.

10

nical evaluation of SEPs in telecommunications industry by Fairfield Resources International,an intellectual property consultancy, indicate that as high as 80 percent of declared essentialpatents are not truly essential (e.g., Fairfield Resources International 2007, Fairfield ResourcesInternational 2010). Cyber Creative Institute find that the overall essentiality rate for patentsdeclared for LTE standard is 58 percent (Cyber Creative Institute 2012). Outcomes frompatent litigations in the telecommunications industry show similar rates of over-declaration.For example, in a 2005 patent litigation in the UK, Nokia alleged that 31 of InterDigital’sSEPs were not essential to the UMTS standard; prior to the trial, InterDigital droppedessentiality claims for 26 (or 84 percent) of these patents (Bekkers and West 2009). Althoughthe question of true essentiality is ultimately a legal question that can be only be determinedin a court of law, detailed technical assessments of SEPs and patent owners’ responses tochallenges in litigation provide useful guidance on the prevalence of over-declaration.

3.2 Determinants of Patent Value and Essentiality

Our study is related to three strands of research on patent value and standard setting: (1)literature on the general relation between patent characteristics and their value, (2) theoreticalmodels of SEPs and the impact of the SSO and standardization on competition, and (3)empirical assessment of characteristics of SEPs. Although many of these previous studiesrecognize the problem of over-declaration, they largely ignore the issue of uncertainty aboutthe true essentiality in the implementation of theoretical and empirical analyses.

First, there exists a large body of literature on using patent characteristics as a measureof the value of a patent. The attribute that has been the focus of these studies is the numberof citations that a patent receives from subsequent patents, or “forward citations”. Viewingtechnological progress as a cumulative process in which new advances and inventions arebuilt on previous knowledge and facilitate subsequent discoveries, researchers have proposedthat the number of forward citations provide information about broad technological impactof the patent. A large number of studies have established that there is a positive relationbetween forward citations and private value of the patents (see e.g., Harhoff, Narin, Schererand Vopel 1999, Lanjouw and Schankerman 2001, Hall, Jaffe and Trajtenberg 2005, Kogan,Papanikolaou, Seru and Stoffman 2012, Jaffe and de Rassenfosse 2016) as well as social valueof patents (Trajtenberg 1990), and this relation has been confirmed also in studies with fieldtrial data (Moser, Ohmstedt and Rhode 2014) and patent auction data (Odasso, Scellatoand Ughetto 2014). Research, however, also shows that the number of forward citation is a

11

noisy measure of the economic or technological value of the patent. As noted by Lanjouwand Schankerman (2001), Rysman and Simcoe (2008) and Moser (2016), several other factorsother than the quality of patents, including publicity, licensing practices, or technologicalfeatures, can impact citation counts. In addition, these studies show that the predictivepower of patent value models is improved by the addition of other patent characteristics,such as patent scope, the year of patent application and/or publication, patent inventor andowner characteristics, technology category, and patent family characteristics, indicating thatmultiple attributes of patents are important correlates of economic and/or technological value.Lanjouw and Schankerman (2004) confirm this by showing that using multiple variablesimproves over a single composite index in measuring patent quality. This strand of literature,however, has not analyzed SEPs as a distinct patent category, and as such, patent attributesassociated with standardization have not been included in the analyses.

Theoretical or discussion studies related to SEPs have focused on assessing the impact ofSSOs and SEPs on the development, adoption, and/or diffusion of standards and technologiesor the effect of SSO rules on competition and bargaining powers of patent owners andlicensors (e.g. Chiao, Lerner and Tirole 2007, Lerner and Tirole 2006, Simcoe 2012, Lemleyand Shapiro 2013, Kühn, Scott Morton and Howard 2013, Lerner and Tirole 2015). However,these studies take as given the true essentiality of patents that are disclosed as SEPs and donot explicitly explore the issue of over-declaration. One exception is Dewatripont and Legros(2013) who develop a theoretical model to study market distortions that arise from patentowners’ private information about the essentiality of the patent. Their model provides severalpredictions on how FRAND-like setting in SSOs affects incentives to invest in R&D and theessentiality rate of patents that firms declare. In a working paper version of Dewatripontand Legros (2013), they consider an extension of the model in which an upstream anda vertically-integrated firm are both involved in patenting. The model predicts that anupstream firm is more likely to over-declare than a vertically-integrated firm, because firmswith a stronger dependence on licensing income from SEPs may have a greater incentiveto over-declare. Another exception is recent work by Lerner et al. (2016). They look atpatent holders’ strategies when to disclose patents by identifying them individually (“specificdisclosure”) and when to make a generic disclosure (or “blanket declaration”). They validatethe theoretical results with cross-SSO data by showing that lower quality patents and firmsin downstream position are more likely associated with generic disclosures.

Recently, an emerging body of empirical studies on SEPs examines the differencesbetween SEPs and non-essential patents and analyzes the relationship between SSOs and the

12

disclosure of standard essential patents. Rysman and Simcoe (2008) examine the economicand technological significance of SSOs by analyzing the forward citation patterns of patentsthat were disclosed to be standard essential to four major SSOs. They find that SEPs are citedmore frequently and over a longer time period than non-SEPs, which suggests that SEPs mayhave higher than average economic and technological value across all patents. Rysman andSimcoe (2008) also note that SEPs are different from non-SEPs in other potential determinantsof patent value noted in prior research. SEPs contain more claims, are more likely to be partof a larger patent family, and are cited by patents from broader set of technology classes thanpatents that are not disclosed to the SSOs. The authors further find that the age distributionof SEP citations is shifted toward later years relative to non-SEP citations with the sameapplication year and that the citations increase following their disclosure to the SSO. Theseresults suggest that SSOs play a significant role in influencing the development of standardsand the adoption of valuable technologies. An important limitation to Rysman and Simcoe(2008) is that they are unable to identify whether a disclosed patent is actually essential tothe final specification of the standard. Consequently, their sample of patents contains “falsepositives” or disclosures that are not actually essential. Layne-Farrar and Padilla (2011)and Baron and Pohlmann (2013b) confirm the finding by Rysman and Simcoe (2008) on theendogenous relationship between the quality of SEPs and development of standards by SSOs.

Layne-Farrar and Padilla (2011) find that SEPs increase in value (as measured by forwardcitations) when included in a standard. The authors suggest that this effect is heterogeneousacross SSOs, application year, standard, and jurisdiction. Baron and Pohlmann (2013b) findthat the disclosure or the inclusion of a greater number of SEPs to the standard increases thelikelihood that the standard would be upgraded but decreases the probability of standardreplacement and that these effects depend on the level of ownership fragmentation in theSEPs. These findings indicate that SEPs may accelerate the technological improvement ofexisting standards but stall significant changes to the standard.

Pohlmann et al. (2016) test for the economic value of SEPs to patent holding companiesby assessing the relationship between the rate of return on assets and the number of SEPsthat the company holds. They find an inverted U-shaped relation between SEP holdingsand financial returns. Holding and licensing SEPs can increase a company’s return, whichencourages firms to become involved in the standardization process and to declare many SEPs.However, the sign of the relation is reversed after a certain number of SEP disclosures. Theauthors attribute this change to an “over-declaration” problem of SEPs, in which a very largenumber of patent disclosures may lead firms to offer fewer products and services and develop

13

lower quality patents. The “over-declaration” problem noted by Pohlmann et al. (2016) isthus different from the over-declaration problem addressed in this paper which concerns thetrue essentiality of patents that are disclosed to the SSOs. They also find that controllingfor firm-fixed effects and patent attributes improves the predictive power of their model,suggesting that the relation between financial return and SEPs can vary across differentcompany and patent types.

Bekkers et al. (2011) provide an important empirical study on the determinants of essentialpatent disclosures and demonstrate the importance of patent and company attributes in SEPs.They assess the role of two main factors – (1) the significance of the technological solution inthe patent and (2) the involvement of the patent applicant in the standardization process –by comparing SEPs declared to the W-CDMA standard with a control group of non-SEPs.Bekkers et al. (2011) find that both factors have significant impact on the probability that apatent is claimed as essential, but the applicant’s involvement in the standardization processhas a stronger impact. The empirical results also demonstrate that various patent and patentapplicant characteristics including forward citations, region of application, company size andR&D expenditures are significantly correlated with the likelihood of a disclosure.

In sum, these theoretical and empirical studies on SEPs reveal the importance of patent,applicant/owner, and SSO characteristics in determining the likelihood and the value ofSEP disclosures. Furthermore, participants to standardization and patentees may differ withrespect to their declaration policies as some firms have more inclusive policies to declarethan others. This is in-line with the findings by Lerner et al. (2016) mentioned earlier. Kangand Bekkers (2015) raise the possibility of opportunistic patenting during standardizationmeetings, though in our discussions with industry and standardization participants, theseconcerns were downplayed and the possibility that problems arising during meetings aresought to be addressed imminently was emphasised. However, the findings by Kang andBekkers (2015) underline the imperfections in the standardization and declaration processesand the differences in SSO members’ policies. In all though, these studies largely ignorethe issue of true essentiality of SEPs in their analyses, though Bekkers et al. (2011) notethat all declarations are not actually essential and determining essentiality is a hard task.To our knowledge, our current study is the first rigorous and, to-date, the most completeexamination of the determinants of true essentiality among SEPs.

14

4 Description of Data Sources

4.1 Data Sources

To conduct our study, we combine data from four different sources: (1) ETSI database onIPR and standards, (2) PA Consulting data on patents that are declared to the 3GPP-LTEstandard and results from PA Consulting’s independent assessment of the SEP’s technicalessentiality, (3) PATSTAT data including bibliographical and legal status information, and(4) COMPUSTAT data on various patent inventor and owner company metrics.

4.1.1 ETSI Database

ETSI maintains a database which provides information about the IPRs that have beendisclosed to be SEPs to the various standards that ETSI has developed. ETSI provides atleast two types of information. The first is the description of the technological standardsand all the relevant versions, and the second is the list of self-declared SEPs. PA Consulting(described below) uses the second type of information for its technical assessment of essentiality.

4.1.2 PA Consulting Database

Patents disclosed to be essential to the LTE (4G) standard in the ETSI database were matchedto the patents evaluated by PA Consulting (PA) for their technical essentiality claims. PA isa consulting firm focusing on the technology and innovation industries. Since the mid-1990s,PA has examined IPR in wireless technology, beginning in GSM, followed by 3G, CDMA,and LTE. Beginning with the 3G standard, PA has developed databases containing PA’sindependent evaluation of the patents that are disclosed to be standard essential to variouswireless standards. Our study relies on PA’s database on the LTE standard. In February2016, PA published the latest version of the “LTE Essential IPR: PA’s 3GPP-LTE Databaseand Report” accompanied by an online database which contains a record of patents that PAhas evaluated.

To assess the technical essentiality of SEPs to the LTE standard, PA’s neutral technicalexperts compared the patent’s independent claims (i.e., those that do not reference otherclaims) to the relevant sections of ETSI’s 3GPP-LTE standard specifications (Release 11).Patents declared as essential and stored on the ETSI database under projects relating toLTE were selected based on the following criteria: (1) relate to LTE, (2) declarations upto end of July 2013, and (3) contain at least 1 declaration from an ETSI member company.

15

Projects that related to the LTE-Advanced standard or those covering features preceedingRelease 8 were not included in the pool of evaluated patents.

Where PA was able to reach a conclusion about the essentiality of the patent, PA rankedthe patents from 1 through 3 according to how essential they are to the standard from atechnical perspective:

• 1 = All the elements of the claim match the technical specifications of the LTE standardand hence the patent is probably essential to the technical standard

• 2 = The patent is relevant to the technical standard (a useful implementation method,but not directly described in the standard)

• 3 = The patent is not relevant to the technical standard in any way or at least oneelement of the claim does not match the standard

For patents where assessment results were unclear, PA errors on the side of caution byclassifying these patents as technically essential (i.e., rank of 1). The PA database is based onETSI IPR database disclosures to the end of July 2013 and contains 6,505 disclosed patentsfor LTE. Of these disclosed patents, PA successfully analyzed 4,628 patents.

The patent evaluation and the creation of the database were conducted in four steps:

1. Identification of LTE SEPs: The set of patents declared to ETSI for LTE were identifiedin the ETSI IPR Database.

2. Location of US family member: The set of disclosed patents were filtered by identifyinga US granted patent or application; duplicates were identified and flagged.

3. Comparison of claims with LTE standard’s technical specifications: The claims of thepatents were evaluated by PA’s engineers and ranked.

4. Creation of Database: Patent numbers, assignees, PA’s rankings, and other data wereincorporated into a database.

In addition to the ranking of essentiality, the PA database also provides information onthe current ownership of the patent from Thomson Reuters and patent family characteristics.It is worth pointing out that PA assesses applications too, and their report states that anessential application may have its claims narrowed down which renders the granted patentnot-essential despite the assessed application being judged essential.

16

4.1.3 PATSTAT

PATSTAT contains bibliographical and legal status patent data from patent authorities fromleading industrialized and developing countries. PATSTAT is extracted from the EPO’sdatabases and is provided as raw data for statistical tools. It contains information relating tomore than 90 million patent documents from 40 patent authorities worldwide.

There are three main databases in PATSTAT:

• EPO worldwide patent statistical database: contains bibliographical data relating tomore than 90 million patent documents from leading industrialized and developingcountries.

• EPO worldwide legal status database: contains legal status data on patent applicationsand granted patents from more than 40 patent authorities worldwide.

• European patent register: contains bibliographic and legal status data on publishedEuropean and Euro-PCT patent applications

4.1.4 COMPUSTAT

Standard & Poor’s Compustat database provides various financial and company informationon the inventor and the current owners of the SEPs in the study: R&D expenditure, numberof employees, returns to assets, and value of intangible assets.

4.1.5 Sample Selection

Our data include only SEPs evaluated by PA Consulting that can be linked to the ETSIdatabase. Of the 79,257 patents in the ETSI database (which contains information ondisclosed SEPs to various standards that ETSI developed including the LTE standard), 3,953patents were matched by their patent application number to the set of SEPs that wereevaluated by PA. From the match sample of SEPs, 35 patents were dropped from the basesample, because the ETSI data shows that they were not declared to the 4G standard in thedata. In addition, only SEPs that were declared to the US patent office were kept in the basesample, which resulted in the removal of four SEPs. In sum, 3,914 SEPs are included in ourbase sample.

17

5 Empirical Methods

We estimate a logit regression model to assess the impact of key attributes of patents andcharacteristics of inventor and owner companies on the likelihood that the disclosed patent isessential:

1(Essentiality)p,f,t = F (Xp, Xf , Tt, Ip, Op) (1)

Essentiality of patent p in family f with application year of t is estimated as a function ofpatent characteristics Xp, patent family characteristics Xf , age characteristics Tt, inventingowner attributes or fixed effects Ip, and declaring owner attributes or fixed effects Op.

5.1 Dependent Variable: Essentiality

Although PA provides a ranking of 1 through 3, we combine rank of 2 and 3 to create adichotomous variable of essentiality in our analysis:

• 1 : PA rank = 1

• 0 : PA rank = 2 or 3

5.2 Summary of Potential Determinants of Essentiality

Existing literature on patents and studies specifically related to SEPs both demonstrate theimportance of several categories of patent and inventor or owner characteristics on patentvalue and the likelihood of disclosure. We consider four groups of potential explanatoryfactors.

5.2.1 Patent Citations

Existing studies indicate that number of citations that a patent receives is positively correlatedwith the economic and technological value of the patent. The previous literature has largelyfocused on the impact of the number of overall forward citations but the relation betweencitations and the value of the patent may differ depending on the type of citation as wellas the timing of these citations. Furthermore, our focus is on true essentiality of a patentdeclared as standard-essential, and correlation between overall forward citations and trueessentiality might be weaker than the correlation between forward citations and economic ortechnological value. In our regression model, we consider the impact of forward citations, self-citations, and citations from non-patent literature. Importantly and novel, we also consider

18

the disparate impact of forward citations received from other patents that were declared tothe standard, and also from those judged essential to the standard (PA Rank = 1). The ideain separating “within” standard citations and “external” citations is to test the hypothesisof cumulative innovation in standardisation. The LTE standard is created by a series ofinventions over many years and it is built on previous generations of telecommunicationsstandards. The best prior art for a SEP is the previous standard because of backwardcompatibility requirement. Therefore if a patent receives citations from subsequent patentswithin the line of standardization, it can be expected to be more relevant for the standardand thus more likely to be judged as essential. Forward citations from the same standardthus indicate a higher likelihood that the SEP is infringed by the implementations of thestandard. Conversely, a citation received from a patent that does not cover the standardcan be expected not to reflect on the actual essentiality of the cited patent to the standardin question. Belenzon (2011) too distinguishes between citations that are coming fromsubsequent patents within the same cumulative line of innovation and citations outside of theinnovation line. He finds that the former “internal” citations are positively related to firm’sstock market value, whereas the latter “external” citations are not. The use of citations,however, is complicated when comparing patents with different ages. Older patents are morelikely to have received a greater number of citations simply by the virtue of the fact thatthey have had longer period of time to receive them. The problem may be aggravated bypotential changes in the patent and citations practices over time. Thus, to mitigate this issue,we age-normalize the citation counts and control for patent application year fixed effects inall of our specifications. Patent citations counts are also standardized to have mean 0 andstandard deviation of 1, such that the coefficient on citation counts can be interpreted as thechange in the likelihood of essentiality associated with a one standard deviation change incitation counts.

5.2.2 Basic Patent Characteristics

Previous studies indicate patent age is associated with patent value. Patents that weredeclared earlier to the standard may have claims for technologies that are more fundamentalto the implementation of the standard. They could have also influenced the development ofthe standard. Thus, patents that were declared early could be more likely to be essential.We include application year fixed effects in all our model specifications.

A patent application that is granted has undergone and passed examination and, thus,may be more likely to be essential. On the other hand, the claims of a granted patent may be

19

more narrow and less likely to cover the standard’s elements, and therefore a granted patentmay be less likely judged essential compared to the patent application. Existing studies alsodemonstrate that patent technology class impacts citation counts and valuation. We considervariables that indicate whether the patent application was granted and indicators for broadpatent technology classes. We also study the impact these categories of attributes that areassociated with the family of patents to which a particular SEP belongs.

A natural place to look at for information about the value of the patent, or about itsessentiality to the standard, is the scope of the patent. As Lerner (1994) notes “The best wayto measure patent scope might be through subjective assessments.” Consequently, economicsliterature has turned to proxies for patent scope, including IPC codes (Lerner 1994) andthe number of claims (Lanjouw and Schankerman 2001). Independent claims form the basisfor the patent’s claim structure, and therefore the analysis on the claims’ quality can benarrowed down to these. We develop two approaches to measure claim quality, and examinehow these measures of patent quality are related to standard essentiality of SEPs. First,we look at the number of independent claims and the length of the first independent claimin the subsample of granted patents.6 A shorter first independent claim corresponds to abroader scope of the patented invention, and thus to higher patent quality. This measureof claim value relates primarily to the likelihood of the standard specification infringing thedeclared SEP. Likewise, a patent with more granted independent claims is more broad in itscoverage of the patented invention. Second, we further compare the first independent claimof the granted US patents in our sample to the first independent claim of the correspondingEP patent within the same patent family, if the EP grant exists. If these two claim lengthsmatch closely, it indicates that the claim has passed examination relatively intact, thus thereshould be less concerns about the validity of the patent in particular. de Rassenfosse, Jaffeand Webster (2016) have recently shown that European patent examiners on average employa stricter examination standard than their US colleagues. Several patent practitioners havetold us to use such similarity measures when starting to assess the general quality of a patent.Our implementation of claim similarity is to compute cosine similarity, a commonly usedmeasure of similarity in text analysis, of the first independent claim of corresponding EPand US grants. Secondly, we identify certain key words that limit the breadth of the claimand therefore reduce the quality of the patent. Words such as “maximum”, “minimum”, and

6Harhoff (2016) reports that both the USPTO and the EPO introduced fee increases in years 2004, 2008,and 2010, which had the effect that patent applications started to have fewer claims and longer independentclaims. We control for this by including application year fixed effects.

20

“only” in a claim limit its breadth and therefore are expected to reduce the quality of thepatent. Patent practitioners typically look for such limiting words in claims when assessingthe quality of the patent. Therefore, we include a dummy variable equal to one if any of theabove stated limiting keywords exist in the first claim. However, we do not find any effect ofthis dummy on judged essentiality, and we thus do not report the results of this specification.We leave more sophisticated claim similarity analysis with e.g. natural language processingfor future work.

5.2.3 Disclosure Attributes

The nature of the patent’s disclosure would likely be associated with the underlying trueessentiality of the patent. SEPs that are declared against a specific technical specificationdocument are more likely to be essential than a patent without a reference to such document.The hypothesis is that a firm who declares against a specific technical specification knowsboth the standard and the patent contents better and is thus able to make a more informeddecision on the appropriateness of the declaration compared to a firm who declares patentsagainst the general standard on high level only. We also examine the impact of declarationagainst earlier generation of the cellular standard on which LTE builds. There are two reasonswhy there might be an effect. Firstly, firms that have been involved in the development ofcellular standards for a long time may have better understanding of the standard, and cantherefore make more informed declaration decisions, compared to a firm that just recentlyjoined standardization at 4G LTE phase and who does not have similar knowledge of thetechnologies. Secondly, LTE standard technologies build on earlier generations, and thereforedeclarations to 2G or 3G standards may involve more fundamental technologies that arecarried over to 4G, and therefore patents covering these areas may be more likely essential.

5.2.4 Patent Inventor and Ownership

Existing studies show that the likelihood of true essentiality might be associated with theidentity of the company holding the patent. We assess how the identity of the patent ownerdeclaring the patent as essential or the patentee is associated with the likelihood of trueessentiality through two approaches. First, we include fixed effects for the inventor and thecurrent owner to control for the variation in true essentiality across the set of companiesassociated with SEPs disclosed to the LTE standard. While useful, the inclusion of companyfixed effects does not reveal what company attributes explain variability in the likelihood

21

of true essentiality. Second, we thus include various attributes for inventing and declaringowner companies related to their size, financial performance, R&D expenditure, and a dummyvariable indicating a non-practicing entity (NPE). We further control for changes in ownershipand whether the essentiality declaration was made by somebody other than the patent’sinventor. The inclusion of these characteristics allows us to assess whether traded SEPs differfrom non-traded SEPs in terms of true essentiality.

6 Descriptive Statistics

Table (I) presents descriptive statistics on the patent and company characteristics. The firsttwo columns stratify the statistics by essentiality ranking, where non-essential SEPs aredefined as those patents with a PA ranking of 2 and 3 and essential SEPs are defined aspatents with PA ranking of 1. The last column presents statistics on the full dataset.

The patents in our base sample received, on average, approximately 39.1 forward citations,although the relatively high standard deviation of around 62.1 indicates a wide variation.Patents that are judged essential received close to three more forward citations on averagethan those that are not judged to be essential patents by PA. The difference in receivedcitations between essential and non-essential patents is more pronounced when consideringthe number of citations from other patents that were judged to be essential by PA. Comparedto non-essential patents, essential patents receive around 70 percent more citations fromother judged essential patents. However, we observe no differences in the average numberof citations from the non-patent literature. As can be expected, the great majority of thepatents belong to the electricity technology category, although approximately 12.2 percent ofthe patents also belong to the physics technology category. There are no noticeable differencesin the technology category groupings between essential and non-essential patents.

About 79.1 percent of the patent applications in our sample have been granted by thepatent authorities, and the proportion of granted patents is slightly higher among non-essentialpatents than essential patents. Only about 2 percent of the patents in our sample are lapsed,which is not surprising given that these are patents that have been declared to be standardessential. On average, the patents in our sample are about five and a half years old (fromdate of declaration); essential patents were declared slightly earlier than non-essential patents,on average.

Just over 20.7 percent of the patents were also declared to either 2G or 3G standardprior to being declared to the 4G standard. There is a substantial difference in the share

22

declared to a specific technical specification document of the LTE standard between essentialand non-essential patents. Over 96.1 percent of essential patents were declared to specifictechnical specification rather than to the general LTE standard on high level while only 81.9percent of non-essential patents are. Non-essential patents are slightly more likely to be thefirst patent declared to the LTE standard in their patent family than essential patents.

On average, there are no noticeable differences in declaring owner attributes betweennon-essential and essential patents: whether the declaring owner is a non-practicing entity,whether the declaring owner is the same as the inventor firm, or how many changes inownership there has been for a given patent. About 7.5 percent of the patents were declaredby an NPE, and over 64.6 percent of the patents were declared by the patent inventor.

Essential patents tend to be owned by firms that are slightly smaller as measured by thenumber of employees and, somewhat surprisingly, lower levels of R&D expenses. However,given the wide variance in these measures, these differences are not statistically significant atconventional levels. Finally, essential patents belong to families with a greater number ofpatents (i.e., family size) and they are part of families that tend to receive higher number ofcitations.

Figure (I) illustrates the declaration rates and the mean essentiality rates of the SEPs inour dataset over time, grouping the patents by the year in which the patent was first declaredto ETSI as essential for the 3GPP standard. Figure (I) reflects that some of these SEPs werefirst declared as essential to 2G or 3G before the start of the formal LTE development in 2005and the freezing of the first LTE standard in 3GPP Standard Release 8 at the end of 2008.Figure (I) highlights two general trends. First, declarations of patents to LTE continue to riseafter the publication of the first LTE standard release with a large number of declarationsonly after 2010. Second, patents declared as essential around the first LTE release are mostlikely judged essential by PA whereas especially the large number of patents declared afterthe first LTE release have lower rates of technical essentiality in PA Consulting’s assessment.

The combination of increasing declaration rates and a decreasing mean essentialityassessments also relates to the need of increased transparency recently voiced by policymakersand practioners. Figure (I) is also a reminder that the development of the cellular 3GPPstandards is an ongoing process with a sequence of generations and new standard releasesadding additional features and specifications over time.

23

7 Empirical Results

We estimate a logit model of SEP essentiality, where essentiality is defined as 1, if the PA rankis 1, and 0 otherwise. All model specifications include application year, application owner,declaration owner, and IPC technology field fixed effects. We employ three specifications thatdiffer in their inclusion of variables related to forward citations, but otherwise include thesame various characteristics associated with the patent, its declaration and ownership, as wellas patent family composition. All results are reported in the form of average marginal effects.

In section 7.1, we first analyze the determinants of essentiality with the broadest datasetthat includes both applications and granted patents. In the section 7.2 following, we focus onthe quality of the patent claim and its impact on essentiality. We introduce new variables forpatent literature based on the scope and quality of the independent claims of the SEP. Wedraw these variables from in-depth interviews with technical patent experts who are involvedin assessments of patents’ quality, technical essentiality, and suitability for litigation. Forthe analysis in this section we limit the data first to granted US patents only, and then topatents that have both granted US and EP family members.

7.1 Determinants of essentiality

The estimation results for the Logit Model (1) are reported in Table (II). The estimatedmagnitude and statistical significance of the patent, declaration, ownership, and patent familyrelated characteristics are robust to the different forms of the forward citation variable thatwe vary in the three different specifications. Our baseline specification (1) includes thestandardized count of normalized forward citations (mean=0, sd=1) and an indicator forforward citation by non-patent literature. We find no relationship between normalized forwardcitation count and the likelihood that the SEP is judged essential. Citations by non-patentliterature are also not related to judged essentiality. These results are somewhat unexpectedgiven the robust results from previous literature that has established a positive link betweenreceived forward citations and the quality of the patent. On the other hand, PA assessed thetechnical merits of SEPs against the technical standard specifications, and in that light, itis less surprising that general forward citations from any subsequent patent or citations byany non-patent literature are not predicting technical essentiality of SEPs declared to theparticular LTE standard. Our finding suggests that true (technical) essentiality, a key factorof SEP value, is not explained by general forward citations that are used as a measure ofpatent value (though they are known to be imperfect predictor of patent value). In order to

24

understant our finding better, we further investigate the association between citation countsand essentiality in the two other model specifications.

It is nonetheless likely that forward citations from patents closely related to the standard inquestion might be informative of essentiality (i.e. infringement of the patent by the standardimplmentations) even if general forward citations are not. We thus include the normalizedand standardized forward citation count from other patents declared to 4G LTE standard inthe specification (2). We normalize this citation count by age across all of these 4G SEPs.Table (II) shows that age-normalized citations from SEPs declared to the same standardare a statistically significant predictor of judged essentiality. A one standard deviation inage-normalized forward citations from other patents declared to the same standard (equivalentto about 20 citations) increases the likelihood of a patent being judged essential by 3.3 percentwhile the coefficient on overall forward citations and citations by the non-patent literatureremain statistically and economically insignificant. In specification (3), we further investigatethe statistical relationship between forward citations and essentiality judgement by alsoincluding the count of forward citations from other declared LTE patents that PA has judgedas essential. We also normalize this citation count by age across all LTE SEPs judgedessential. Inclusion of this additional citation count as an independent variable shows thatthe correlation between the age-normalized forward citations from other declared SEPs andessentiality is entirely driven by the citations from SEPs that have been judged as essential.A one standard deviation in age-normalized forward citations from other patents declared andjudged essential to the same standard (equivalent to about 3 citations) increases the likelihoodof a patent being judged essential by 3.0 percent. Specification (2) nevertheless highlightsthat forward citations from other declared SEPs of the same standard is a statistically andeconomically significant predictor of technical essentiality, even if their technical essentialityis unknown. Forward citations within the same line of technical standardisation are thusa good predictor of essentiality, which helps in ex post assessment of essentiality. While astatistical assessment of essentiality may not be appropriate for individual patents, it givesus a cost-efficient method to establish a meaningful indication of ownership of truly essentialSEPs on (larger) portfolio level. This may be helpful in litigation or regulatory review contextwhen it is unfeasible or financially impossible to acquire patent specific technical assessmentsof all SEPs individually. Forward citations are particularly helpful in situations when weare interested in historical declarations, but they naturally are not available when assessingcontemporaneous SEP declarations. For such cases, we need predictors of essentiality basedon existing data.

25

We use specification (2) in Table (II) to discuss the other patent characteristics notrelated to forward citations. Granted patents are 8.0 percent less likely to be judged essentialthan patent applications. This finding may be attributed to the narrowing of the patentclaims as well as to changes in the standard specifications during and after the patent reviewprocess. Allowing declarations of patent applications increases the ambiguity around trueessentiality, but a policy banning these declarations does not necessarily lead to improvement.Firstly, it may result in more situations where firms fail to declare actually essential patentsby mistake. Secondly, it may make SEP licensing more cumbersome. A potential licenseewill be interested in getting a license to all possible SEPs, including applications, in orderto eliminate the risk of future litigation by the patent holder. In order to assess the truestrength of the SEP portfolio, the licensee needs information about the SEP holdings of thepatent owner, and if applications were not allowed to be declared, the licensee would nothave any information about potentially valuable essential patents which are not yet beengranted. This problem is aggravated by the time lag between filing a patent application(and often combined with making a contribution to the standard around the same time) andthe application becoming first public and then later having the application being granted.More work on grant lags, changes in essentiality status during application prosecution, andpotential effects for license negotiations is needed to establish a robust policy recommendation.Lapsed patents on the other hands are not less likely to be judged essential, and neither arepatents declared first in the family. We account for the relationship between the time ofdeclaration and essentiality assessments using dummy variables for the year in which thepatents were first declared to ETSI as essential. Using the year 2009 with the first freezeof the LTE standard (Release 8) in 3GPP at the end of 2008 as the base category, thesedummy variables clearly show that patents first declared around the development of thisfirst LTE release are most likely be judged essential, whereas especially declarations in lateryears are ceteris paribus less likely to be judged essential. Note that we estimate these effectswhile simultaneously controlling for application-year fixed effect. This pattern shows thatcontributions in early phase of standardization may be more fundamental and thus patentscovering areas specified in the early releases have therefore higher likelihood of essentiality. Itmay also be the case that fundamental innovations are rarer, and therefore the pool of patentsfrom which declarations are made has less patents but they are more relevant to the standardon average compared to later years. Declarations to later releases on the other hand can coverareas that, while describing useful solutions of specific technical problems, may not be solesolutions to the problem and thus not implemented by the standard, or the solutions described

26

in the SEPs may be helpful for the implementation of the standard, but still fall out-of-scopeof the exact standard specification (remember that the requirement is to delcare all patentsthat might be essential). Furthermore, there can be more alternatives for the standardspecification associated with the later standard areas, and therefore a declared patent maymore likely cover those specifications that do not end in the final standard. An alternativeexplantion to the found pattern is that declarations made in the early standardization yearsare made by firms who are on average more experienced compared to later years. Theseexperienced firms may have better knowledge of the preceeding technologies which LTE isbuilt on, better knowledge how standardization process works, and they may have higherquality internal review processes for declarations, compared to new comer firms who just jointelecommunications standardization at the 4G phase.

A specific, earlier declaration of essentiality to a previous generation of the studiedstandard (in this case 2G GSM or 3G UMTS) does not reduce the likelihood of a patent alsodeclared to 4G LTE as being judged essential to this current standard generation. However,patents that are declared so that the declaration refers to a specific technical specificationdocument rather than referring only to the general LTE standard are substantially morelikely to be judged essential. Patents declared to a specific technical specification of thestandard are more than 38.6 percent more likely to be judged essential than patents witha general declaration to LTE, and this result is statistically significant at the one-percentlevel. Declaration to a specific technical specification is potentially a sign of thorough internalreview and potentially also demonstrates deeper understanding of the standard. This majorresult strongly encourages a move towards allowing only declarations that refer to specifictechnical specifications of the standard by the SSOs. Declarations, which simply refer to LTErequire limited review of the declarations, and naturally capture all possibly essential patentsmore broadly, appear to be however much too impricise and one of the most significantcontributors to over-declaration of SEPs.

Family size, frequently used as a proxy for patent value, is not associated with thetechnical essentiality of a declared SEP, and reflects our finding on general forward citationswhich indicates that the traditional proxies for patent value fail to adequately explain trueessentiality. Ownership fragmentation of the patent family, measured as the number ofdistinct owners of the family, is also not related to technical essentiality. Declaration of thepatent as essential to LTE by either the original inventor or by a non-practicing entity (NPE)also does not affect the likelihood of the patent as being judged as essential by PA’s technicalexperts. Neither does the number of ownership changes so we do not find any evidence of

27

traded SEPs being less likely to be technically essential.Our regression specifications further include characteristics of the firms applying for the

patent and declaring it as essential. Return on assets is a proxy variable for investment, R&Dexpenses ($ millions) capture the overall level of investment in technological innovation anddevelopment, and the number of employees (hundreds) is a measure of firm size. Note, thatwe include these variables in addition to both application-owner and declaration-owner fixedeffects, and the results are robust to the exclusion of these firm-level fixed effects from thelogit regression. We do not find empirical evidence that these company-specific attributes ofapplicants and declarants are associated with the patent’s judged essentiality after controllingfor these fixed effects. The estimated coefficients on R&D expenses of the declarant and thereturn on assets of the applicant are statistically significant at the 5 percent or 10 percentlevel but neither effect is economically significant. We do not report firm-level fixed effects toprotect PA’s clients.

7.2 Claim Analysis of Granted Patents

We also explore how characteristics related to the quality and scope of claims of grantedpatents relate to technical essentiality. Table (III) reports the results of our regressionsusing the subsample of granted US patents with additional explanatory variables. Ourprimary results discussed in the previous sections hold in direction, magnitude, and statisticalsignificance for these regressions. The existence of a corresponding EP patent grant (ifapplied) is however not correlated with the essentiality assessment of the US grant. Havingone more granted independent claim increases the likelihood of technical essentiality by 1.1percent. Our interpretation for this finding is that the patent’s claim on the applicability ofthe patented invention increases in the number of independent claims, and that a patent withmore independent claims is thus more likely to cover an essential feature of the standard. Wefurther find that the length of the first claim, measured by the number of words, is negativelyrelated to PA’s essentiality assessment. Patent practitioners often start their work withlooking at the length of the independent claims. A shorter claim is regarded as a more usefuland valuable, as it is more general and thus covers a broader set of implementations. We findthat increasing the length of the first claim by one standard deviation, around 14 words inour dataset, lowers the likelihood of essentiality by 3.2 percent. This result agrees with thegeneral view from practice that longer claims are likely less general in nature and thus morespecific and narrow which increases the likelihood that the claim does not exactly match the

28

standard specification text, and therefore the standard specification does not infringe thedeclared SEP. This result is interesting as it agrees with the proxies for patent value capturinginfringement and used in practice, and it agrees with the results on forward citations fromother declared (and judged) essential patents.7

We further explore to what extent differences in the similarity of the first claim betweengranted US and EP patents relates to technical essentiality. Claim similarity is a measureused by practitioners as a starting point for patent assessment, in particular the assessmentof patent validity. If the claims found in a US and an EP patent match closely, it indicatesthat the invention described in the claim is clear and the claim text is well-drafted, and thepatent is more likely valid. The value of such claim can be expected to be above average.Table (IV) reports our regression specifications for the subsample of these successful twinpatent applications. We include the cosine similarity of the term-document matrix of thefirst claim of each EP and US patent. While this similarity is often used by practitioners,this measure does not correlate with the technical essentiality assessment PA Consulting hasperformed for the US family member. This finding is explained by the distinction betweenthe concepts of validity and essentiality that was emphasised by the patent practitioners weintereviewed.

8 Concluding Remarks

The over-declaration of standard-essential patents has become an important topic for industryprofessionals and policymakers alike. However, economists have ignored this importantissue in their research despite having studied standardization for a long time. Transparency,especially with regards to telecommunications standards and the patents essential to them,will likely become even more important in the future. The Internet of Things (IoT) willlead to the deployment of telecommunications standards such as 4G LTE or the future5G in a broad range of industries. To our knowledge, this article is the first study of thedeterminants of true essentiality of standard-essential patents. We study the determinantsof true essentiality of patents declared essential to the 4G LTE cellular standard using anovel and rich dataset combining various patent characteristics with independent technicalassessments of standard essentiality.

7In an unreported specification, we also include a dummy variable equal to one if any of the limitingkeywords “minimum”, “maximum”, and “only” is used in the first claim. However, we do not find any effectof the dummy variable on judged essentiality.

29

We employ a logistic regression to assess the relation between essentiality and a broad setof variables that prior economics literature has found to correlate with patent value. Ourresults suggest that models explaining patent value are not sufficient to explain the technicalessentiality of an SEP. We thus investigated new variables linked to standardization processand variables motivated by the work by patent practitioners. We find that the strongestpredictor for essentiality is when the declaration of the SEP refers to a specific technicalspecification of the standard. This leads to a policy recommendation that SSOs shouldrecommend patent holders to declare against technical specifications. This policy wouldreduce over-declaration, but it should be kept in mind that it puts more burden on patentholders and may lead to a higher probability that a patent holder fails to make a declarationby accident. We also find support for claim quality based variables, but this area requiresfurther work. In a future version of this paper we will expand the analysis to study additionaldeterminants of SEP essentiality to further increase the policy relevance of this study.

Standard development and declarations of SEPs are very complex processes in telecom-munications, and any policy recommendation needs to be carefully balanced. Statisticalanalysis is helpful in the analysis of large data, but the results should not be applied inassessments of individual patents. This should be kept in mind when developing policies fortelecommunication standards. However, our results are encouraging and provide guidancefor the policy debate, and hopefully attract more research in order to further the discussionaround SEPs and SSO policies.

30

References

Baron, J and T Pohlmann (2013a) “Who cooperates in standards consortia-rivals or comple-mentors?,” Journal of Competition Law and Economics, Vol. 9, pp. 905–929.

Baron, J. and T. Pohlmann (2013b) “Who cooperates in standards consortia-rivals orcomplementors?,” Journal of Competition Law and Economics, Vol. 9, pp. 905–929.

Baron, Justus, Yann Ménière, and Tim Pohlmann (2014) “Standards, consortia, and innova-tion,” International Journal of Industrial Organization, Vol. 36, pp. 22–35.

Baron, Justus, Tim Pohlmann, and Knut Blind (2015) “Essential patents and standarddynamics.”

Baron, Justus and Daniel F Spulber (2015a) “Technology Standards and Standard SettingOrganizations: Introduction to the Searle Center Database,” Northwestern UniversityWorking Paper.

Baron, Justus and Daniel F. Spulber (2015b) “Technology Standards and Standards Organi-zations: Introduction to the Searle Center Database.”

Bekkers, Rudi, René Bongard, and Alessandro Nuvolari (2011) “An empirical study on thedeterminants of essential patent claims in compatibility standards,” Research Policy,Vol. 40, pp. 1001–1015.

Bekkers, Rudi and Arianna Martinelli (2012) “Knowledge Positions in High-Tech Markets:Trajectories, Standards, Strategies, and True Innovators,” Technological Forecasting &Social Change, Vol. 79, pp. 1192–1216.

Bekkers, Rudi and Andrew Updegrove (2012) “A Study of IPR Policies and Practices of aRepresentative Group of Standards Setting Organizations Worldwide,”Technical ReportSeptember, National Academies of Science, Boston.

Bekkers, Rudi and Joel West (2009) “The limits to IPR standardization policies as evidencedby strategic patenting in UMTS,” Telecommunications Policy, Vol. 33, pp. 80–97.

Belenzon, Sharon (2011) “Cumulative Innovation and Market Value: Evidence from PatentCitations,” The Economic Journal, Vol. 122, pp. 265–285.

31

Chiao, Benjamin, Josh Lerner, and Jean Tirole (2007) “The rules of standard-setting or-ganizations: an empirical analysis,” The RAND Journal of Economics, Vol. 38, pp.905–930.

Cyber Creative Institute (2012) “Evaluation of LTE Essential Patents Declared to ETSISummary ,”Technical report.

Dewatripont, Mathias and Patrick Legros (2013) “’Essential’ patents, FRAND royalties andtechnological standards,” Journal of Industrial Economics, Vol. 61, pp. 913–937.

European Commission (2014) “Patents and Standards - A Modern Framework for IPR- basedStandardization,”Technical report.

(2016a) Landscaping study of standard essential patents in Europe.

(2016b) Study on Transparency, Predictability and Efficiency of SSO-based Stan-dardization and SEP Licensing.

Fairfield Resources International (2007) “Analysis of Patents Declared as Essential to GSMas of June 6, 2007,”Technical report.

(2010) “Review of Patents Declared as Essential to LTE and SAE (4G WirelessStandards) Through June 30, 2009,”Technical report.

Farrell, Joseph and Carl Shapiro (2008) “How Strong Are Weak Patents?,” American EconomicReview, Vol. 98, pp. 1347–1369.

Galasso, Alberto and Mark Schankerman (2015) “Patents and Cumulative Innovation: CausalEvidence from the Courts,” Quarterly Journal of Economics, Vol. 130, pp. 317–369.

Goodman, David J and Robert A Myers (2005) “3G cellular standards and patents,” in2005 International Conference on Wireless Networks, Communications and MobileComputing, Vol. 1, pp. 415–420, Hawaii: IEEE.

Hall, Bronwyn H., Adam Jaffe, and Manuel Trajtenberg (2005) “Market Value and PatentCitations,” RAND Journal of Economics, Vol. 36, pp. 16–38.

Harhoff, Dietmar (2016) “Patent Quality and Examination in Europe,” American EconomicReview: Papers & Proceedings, Vol. 106, pp. 193–197.

32

Harhoff, Dietmar, Francis Narin, Frederic M. Scherer, and Katrin Vopel (1999) “CitationFrequency and the Value of Patented Inventions,” The Review of Economics andStatistics, Vol. 81, pp. 511–515.

Jaffe, Adam B. and Gaétan de Rassenfosse (2016) “Patent Citation Data in Social ScienceResearch: Overview and Best Practices,” NBER Working Papers.

Kang, Byeongwoo and Rudi Bekkers (2015) “Just-in-time Patents and the Development ofStandards,” Research Policy, Vol. 44, pp. 1948–1961.

Kühn, Kai-Uwe, Fiona Scott Morton, and Shelanski Howard (2013) “Standard SettingOrganizations Can Help Solve the Standard Essential Patents Licensing Problem,” CPIAnti-trust Chronicle.

Kogan, Leonid, Dimitris Papanikolaou, Amit Seru, and Noah Stoffman (2012) “TechnologicalInnovation, Resource Allocation, and Growth,”Technical report, National Bureau ofEconomic Research, Cambridge, MA.

Lanjouw, Jean O. and Mark Schankerman (2001) “Characteristics of Patent Litigation: AWindow on Competition,” The RAND Journal of Economics, Vol. 32, pp. 129–151.

(2004) “Patent Quality and Research Productivity: Measuring Innovation withMultiple Indicators,” The Economic Journal, Vol. 114, pp. 441–465.

Layne-Farrar, A. and A.J. Padilla (2011) “Assessing the link between standards and patents,”International Journal of IT Standards and Standardization Research, Vol. 9, pp. 19–49.

Lemley, Mark a and Carl Shapiro (2013) “A Simple Approach to Setting Reasonable Royaltiesfor Standard-Essential Patents,” Berkeley Technology Law Journal, Vol. 28, pp. 1135–1166.

Lerner, Josh, Haris Tabakovic, and Jean Tirole (2016) “Patent Disclosures and Standard-Setting,” NBER Working Paper 22768.

Lerner, Josh and Jean Tirole (2006) “A model of forum shopping,” American EconomicReview, Vol. 96, pp. 1091–1113.

(2015) “Standard-Essential Patents,” Journal of Political Economy, Vol. 123, pp.547–586.

33

Lerner, Joshua (1994) “The Importance of Patent Scope: An Empirical Analysis,” The RANDJournal of Economics, Vol. 25, pp. 319–333.

Maskus, Keith and Stephen A. Merrill eds. (2013) Patent Challenges for Standard-Setting inthe Global Economy, Washington, D.C.: National Academies Press, pp.176.

Mitchell, Tom M (1997) Machine learning, New York: McGraw-Hill Sci-ence/Engineering/Math, pp.432.

Moser, Petra (2016) “Patents and Innovation in Economic History,”Technical report, NationalBureau of Economic Research, Cambridge, MA.

Moser, Petra, Joerg Ohmstedt, and Paul W. Rhode (2014) “Patent Citations and the Size ofthe Inventive Step. Evidence from Hybrid Corn,” Working Paper.

Odasso, Cristina, Giuseppe Scellato, and Elisa Ughetto (2014) “Selling Patents at Auction:An Empirical Analysis of Patent Value,” Industrial and Corporate Change, Vol. 24, pp.417–438.

OECD (2011) “Standard Setting,” in Secretary General ed. Competition CommitteeDAF/COMP(2010)33, pp. 1–302: OECD.

PA Consulting Ltd (2015) “ESSENTIAL INTELLECTUAL PROPERTY IN 3GPP-FDD PA’ s 3GPP-FDD,”Technical report, PA Consulting Group, Melbourn.

Pohlmann, Tim (2016) “Landscaping standard-essential patents,” IAM Magazine, Vol. 75.

Pohlmann, Tim, Peter Neuhäusler, and Knut Blind (2016) “Standard essential patents toboost financial returns,” R and D Management, Vol. 46, pp. 612–630.

de Rassenfosse, Gaétan, Adam B. Jaffe, and Elizabeth Webster (2016) “Low-quality Patentsin the Eye of the Beholder: Evidence from Multiple Examiners,” Working Paper 22244,National Bureau of Economic Research.

Rysman, M. and T. Simcoe (2008) “Patents and the Performance of Voluntary Standard-Setting Organizations,” Management Science, Vol. 54, pp. 1920–1934.

Sawant, Gaurav S and Parijat Oak (2016) “The Evolving Landscape of Standard EssentialPatents,”Technical report, Thomson Reuters, Philadelphia.

34

Simcoe, Timothy (2012) “Standard setting committees: Consensus governance for sharedtechnology platforms,” American Economic Review, Vol. 102, pp. 305–336.

Trajtenberg, Manuel (1990) “A Penny for Your Quotes: Patent Citations and the Value ofInnovations,” The RAND Journal of Economics, Vol. 21, pp. 172–187.

Figures

Figure I: Declaration rates and mean essentiality assessments by 3GPP under developmentat time of declaration to ETSI.

Begin of LTE standardization, Rel. 6 Rel. 7 Rel. 8 Rel. 9 Rel. 10 Rel. 11 Rel. 12

27.7 29.1

39.0 40.9

52.8

41.537.8

33.8

26.422.8

33.3

UMTS/W-CDMA LTE

0

10

20

30

40

50

60

70

80

90

100

Tech

nica

l Ess

entia

lity

(%)

0

100

200

300

400

500

600

700

800

900

1000

1100

No.

of D

ecla

ratio

ns

20052008

20092010

20112013

2014

before 2005

20052006

20072008

20092010

20112012

20132014

Earliest Declaration Year

Declarations (#) Technical Essentiality (%)

US SEPs declared essential to LTE with technical essentiality assesment by PA Consulting.

Tables

35

Table I: Descriptive Statistics.

Non-Essential Essential TotalMean (SD) Freq. (%) Mean (SD) Freq. (%) Mean (SD) Freq. (%)

Forward Citations 37.48 (58.44) 42.03 (68.36) 39.08 (62.14)Citations from Other Declared Patents 10.48 (18.56) 14.48 (25.15) 11.89 (21.20)Citations from Other Judged Essential Patents 1.01 (2.35) 1.73 (3.81) 1.26 (2.97)Cited by Non-Patent Literature (0/1) 0.28 (0.50) 0.26 (0.48) 0.28 (0.50)IPC Patent ClassificationHuman Necessities (0/1) 6 (0.2%) 0 (0.0%) 6 (0.2%)Performing Operations, Transporting (0/1) 4 (0.2%) 0 (0.0%) 4 (0.1%)Chemistry, Metallurgy (0/1) 5 (0.2%) 1 (0.1%) 6 (0.2%)Mechanical Engineering, Lighting, Heating,Weapons (0/1)

0 (0.0%) 1 (0.1%) 1 (0.0%)

Physics (0/1) 313 (12.3%) 165 (12.0%) 478 (12.2%)Electricity (0/1) 2,474 (97.5%) 1,340 (97.3%) 3,814 (97.4%)Not Available (0/1) 3 (0.1%) 3 (0.2%) 6 (0.2%)Declaring Owner is NPE (0/1) 179 (7.1%) 116 (8.4%) 295 (7.5%)Patent Basic AttributesGranted (0/1) 2,045 (80.6%) 1,052 (76.4%) 3,097 (79.1%)Lapsed (0/1) 51 (2.0%) 24 (1.7%) 75 (1.9%)Patent Declared First in Family (0/1) 873 (34.4%) 430 (31.2%) 1,303 (33.3%)Patent Age (years) [from declaration] 5.55 (2.80) 5.89 (2.44) 5.67 (2.68)Patent Age (years) [from declaration] Squared 38.58 (46.26) 40.61 (39.50) 39.30 (44.01)Patent Declaration AttributesDeclared to 2G (0/1) 203 (8.0%) 97 (7.0%) 300 (7.7%)Declared to 3G (0/1) 519 (20.5%) 247 (17.9%) 766 (19.6%)Declared to 4G (0/1) 2,537 (100.0%) 1,377 (100.0%) 3,914 (100.0%)Family AttributesNumber of distinct owners in the family 1.65 (0.93) 1.55 (0.90) 1.62 (0.92)Continued on next page

36

Non-Essential Essential TotalMean (SD) Freq. (%) Mean (SD) Freq. (%) Mean (SD) Freq. (%)

Family Size 19.71 (74.77) 23.60 (92.65) 21.08 (81.52)Declaring Owner AttributesDeclaring Owner Same as Inventor (0/1) 1,678 (66.1%) 849 (61.7%) 2,527 (64.6%)Declared to Specific Technical Standard (0/1) 2,077 (81.9%) 1,323 (96.1%) 3,400 (86.9%)Number of Changes in Ownership 4.41 (2.13) 4.29 (2.08) 4.37 (2.11)Owner Attributes around ApplicationYearOwner - Number of Employees (000s) 78.19 (75.71) 62.73 (65.53) 72.75 (72.66)Owner - R&D Expenses ($100,000s) 3,164 (2,749) 2,756 (2,674) 3,021 (2,730)Owner - Return to (%) 7.56 (6.58) 7.46 (6.55) 7.53 (6.57)Owner Attributes around DeclarationYearOwner - Number of Employees (000s) 83.53 (93.72) 66.33 (76.56) 77.48 (88.44)Owner - R&D Expenses ($000,000) 3,735 (3,435) 3,028 (2,910) 3,487 (3,277)Owner - Return on Assets (%) 7.36 (6.24) 6.79 (5.97) 7.16 (6.15)A patent is considered essential if it received a rank of 1 in the PA dataset. It is considered non-essential if it received a rank of 2 or 3.Percentages are calculated as a fraction of the number of patents in the same group.

37

Table II: Logit Regression Results.

(1) (2) (3)Forward Citations 0.010 -0.025 -0.021

(0.022) (0.024) (0.024)Forward Citations from Declared Essentials 0.033*** 0.011

(0.008) (0.012)Forward Citations from Judged Essentials 0.030***

(0.011)Cited by Non-Patent Literature (0/1) -0.012 -0.005 -0.005

(0.019) (0.019) (0.019)Granted Patent -0.076*** -0.080*** -0.080***

(0.022) (0.022) (0.022)Lapsed (0/1) 0.001 0.004 0.004

(0.110) (0.111) (0.108)Patent Declared First in Family (0/1) 0.002 -0.001 -0.000

(0.020) (0.020) (0.020)Earliest ETSI Declaration before 2005 0.002 0.002 -0.000

(0.053) (0.053) (0.053)Earliest ETSI Declaration in 2005 0.056 0.054 0.045

(0.072) (0.072) (0.072)Earliest ETSI Declaration in 2006 0.147* 0.141* 0.138*

(0.076) (0.075) (0.075)Earliest ETSI Declaration in 2007 0.149*** 0.133*** 0.127***

(0.044) (0.044) (0.045)Earliest ETSI Declaration in 2008 0.171*** 0.169*** 0.170***

(0.034) (0.034) (0.034)Earliest ETSI Declaration in 2010 0.047* 0.053** 0.055**

(0.026) (0.026) (0.026)Earliest ETSI Declaration in 2011 -0.039 -0.030 -0.027Continued on next page

38

(1) (2) (3)(0.029) (0.029) (0.029)

Earliest ETSI Declaration in 2012 -0.096*** -0.086*** -0.082**(0.033) (0.033) (0.033)

Earliest ETSI Declaration in 2013 -0.084** -0.072** -0.069*(0.036) (0.036) (0.036)

Earliest ETSI Declaration in 2014 0.103 0.112 0.111(0.072) (0.072) (0.072)

Family Size -0.000 0.000 0.000(0.000) (0.000) (0.000)

Number of distinct owners in the family -0.016 -0.013 -0.013(0.013) (0.013) (0.013)

Declaration to 2G -0.025 -0.014 -0.012(0.037) (0.037) (0.037)

Declaration to 3G -0.036 -0.033 -0.030(0.025) (0.025) (0.025)

Declared to Specific Technical Specification (0/1) 0.386*** 0.386*** 0.382***(0.029) (0.029) (0.029)

Declared Owner is NPE (0/1) -0.069 -0.090 -0.084(0.157) (0.162) (0.161)

Declared Owner Same as Inventor (0/1) 0.011 0.017 0.017(0.071) (0.071) (0.071)

Number of owner changes for patent 0.010* 0.009 0.009(0.006) (0.006) (0.006)

Declaration Year: Owner - Return on Assets (%) -0.001 -0.001 -0.000(0.002) (0.002) (0.002)

Declaration Year: Owner - R&D Expenses (millions) -0.000** -0.000** -0.000**(0.000) (0.000) (0.000)

Declaration Year: Owner - Number of Employees (hundreds) 0.000* 0.000* 0.000*(0.000) (0.000) (0.000)

Application Year: Owner - Return on Assets (%) -0.003* -0.003** -0.004**Continued on next page

39

(1) (2) (3)(0.002) (0.002) (0.002)

Application Year: Owner - R&D Expenses (millions) 0.000 0.000 0.000(0.000) (0.000) (0.000)

Application Year: Owner - Number of Employees (hundreds) -0.000 -0.000 -0.000(0.000) (0.000) (0.000)

Application Year Fixed Effects Yes Yes YesIPC Fixed Effects Yes Yes YesDeclaration Owner Fixed Effects Yes Yes YesApplication Owner Fixed Effects Yes Yes YesObservations 3850 3850 3850Standard errors in parentheses* p<.1, ** p<.05, *** p<.01Patent citation counts normalized and standardized.

40

Table III: Logit Regression Results, Granted US Patents Only.

(1) (2) (3)Forward Citations -0.004 -0.037 -0.035

(0.024) (0.026) (0.026)Forward Citations from Declared Essentials 0.032*** 0.008

(0.009) (0.013)Forward Citations from Judged Essentials 0.032***

(0.012)Cited by Non-Patent Literature (0/1) -0.010 -0.002 -0.002

(0.020) (0.019) (0.019)Independent claims 0.008** 0.006* 0.007*

(0.003) (0.003) (0.003)Length claim one -0.025*** -0.026*** -0.026***

(0.009) (0.009) (0.009)Granted also by EPO -0.024 -0.027 -0.027

(0.018) (0.018) (0.018)Lapsed (0/1) 0.017 0.019 0.018

(0.102) (0.103) (0.101)Patent Declared First in Family (0/1) -0.009 -0.012 -0.012

(0.022) (0.022) (0.022)Earliest ETSI Declaration before 2005 -0.013 -0.016 -0.018

(0.054) (0.054) (0.055)Earliest ETSI Declaration in 2005 0.028 0.025 0.016

(0.077) (0.076) (0.076)Earliest ETSI Declaration in 2006 0.172** 0.163** 0.161**

(0.080) (0.079) (0.079)Earliest ETSI Declaration in 2007 0.137*** 0.117** 0.115**

(0.049) (0.049) (0.049)Earliest ETSI Declaration in 2008 0.162*** 0.159*** 0.160***Continued on next page

41

(1) (2) (3)(0.038) (0.038) (0.038)

Earliest ETSI Declaration in 2010 0.028 0.033 0.035(0.028) (0.028) (0.028)

Earliest ETSI Declaration in 2011 -0.038 -0.031 -0.028(0.033) (0.033) (0.033)

Earliest ETSI Declaration in 2012 -0.114*** -0.104*** -0.098**(0.038) (0.038) (0.038)

Earliest ETSI Declaration in 2013 -0.115*** -0.102** -0.099**(0.041) (0.041) (0.041)

Earliest ETSI Declaration in 2014 0.038 0.046 0.044(0.088) (0.087) (0.088)

Family Size 0.000 0.000 0.000(0.000) (0.000) (0.000)

Number of distinct owners in the family -0.005 -0.002 -0.002(0.014) (0.014) (0.014)

Declaration to 2G -0.041 -0.030 -0.028(0.040) (0.040) (0.040)

Declaration to 3G -0.051* -0.046 -0.042(0.028) (0.028) (0.028)

Declared to Specific Technical Specification (0/1) 0.387*** 0.383*** 0.381***(0.033) (0.033) (0.033)

Declared Owner is NPE (0/1) -0.083 -0.099 -0.094(0.158) (0.163) (0.163)

Declared Owner Same as Inventor (0/1) 0.014 0.021 0.019(0.076) (0.076) (0.075)

Number of owner changes for patent 0.011* 0.010* 0.010(0.006) (0.006) (0.006)

Declaration Year: Owner - Return on Assets (%) 0.001 0.001 0.001(0.002) (0.002) (0.002)

Declaration Year: Owner - R&D Expenses (millions) -0.000*** -0.000*** -0.000***Continued on next page

42

(1) (2) (3)(0.000) (0.000) (0.000)

Declaration Year: Owner - Number of Employees (hundreds) 0.000* 0.000* 0.000*(0.000) (0.000) (0.000)

Application Year: Owner - Return on Assets (%) -0.002 -0.002 -0.002(0.002) (0.002) (0.002)

Application Year: Owner - R&D Expenses (millions) -0.000 -0.000 -0.000(0.000) (0.000) (0.000)

Application Year: Owner - Number of Employees (hundreds) 0.000 0.000 0.000(0.000) (0.000) (0.000)

Application Year Fixed Effects Yes Yes YesIPC Fixed Effects Yes Yes YesDeclaration Owner Fixed Effects Yes Yes YesApplication Owner Fixed Effects Yes Yes YesObservations 3019 3019 3019Standard errors in parentheses* p<.1, ** p<.05, *** p<.01Patent citation counts normalized and standardized.

43

Table IV: Logit Regression Results, Granted US-EP Patents Only.

(1) (2) (3)Forward Citations 0.057 -0.004 0.001

(0.053) (0.055) (0.055)Forward Citations from Declared Essentials 0.045*** 0.017

(0.014) (0.020)Forward Citations from Judged Essentials 0.041**

(0.020)Cited by Non-Patent Literature (0/1) -0.023 -0.015 -0.014

(0.032) (0.031) (0.031)Independent claims 0.012** 0.011** 0.011**

(0.005) (0.005) (0.005)Length claim one -0.029** -0.032** -0.031**

(0.013) (0.013) (0.013)Similarity of first claim EPO-US -0.086 -0.081 -0.084

(0.082) (0.083) (0.083)Lapsed (0/1) 0.092 0.104 0.096

(0.150) (0.145) (0.141)Patent Declared First in Family (0/1) -0.011 -0.017 -0.017

(0.038) (0.039) (0.039)Earliest ETSI Declaration before 2005 -0.004 -0.009 -0.014

(0.072) (0.072) (0.072)Earliest ETSI Declaration in 2005 0.040 0.037 0.025

(0.100) (0.101) (0.102)Earliest ETSI Declaration in 2006 0.055 0.036 0.029

(0.101) (0.098) (0.098)Earliest ETSI Declaration in 2007 0.035 0.004 -0.003

(0.068) (0.068) (0.068)Earliest ETSI Declaration in 2008 0.130** 0.123** 0.126**Continued on next page

44

(1) (2) (3)(0.057) (0.058) (0.058)

Earliest ETSI Declaration in 2010 0.038 0.045 0.048(0.040) (0.040) (0.040)

Earliest ETSI Declaration in 2011 -0.095* -0.087 -0.083(0.055) (0.054) (0.054)

Earliest ETSI Declaration in 2012 -0.061 -0.049 -0.041(0.060) (0.061) (0.061)

Earliest ETSI Declaration in 2013 -0.118* -0.102 -0.099(0.067) (0.067) (0.066)

Earliest ETSI Declaration in 2014 -0.086 -0.077 -0.088(0.128) (0.128) (0.129)

Family Size -0.001 0.000 0.000(0.001) (0.001) (0.001)

Number of distinct owners in the family 0.001 0.004 0.005(0.021) (0.021) (0.021)

Declaration to 2G 0.013 0.028 0.032(0.053) (0.052) (0.052)

Declaration to 3G -0.041 -0.034 -0.027(0.042) (0.042) (0.042)

Declared to Specific Technical Specification (0/1) 0.434*** 0.416*** 0.414***(0.057) (0.056) (0.055)

Declared Owner is NPE (0/1) 0.055 0.034 0.035(0.090) (0.091) (0.090)

Declared Owner Same as Inventor (0/1) -0.001 0.008 0.011(0.100) (0.097) (0.097)

Number of owner changes for patent 0.002 -0.001 -0.000(0.009) (0.009) (0.009)

Declaration Year: Owner - Return on Assets (%) 0.000 0.000 0.000(0.003) (0.003) (0.003)

Declaration Year: Owner - R&D Expenses (millions) -0.000 -0.000 -0.000Continued on next page

45

(1) (2) (3)(0.000) (0.000) (0.000)

Declaration Year: Owner - Number of Employees (hundreds) 0.000 0.000 0.000(0.000) (0.000) (0.000)

Application Year: Owner - Return on Assets (%) -0.003 -0.003 -0.003(0.003) (0.003) (0.003)

Application Year: Owner - R&D Expenses (millions) -0.000 -0.000 -0.000(0.000) (0.000) (0.000)

Application Year: Owner - Number of Employees (hundreds) 0.000 0.000 0.000(0.000) (0.000) (0.000)

Application Year Fixed Effects Yes Yes YesIPC Fixed Effects Yes Yes YesDeclaration Owner Fixed Effects Yes Yes YesApplication Owner Fixed Effects Yes Yes YesObservations 1313 1313 1313Standard errors in parentheses* p<.1, ** p<.05, *** p<.01Patent citation counts normalized and standardized.

46