Research Policy 40 (2011) 1001– 1015 Contents lists available at ScienceDirect Research Policy jou rn al h om epage: www.elsev ier .com An empirical study on the determinants of essen compatibility standards Rudi Bekkersa,b,∗, René Bongardc, Alessandro Nuvolarid a Eindhoven Centre for Innovation Studies (ECIS), School of Innovation Sciences, Eindhoven University of Tec b Dialogic Innovatie & Interactie, Utrecht, The Netherlands c Department o d LEM—Labora a r t i c l Article history: Received 4 No Received in re Accepted 7 Ma Available onlin Keywords: Essential patents Compatibility standards IPR policies Mobile telecommunications W-CDMA Economics of i an inc ispen that ile fo tial stream of licensing revenues. In this paper we study the determinants of essential patent claims in compatibility standards. In particular, we assess the role of two main factors: the significance of the technological solution contained in the patent and the involvement of the applicant of the patent in the standardisation process. We examine the case of W-CDMA, one of the most successful standards in mobile telecommunications. We compare the patents claimed essential for this standard with a control group of randomly selected, unclaimed patents covering the same time period and technology classes. 1. Introdu Standard patents. W technology sionary righ particularly patents tha offer a serv  The views sarily reflect th ∗ Correspon P.O. Box 513, 5 fax: +31 40 24 E-mail add (R. Bongard), a 1 This is refl dards Institute (but not comm state of the art otherwise disp a standard with retrieved from 0048-7333/$ – doi:10.1016/j.nformation systems We find empirical evidence that both factors have significant impact on the probability that a patent is claimed as essential, but the involvement in the standardisation process is a stronger determinant than the technical value (‘merit’) of the patent. On the basis of our findings, we offer policy recommendations. © 2011 Elsevier B.V. All rights reserved. ction s have a fascinating and complex relationship with hereas standards aim at ensuring equal access to the for all stakeholders, patents establish temporary, exclu- ts on the use of inventions. The ensuing tension may be problematic in the case of so-called essential patents, i.e. t are indispensible in order to manufacture a product or ice based on the standard in question.1 To prevent the expressed in this paper are those of the author(s) and do not neces- e policies of Statistics Netherlands. ding author at: Eindhoven University of Technology, Room IPO 2.36, 600BM Eindhoven, The Netherlands. Tel.: +31 40 2475621; 74646. resses: r.n.a.bekkers@tue.nl (R. Bekkers), r.bongard@hotmail.com lessandro.nuvolari@sssup.it (A. Nuvolari). ected in the definition that the European Telecommunications Stan- (ETSI) adopted for ‘essentiality’: ‘[. . .] that it is not possible on technical ercial) grounds, taking into account normal technical practice and the generally available at the time of standardization, to make, sell, lease, ose of, repair, use or operate equipment or methods which comply with out infringing that IPR.’ (ETSI Rules of Procedure, 26 November 2008, http://www.etsi.org). situation in which standards could not be implemented in practice due to the existence of patents covering inventions related with these essential components of the product or service in question, standards bodies have developed policies that establish rules for their members concerning the availability and licensing conditions for essential patents. Most of these policies require the members of the standards body to disclose (‘declare’) their essential patents. Prima facie, such disclosures may be seen as revealing the technical value of the patent for the standard in question: i.e. the patented technology is chosen to become part of the standard by virtue of its contribution to technical performance, cost-effectiveness, etc. However, the technical content of standards is actually drawn up by members of a standards body, thus offering members that are involved several types of opportunities to influence the very technical content of standards. As a consequence, patents may be brought into the standard (and subsequently declared to be essen- tial) because of strategic reasons. After all, ownership of essential patents brings many advantages, such as licensing revenues, and ‘exchange chips’ that open the door to attractive cross-licensing arrangements. This paper examines to what degree patents are claimed to be essential in a standardisation process on the basis of the merit of their specific technological contribution, or the result of the patent see front matter © 2011 Elsevier B.V. All rights reserved. respol.2011.05.004f Business Statistics, Statistics Netherlands, Heerlen, The Netherlands tory of Economics and Management, Sant’Anna School of Advanced Studies, Pisa, Italy e i n f o vember 2009 vised form 18 April 2011 y 2011 e 11 June 2011 a b s t r a c t In the field of compatibility standards, patents (i.e. those patents that are ind to the standard). It is widely believed tool in cross-license negotiations, wh/ locate / respol tial patent claims in hnology, Eindhoven, The Netherlands reasing number of companies claim to own so-called essential sable for designing and manufacturing products conforming the ownership of such patents is a very valuable bargaining r non-producing firms such patents may result in a substan- 1002 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 owner’s involvement in the standardisation process (and the strate- gic opportunities that come along), or both. To do so, we perform an extensive a tial patents standard fo is organised essential pa claiming be to test thes to the tech data set fol analyses. Se esis carried paper and d 2. Essentia Standard lar firm dev successful t question m or sell serv detailed sp obtains a d try to enfo dard (e.g. th Windows). negotiation holders (Fu is the creat holders and standards b other. Form ing opennes process, op rules (Krec different de will refer to as such by 98/34/EC d regional sta these criter the exact c (also called other; there the many su peting again Althoug patents an issue until had occurre when the Eu developed place again European te (Granstrand relations be clash has be and Lea, 19 may be con to develop The mos tions now d refers to th that parties While the (F)RAND policies in principle cover all types of Intel- lectual Property Rights (IPR), including copyrights, trademarks, se ri atego ence able (and al S tia a bod D’s p ill be ake a ch th ey w e ess ident e lic not the cceed at a gard crea latio ally, rger nsure port tents d by mpli re al ndar tent critic scrim nee tent, e ess ent e tec ality ally ting mu aim t overa km/h ) req whi ) at lo logic equi pate hat m ed. H to b e cas ise th r cas t stil ring ost-enalysis on a publicly available list of self-declared essen- for W-CDMA, the most successful third-generation r mobile telecommunications worldwide. The paper as follows. Section 2 introduces the phenomenon of tents and develops two hypotheses concerning their haviour. Section 3 introduces the empirical data we use e hypotheses. In particular, it contains an introduction nology in question (i.e. W-CDMA), a discussion of our lowed by descriptive statistics and some preliminary ction 4 presents the results of the tests of the hypoth- out with a multivariate model. Section 5 concludes the iscusses its implications. l patents and compatibility standards s can evolve in several ways. Sometimes, a particu- elops a set of technical specifications that becomes so hat it is regarded as an industry standard. The firm in ight or might not allow other firms to make products ices based on its industry standard, by publicising the ecifications and/or granting licenses. In case the firm ominant market position, competition authorities may rce that third parties have better access to the stan- e decisions of the European Commission on Microsoft In most cases, however, standards are the result of a process between two or more (independent) stake- nk and Methe, 2001). The explicit aim of this process ion of a standard that will be widely adopted by stake- /or other parties. We can distinguish between formal odies on the one hand, and consortia and fora on the al bodies must adhere to a number of criteria, includ- s to participate, consensus-based decision-making/due en access to documentation, and intellectual property hmer, 1998, 2005; Andersen, 2008). Although various finitions have been proposed in the literature, this paper formal standards bodies as those that are recognised public authorities (for example, in Europe, Directive esignates CEN, CENELEC, and ETSI to be recognised ndards bodies). Consortia and fora might also meet ia, but may also choose not to do so. Depending on ontext, firms may prefer one type of standards body standards developing organisations SDOs) over the fore formal standards bodies, consortia and fora (and b-types one can distinguish here) may be seen as com- st each other in the same “market”. h there were some early cases of tension between d standards this was not considered to be a major the late 1980s. By that time, the first major clash d in the telecommunications sector, more specifically ropean Telecommunications Standards Institute (ETSI) GSM, a standard for mobile telephony. The clash took st a background of liberalisation and privatisation in the lecommunications sector, the so-called pro-patent era , 2005), and the disappearance of exclusive, long-term tween network operators and their suppliers. This first en extensively documented in the literature (Shurmer 95; Iversen, 1999; Wilkinson, 1991; Bekkers, 2001) and sidered as the starting point for standard setting bodies specific rules concerning patent ownership. t common way in which standard-developing organisa- eal with patents is known as (F)RAND, an acronym that e Fair, Reasonable and Non-discriminatory conditions need to ensure for the licenses of their essential patents. databa only c date; h change 1990s all form consor W3C, a (F)RAN not, w will m of whi SDO th becom IPR is to mak choose change not su Note th (e.g. re cannot legal re Usu of a la This e it is im tial pa define that co there a the sta the pa ularly non-di ing fee the pa that ar implem 2.1. Th essenti Usu conflic telecom might uous c to 300 life, (5 ers, (6) and (7 techno these r sity to likely t patent chosen In som to real In othe tion bu by offe more cghts, industrial design rights, patents are virtually the ry of such rights that have been claimed essential to this paper will use the terms patents and IPRs as inter- . The (F)RAND policy was pioneered by ETSI in the early revised and refined several times). Nowadays, virtually DOs have adopted a (F)RAND policy, as well as many nd fora, though there are notable exceptions such as y that requires all patents to be royalty-free. In short, rinciples are as follows: (1) Holders of IPR, member or rewarded in a suitable and fair manner; (2) Members reasonable effort to inform the SDO of relevant IPRs ey are aware. If they propose a technical design to the ill also, in good faith, draw attention to IPRs that could ential once that proposal is adopted; (3) If an essential ified, the SDO will request its holder, member or not, enses available under (F)RAND terms; (4) Members can to license an IPR; if they persevere, the SDO will try to standard so it no longer draws upon that patent. If it does , it will withdraw the standard or stop working on it. lthough this policy does create rights for non-members ing licensing conditions that members can impose), it te obligations for non-members as it does not have any nship with such third parties. standards only define the interfaces between elements system, not the way these elements need to be made. s a maximum incentive to innovate. In this context, ant to note that the (F)RAND policy refers to ‘essen- ’. These are patents that are so basic to the interfaces the standard that it is impossible to design any device es with these interfaces without infringing the patent. If ternative ways to design something that complies with d (even when they are more expensive to implement), in question is no longer essential. Obviously, a partic- al issue is how the requirement of fair, reasonable and inatory conditions is to be interpreted. Does the licens- d to be reasonable from the perspective of the value of or reasonable given the total number of other patents ential to the standard and the total licensing fee that ers can afford or want to afford? hnological value of the patent as a determinant of a standard must attain a number of different, often functional requirements. For example, in the area of nications, a standard for mobile internet data services o (1) offer a high data rate (speed), (2) in a large, contin- ge area, (3) allowing the user to move with speeds up , (4) consuming low power in order to optimise battery uiring a minimum number of cell sites or antenna tow- le being robust to noise and other types of interference w costs for base stations and terminals. Clearly, certain al solutions may be able to deal with one or more of rements better than others, and given the high propen- nt prevailing in many “high tech” sectors, it is very any of these attractive technological solutions will be ence, it is likely that patented technologies may then be e included in a standard because of their attractiveness. es, a patented technology may be the only feasible way e functional requirements of the standard in question. es, the patented technology may not be the only solu- l be the best way to achieve the standard requirements, a higher performance or making the implementation ffective, etc. In both cases, the patented technology has R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 1003 a high degree of merit—or technological value, as one might call it. This leads to our first hypothesis: Hypothesis increases th to a standar If we ass technologic its inclusion will be disc that essent patents, all have attem characterist opposition, of this litera citations re cator of pa Carpenter e consistently correlated w patents (see 2004). The of the valu receives ma tion outline of subseque ment is that that it has b scope of pr points to th in the origin it seems re patent will is the case, cally have a patents. On as a result o this possibi that, given t set, the risk Several r 1 in other patents dec more frequ goes one st place after t tial patents citations – t 2.2. Firm st Firms m the standar the specific standard. O behaviour. nificant rev the door to with the GS for market now genera others from essential pa To unde to get essen typical standardisation process at formal standards bodies (see Schmidt and Werle, 1998 for a more in-depth discussion of deci- proc tens n the ew st en it ber man that nce”) shed ers c ir re an) a act te in th bly, d accep ding es p desig ecis omm ng o e exa uous tions of r re co iged sensu fined deci ajor s (th ny (s s neg pant ften heir stand ts of patin cont their sam t, of on’ p s int n Tec rateg rdisa hesis stand bein e the stra r par expl ten in n-es reatl ial an irely 1. The intrinsic technological value of a patent e likelihood of that the patent will be claimed essential d. ume that the intrinsic technological value of a patented al solution for the standard is the sole determinant of in the standard (and that strategic considerations, as ussed below, do not play any role) then we can expect ial patents are likely to be of higher value than other other things being equal. Economists of innovation pted to assess the value of patents using a number of ics such as citations received, renewals, family size, etc. (Van Zeebroeck, 2011 provides a detailed survey ture). So far, the number of forward citations (i.e., the ceived by a patent) is surely the most popular indi- tent value. Following the pioneering contributions of t al. (1981) and Trajtenberg (1990), various studies have established that forward citations are systematically ith the economic value or the industrial importance of for example, Albert et al., 1991; Sampat and Ziedonis, idea behind the use of forward citations as indicator e of a patent is relatively straightforward: if a patent ny citations, this means that the technological solu- d in the patent serves as a basis for a large number nt technological developments. Another related argu- if a patent receives many citations, this may also mean een frequently used by patent examiners to reduce the otection claimed by subsequent patents and this again e significance of the technological solution contained al patent (Van Zeebroeck, 2011). For all these reasons, asonable to presume that the technological value of a be captured by the number of forward citations. If this then we can expect that essential patents will typi- higher number of forward citations than non-essential e might argue that some patents may receive citations f being declared essential. We shall consider explicitly lity in more detail in the next section and we conclude he procedures adopted for the construction of our data of endogeneity is limited. ecent papers have indeed already explored Hypothesis contexts. Rysman and Simcoe (2008) conclude that lared as essential to standards bodies are cited much ently than a set of control patents. Layne-Farrar (2008) ep further by looking specifically at patenting taking he standard has been set, and concludes that also essen- in this group are more valuable – i.e. receive more han average patents. rategies and essential patents ay try to have some of their own patents included in ds as essential patents for reasons that may go beyond “technological” contribution these patents make to the n reflection, there may be several incentives for this Essential patents are not only a potential source of sig- enue but are also valuable ‘exchange chips’ that open attractive cross-licensing agreements. As demonstrated M case, ownership of essential patents can be crucial entry (Bekkers et al., 2002). Although (F)RAND policies lly prevent patent owners from systematically blocking the market, those market players who do not own any tents are usually not in a very enviable position. rstand better how firms can employ strategies in order tial patents, it is useful to look in more detail at the sional for a ex dards i on a n but oft of mem lish the firms) Refere establi Memb ing the chairm the ex bodies Assem on the – inclu ally tak major these d nical C are goi fore th contin of solu the set know a are obl of con ally de makes some m cedure but ma tinuou partici quite o from t in the minan partici such a ing in do the at leas inclusi patent firms i with st standa Hypot in the patent Onc further to othe aim to cess of and no often g essent for entesses in standardisation committees, and Funk, 2002, ive treatment of competition between and within stan- field of mobile telephony). The way in which the work andard commences differs between standards bodies, is triggered by a proposal that is backed by a number s. A set of requirements is defined in order to estab- date of the participants (usually employees of member are going to make the standard (often called “Terms of . Subsequently, one or more Technical Committees are in order to develop draft specifications for the standard. an decide to participate in these committees (by send- presentatives to the meetings, or even by providing the nd, in this way, can take an active part in determining chnological content of the standard. Although higher e standards bodies (such as a General or a Technical epending on the standards body) will still need to vote tance of a standard, the real technical inclusion process decisions to incorporate patented technologies – usu- lace in the Technical Committees. In exceptional cases, n decisions might be taken to a higher level, but once ions are made, the detailed work goes back to the Tech- ittee. In these committees, discussions and negotiations n about the exact definition of the standard, and there- ct set of technologies the standard is drawing on. In a process, participants in these meetings propose all sorts and technologies in order to draft a standard that meets equirements. If firms suggest technologies which they vered by patents – in particular their own patents – they to disclose this information. On the basis of the pursuit s (which does not mean all need to agree, but is usu- as the absence of persistent resistance), the committee sions on whether to include suggested technologies. On decisions, standards bodies resort to formal voting pro- at often require majority voting or 70% majority voting), maller) technology decisions are made as a part of con- otiation processes in relatively small groups, where the s usually know and respect each other very well and see each other as friends. The benefits that firms derive inclusion in the social network created by cooperation ard has been found to be one of the strongest deter- the willingness of firms to contribute to standards by g in Technical Committees (Bar and Leiponen, 2008). In ext, there might be a tendency to accept others bring- patented technologies, as long as you are allowed to e thing. In this way, the interests of all parties – or all the participants – are catered for. This ‘technology rocess offers firms a number of opportunities to drive o a standard. Both the specific, strategic behaviour of hnical Committees as well as the use of voting power ic intend require a certain degree of involvement in the tion process. Hence our second hypothesis: 2. The active involvement of the holder of a patent ards’ creation process increases the likelihood of that g claimed essential to a standard. standard is established, firms may adopt a diverse set of tegies for exploiting these patents and obtaining access ties’ patents, where necessary. While our paper does not ore this in great depth, it is worth noting that this pro- volves the creation of larger portfolios of both essential sential patents (where the value of some of the latter is y overlooked), cross-licensing involving bundles of both d non-essential patents, and sometimes even patents different products, agreements on future patents, etc. 1004 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 Firm strategies for patents in standards receive increasing interest in the academic community, including contributions from Lemley (2002), Chiao et al. (2007), and Simcoe et al. (2009). The effects of ‘royalty stacking’ are discussed by Shapiro (2001) and Lemley and Shapiro (2006), and are critically commented on by Geradin et al. (2008 the extreme (see, for ins trolls (Reitz massively ‘ essential (G to do so in o are expecte and in spiri are actually ties when su endanger th 2.3. Researc The two may be sign dard, name contained i in the creati such as influ patents of t ment of the to do so, we holders as b phony/telec affecting th constructed essential pa time. Follow study the d for a contin ing the dete 3. Data: th The W-C mobile tele the success lowering th CDMA was that can ac based audio transmissio flexibility: and the spe time. To me be develope (2G) to thir technology value netw standardisa The first third-gener before the fi research pr ied suitable 2 This sectio of the develop known as ‘CoDIT’ and ‘FRAMES FMA2′) formed the basis for what eventually became W-CDMA. However, the 3G developments were largely ignored by GSM operators, who were focusing on increasing subscribers to their existing 2G systems (Garrard, 1998, pp. 478). In 1997, Japan, whose first and second generation standards had chie actur ted t ndard uild ed a t an F o no nies d no cent s of h s. The echn pt gro al w sed l ted s ba Japa urne roces winn RIB ies to pati ng st d to edia with stan only es fu ne s trie e Uni ted. A n Mu som this t bes vari essin rds. The w mou tes a GDP hon dy ar A is men repre mu e an dica DMA ce: h). Considerable attention is paid to some strategies at end of the spectrum, such as ‘holdup’, patent ambush tance, Farrell et al., 2007 and Sidak, 2009), and patent ig et al., 2007). Others have also claimed that firms over-claim’ patents, i.e. unjustly declare patents to be oodman and Myers, 2005). While it may be attractive rder to raise licensing revenues, at the same time, firms d to respect the IPR policies of standards bodies in letter t and claim their patents to be essential only when they so. Claiming non-essential patents might lead to penal- ch patents are challenged in court and might eventually e licensing income by all licensees. h question and our approach preceding sections outline two basic determinants that ificant when claiming a patent “essential” for a stan- ly: (1) the intrinsic value of the technological solution n the patent, (2) the involvement of the patents’ owner ng of the standard, providing room for strategic conduct encing the standards’ content into the ‘direction’ of the he firm. This paper aims to provide a detailed assess- role played by these two types of determinants. In order analyse a set of patents that has been declared by their eing essential to the W-CDMA standard for mobile tele- ommunications. In order to identify the determinants e probability of patents being claimed essential, we have a set of (non-essential) control patents that mirrors our tents in both technical subject and distribution over ing Hegde et al. (2007), who applied this approach to eterminants of the decisions of firms whether to apply uation patent, we use a binomial logit model for assess- rminants of essentiality declaration decisions. e case of W-CDMA DMA standard is a third-generation (3G) standard for communications. It was designed as a successor to ful, second-generation GSM standard. In addition to e costs per call in comparison with earlier systems, W- also designed to support multimedia mobile phones commodate web-based applications and offer phone- and video facilities. This required not only much higher n speeds than in earlier generations but also much more different users might require totally different speeds, eds demanded by a given user may change a lot over et these requirements, new radio technologies had to d. Furthermore, the transition from second generation d-generation (3G) technologies was much more than a upgrade, it also required a reconfiguration of the whole ork (Tilson and Lyytinen, 2006), which also affected the tion process. research and standardisation activities for a European ation standard commenced in the early 1990s, even rst commercial GSM network was launched.2 Several ogrammes funded by the European Commission stud- radio technologies, and one of these (in earlier stages n does provide some references, but for a more extensive overview ment and for a detailed list of references we refer to Bekkers (2001). never a manuf attemp 3G sta ers to b specifi Europe DoCoM compa engage and Lu chance region cess. T ‘conce propos was ba suppor ter wa in the what t sion p as the body A countr be com resulti referre Multim nation mobile not the provid cdmaO in coun and th is limi Divisio help of cess of itself a For for ass standa vance. 2010 a estima global these p is alrea W-CDM require tions, telecom is a larg tions in to W-C 3 Sour 2011.ved any success outside its domestic market and whose ers had only a minimal role in supplying GSM products, o leapfrog the other world regions by swiftly adopting a and contracting both Japanese and other manufactur- test systems. In fact, the Japanese standards body ARIB echnology that was largely based on the outcome of the MA2 research programme. Japan’s largest operator NTT t only ordered its experimental network with Japanese such as NEC, Fujitsu, Matsushita and Mitsubishi but also n-Japanese firms, including Ericsson, Nokia, Motorola, . By involving foreign suppliers, it tried to increase its aving the W-CDMA technology adopted in other world Japanese moves prompted Europe to speed up its pro- ical proposals by participants were clustered into five ups’, of which the so-called alpha proposal and the delta ere regarded as the two most promising ones. The first argely identical to the technology selected by Japan and by Nokia and Ericsson, among others, whereas the lat- cked by Alcatel and Siemens, who were not involved nese activities at that point in time. In early 1998, in d out to be the most difficult and controversial deci- s ETSI ever experienced, the alpha proposal came out er. Soon after, ETSI agreed with the Japanese standards and several other telecom standards bodies from other align their efforts to ensure their 3G standard would ble, and the 3GPP partnership project was born. The andard is known as W-CDMA, but within Europe is also as UMTS and in Japan as FOMA (Freedom of Mobile Access). Typically, W-CDMA is implemented in combi- GSM, its successor and the most successful generation dard. W-CDMA is the most successful 3G standard, but one. One competing technology is called cdma2000. It ll backward compatibility with the second-generation tandard, and for that reason it is mostly implemented s that applied its predecessor, particularly South-Korea ted States (Lee et al., 2009). Its overall success, however, third competitor is Time Division Synchronous Code ltiple Access (TD-SCDMA), developed in China with the e international companies like Siemens. The future suc- standard is unclear, and is likely to be limited to China t. ous reasons, W-CDMA is a particularly attractive case g the determinants of essential patents in compatibility First of all, it is a standard with a major economic rele- orldwide market for mobile telephony as of September nted to just above 5 billion subscribers,3 The World Bank global telecommunications spending of about 2.5% of in 1990 (Insight Research Corporation, 2009). Although es do not yet all support W-CDMA, their share in Europe ound 70% and growing quickly (IDATE, 2007). Secondly, an attractive case because in order to meet the design ts, it included a number of radical technological solu- senting a new technological trajectory in the area of nications. Finally, and possibly most importantly, there d complete database publicly available of firms’ declara- ting which of their own patents they consider essential . The availability of such data is a relatively rare occur- ttp://www.4gamericas.org, at ‘statistics’, consulted on January 24, R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 1005 rence in this field. For all these reasons, the W-CDMA case seems an almost ideal test bed for assessing the determinants of essential patents. 3.1. Data sources In this s all patents standard. A line by ETSI These decla matter of pr In other wo ment of the W-CDMA t patent own dard and pa of registrati of its format on a progr database fo ing procedu The databas contains 18 that are not tion of esse projects rel in total) hav all patents f ing that firm market. The US p matched w Brownie Ha NBER (Nati detailed in between Jan to these pat study we us by Bart Ve employing patents gran way it was the NBER da citations, et essential fo patents reg this study w is that US p value of the legal obliga (‘duty of can it risks to be tary; an app (Criscuolo a of studies s Albert et al. regarded as to compare we constru claimed to 4 An on-lin http://webapp patents are randomly selected from a defined population.5 These patents are also matched with the NBER US Patent Citation Data File. In order to make the non-essential patents comparable to the essential patents we have established a number of requirements that the random selection queries procedures had to satisfy. Firstly, the patents in the control data set should obviously not occur in t of essential W-CDMA patents, since we do not want any ial patents in the control data set.6 Secondly, the distribu- patents over patent classes in the control class must be to the distribution of the original set of essential patents we only included those classes with a share of 3% or higher). sures that essential W-CDMA patents are compared with ol set of non-essential patents covering the same technical Finally, the essential and non-essential patent set must cover e time-periods in terms of application year. Since the oldest ial patents were applied for in 1979, the patents in the con- ta set were retrieved from the period 1979–2003. Following rocedures, we have constructed a control group of about non with e cor of n d as s (0 = ays a hnol ned arch isten chno he v pate at th nolo learl ns th r of nges perc p. 43 d-eff g the e ave tion logic rent ce we logic t ex sampl x pate andar sidere nts (1 func preve oved should urce- ndard r, shou as bee of es ximat 2% (2 essentudy we use the latest available data set containing declared by their owners as essential to the W-CDMA database of all these declarations is made available on- 4; we retrieved the relevant declaration in March 2008. rations are sent to ETSI by its individual members. As a inciple, ETSI is not responsible for judging these claims. rds, to be included in the database there is no assess- “essential” nature of the content of these patents for echnology. The database contains information on the er, patent title, patent and application number, the stan- rt of the standard the patent related to, and the country on. However, the data and in particular the consistency , leave much to be desired (ETSI has currently embarked am to improve that). Since ETSI does not check this r errors or duplicates, an extensive cleaning and sort- re is necessary to prepare the database for this study. e we used was updated by ETSI in March 2008, and it ,738 patents. 7090 patents are essential to standards relevant for our research, for example GSM. Our selec- ntial patents is restricted to those patents belonging to ated to W-CDMA. Withdrawn and inactive patents (88 e been deleted from the selection. We continued with or which we could identify a US patent number, assum- s would at least seek to have a patent in this important atent application numbers from the ETSI database were ith the ‘NBER Patent Citation Data File’ compiled by ll and her co-workers (Hall et al., 2001). The original onal Bureau of Economic Research) data set comprises formation on almost 3 million U.S. patents granted uary 1963 and December 1999 and all citations made ents between 1975 and 1999 (over 16 million). In this e an updated version of the NBER data file constructed rspagen (see Fontana et al., 2009 for another study this updated version). This updated version contains ted up to the year 2003, including citations pairs. In this possible to retrieve the patent information included in tabase (inventor, applicant, title, primary USPTO class, c.) concerning 752 US patents that have been declared r the W-CDMA standard. The ETSI data set also contains istered in different countries and patent offices, but in e only use the USPTO patents. The reason for this choice atent citations are probably more representative of the patent, because in the US system, an applicant has a tion to cite all relevant patents and prior knowledge dor’). If a patent applicant fails to cite relevant patents, sued. In Europe, patent citation is more or less volun- licant cannot be punished for not citing relevant patents nd Verspagen, 2008). As mentioned above, a number uch as Carpenter et al. (1981), Trajtenberg (1990) and (1991) have established that forward citations may be a valid proxy for the value of patents in the US. In order essential W-CDMA patents with non-essential patents, cted a control set of comparable patents that are not be essential to W-CDMA. This is a sampled set where e version of the ETSI IPR database can be found at .etsi.org/ipr/. the lis essent tion of similar (here, This en a contr fields. the sam essent trol da these p 10,000 classes and th group structe patent able pl the tec mentio of rese the ex and te mean t in the note th of tech ages. C citatio numbe ble cha have re 2001, p a “fixe dividin with th applica techno of diffe Sin techno to wha 5 The in the si CDMA st only con tial pate selection order to were rem 6 We and reso other sta howeve cedure h estimate is appro of about claimed-essential telecom patents. Table 1 shows the patent shares higher than 3% in the essential patents group responding shares in the randomly extracted control on-essential patents. Our variable of interest is con- a dummy distinguishing essential from non-essential non-essential, 1 = essential). The forward citations vari- key role in this study, as it represents our indicator of ogical importance of each patent (Hegde et al., 2007). As in the previous section, we follow an established line relying on a number of studies that have established ce of a positive correlation between forward citations logical importance. With technological importance we alue (or ‘merit’) of the technological solution described nt for the technological field in question. We should e number of forward citations is not a good measure gical importance when comparing patents of different y, older patents had more opportunities of receiving an newer patents, so they will typically have a higher forward citations. This problem is aggravated by possi- in patent and citation practices over time, which may ussions on the number of forward citations (Hall et al., 4–437). In order to deal with this issue, we carried out ect” adjustment of the number of forward citations by number of forward citations of each individual patent rage number of forward citations of patents of the same year. In this way, we have constructed an indicator of al importance that can be used for comparing patents ages (Hall et al., 2001, pp. 437–441). use received citations as our indicator of the intrinsic al significance of the patent, it is important to consider tent a patent could actually receive (more) citations e of 10,000 control patents is drawn from the population of all patents nt classes that contain 90% of all patents claimed essential to the W- d (see Table 1 for a specification of these classes). For the selection we d those patents that fall in the same time frame as that of our essen- 979–2003). The patents were randomly selected using the random tion in a database application (we used Microsoft Access for this). In nt the extraction of duplicate patents, the essential W-CDMA patents from the population before the random sampling procedure. note that as information on essential patents is scarce, incomplete, consuming to collect, we cannot prevent that patents essential for s than W-CDMA are included in the control data set. This problem, ld not introduce a major bias in our results as the random query pro- n applied to a very large patent population. A reasonable upper bound sential patents for mobile telecom patents not claimed for W-CDMA ely 2000 patents. Thus, we have an average upper bound probability 000 out of 88,507) that a patent in our control set would have been tial in a non-W-CDMA standard. 1006 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 Table 1 Comparison of the distribution of USPTO technical classes across W-CDMA essential patents and across the sampled control croup. Primary class no. Primary class title Patents in primary class Claimed essential patents in data set (%) Non-essential patents in sampled control group (%) 370 Multiplex communication 20,831 225 (30%) 3,194 (33%) 455 Telecommunications 18,340 193 (26%) 2,716 (29%) 375 Pulse or digital communications 16,863 144 (19%) 2,045 (21%) 704 Data processing: speech signal processing, linguistics, language translation, and audio compression/decompression 6,197 48 (7%) 679 (7%) 714 Error detection/correction and Fault detection/recovery 13,842 33 (4%) 469 (5%) 342 Communications: directive radio wave systems and devices (e.g., radar, radio navigation) 12,434 32 (4%) 459 (5%) Other n/a 77 (10%) 0 (0%) 7 because of believe this patent may tial (indepe a set of ver ing) patent aware that and (ii) bel Concerning year. It typ becomes pu that the US Then, if the patent, it w and some m used to calc mentioned our databas patents up understand tain patent For W-CDM would beco ogy decisio designs we ODFM and c decade late successor, d firms may h to become official stan the first rel Furthermor surp ial—t of th ns of decl publ d ess at th t, th nt th tent d be ed i ta re pate (an e tim in 2 t, an ns, th iality cern mak tent f can he C nt co and elev ated Table 2 Average delay Application 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 88,507 the fact that it is claimed to be essential. Overall, we is not a major concern in our case. The concern that a receive citations just because it was claimed as essen- ndently from its intrinsic value) is implicitly based on y specific assumptions: (i) the holder of the later (cit- is aware of the existence of the focal patent, (ii) he is the focal patent is essential or likely to be essential ieves it is in its own interest to cite the focal patent. (i), we perform our calculations based on application ically takes a year or more before a patent application blic and before a patent is granted (if at all). Also note PTO system only allows citations to granted patents. applicant of that later patent decides to cite the focal ill take some time before this later (citing) is granted ore time before it will appear in the database we have ulate the citations received. This delay limits the afore- risks for the patents applied for in the later years of e. (Note that we consider all incoming citations from to the year 2003). Concerning (ii), it is important to at what point in time parties could be aware that a cer- would eventually be claimed essential for the standard. A, there was no way to tell which basic technologies me relevant or not until the landmark basic technol- n in early 1998. Before that moment, totally different re in the race: not only CDMA-based but also based on ompletely different basic technologies. (In fact, about a r, ODFM would be selected as the basis for W-CDMA its ubbed LTE – but that goes beyond our story.) Although ave speculated about some ‘basic technology’ patents essential, they could not have known for sure until the can be essent claim ficatio a firm to the claime note th data se mome ‘old’ pa declare publish the da recent in 1997 averag arrived data se citatio essent Con need to in a pa ‘duty o 37 of t applica scope – cite a r invaliddard was decided upon and published (for W-CDMA, ease called ‘Release-99′ was published in March 2000). e, even when both patent and standard are published, it that they be have substa that citatio between patent application and declaration at ETSI. year Patents Average delay (years) Application 1 21.00 1992 1 21.00 1993 2 20.50 1994 1 18.00 1995 3 18.00 1996 7 16.71 1997 9 15.56 1998 9 15.33 1999 10 12.50 2000 21 11.33 2001 30 9.30 2002 52 (100%) 9,562 (100%) risingly hard to assess whether a certain patent is indeed his requires the production of a ‘claim chart’ where each e patent is meticulously compared with all the speci- the standard, sometimes 10,000 of pages. Only when ares a patent to be essential, and this claim is added ic database, other firms have good information about entiality. In the context of this study, it is important to ere is a considerable delay for declaring patents. In our e average time between the patent application and the e essentiality claim was sent to ETSI is seven years. For s this is obvious: a patent filed back in 1980 cannot be ing essential before the definitive standard is officially n 1999, so the delay will be at least 19 years. However, veals that there is also a considerable delay for more nts (Table 2). For instance, patents that were applied for d presumably are granted in 1998 or 1999), still have an e lag of seven years before declaration (i.e. the claims 004, on average). We conclude that in our particular d the time period for which we considered incoming ere is only a very low chance that knowledge of patent may have affected the decision to cite essential patents. ing (iii), we would like to emphasise that companies e careful considerations before citing an existing patent application at the US Patent Office, having to meet the dor’ obligation in US patent law (Section 1.56, Chapter ode of Federal Regulations). By including a citation, an nfirms the existence of prior art and thereby limits the value – of its own patent. (On the other hand, failing to ant existing patent increases the risk of a patent being at a later date.) Patent examiners may also add citations lieve to be missing from the application. Thus, citations ntial legal consequences and are not added in the way ns are added to academic papers, for instance. year Patents Average delay (years) 27 9.33 44 8.77 72 7.86 69 7.54 80 7.06 92 6.35 104 5.38 91 4.85 41 4.27 22 3.77 31 2.19 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 1007 Table 3 Descriptive statistics. Variables Observations Mean Essential 10,311 0.072 Forward citations 10,311 9.444 Adjusted for Total patent Technologic Work items Voting weig R&D expend Sales Turnov Employees ( Continui cant names the Compu 2718 differ sive manua United Stat on Compus and assigne “mother” co 1850 differ we have tak before that merged afte our data set Compustat employees, ditures ove our applica collected re the standar In order the standar The first va in actual dra features, or tified if at l under way, to and prog (Bar and Lei 3GPP suppo actual degr of new wo (2008), wh project coo ing groups, members in ever, we sho “institution number of w degree of pa uine techno areas in qu the voting are assigne ETSI. In tur telecommu in units and contributin end of our d 7 For details USP ial fo ue (c r the tion num e 18 by 5 scrip le 3 es of s esse ely lo on-e I me um tems is ze e. Mo 5,10 senti ethod one S e ke st es adjus n wi itatio port aten inde port es m er) a 1 sh ial W sent s dec f Coward citations 10,311 1.003 s 10,311 14989.500 al concentration (Herfindahl) 10,311 0.120 supported 10,311 7.481 ht 10,311 12.413 itures (×US$1000) 7,529 3758.062 er (×US$1000) 7,692 36271.400 ×1000) 5,824 123.226 ng with our data set, we have also matched the appli- of our patent data set with the names of companies in stat database.7 Our original patent data set contained ent company names as applicants. By means of exten- l research on a number of internet sources such as the es Securities and Exchange Commission website and tat, we retrieved information on ownership structure d patents applied by “daughters” to their respective mpanies. In this way we reduced our applicant set to ent companies. Concerning mergers and acquisitions, en 1999 as reference point: companies which merged year were considered one company, companies which r that year were added as two different companies in . This is not likely to impact greatly on our results. From we then retrieved data on the annual average number of the average sales turnover and the average R&D expen- r the period 1997–2001 for the companies included in nt set. As much as possible, all the additional data we fers to the 1999/2000 time frame, the period in which d was finalised. to capture the strategic involvement of companies in disation, we have constructed two additional variables. riable we apply is a proxy for the involvement of firms fting of the content of a specific standard. In 3GPP, new work items, are proposed, accepted, developed, and cer- east four firms support this work. Once a work item is the supporting companies are expected to “contribute ress the new work item throughout the drafting phases” ponen, 2008). As such, the number of work items within rted by a given firm, as registered by 3GPP, reflects the ee of participation of companies in the development rk. This data has been collected by Bar and Leiponen o also describe the underlying 3GPP procedures with rdination groups, technical specification groups, work- and work items. Of the over 300 firms that were 3GPP 2000, only 58 supported one or more work items. How- uld note that while the voting weight reflects the more al” role of companies in the standardisation process, the ork items supported probably reflects a more “active” rticipation in the process possibly also because of gen- logical capabilities of the different companies in the 10,314 essent for val and fo dardisa by the we hav owned 3.2. De Tab variabl guishe relativ many n an ETS maxim work i imum incom patent This es and M Teleph to be th the mo effect divisio ward c patent of US p tration patent variabl turnov Fig. essent 208 es patent oper oestion. The second variable measuring involvement is weight in the standardisation process. Voting weights d to companies in relation to the contribution fees to n, these are based on companies’ declarations of their nication-related revenues. Voting weights are assigned can range from 1 to 45. Appendix A contains a table with g fees and voting weights in ETSI. To summarise, at the ata construction efforts, we have a data set containing on the contents of this see http://www.compustat.com. mobile tele W-CDMA a comm’s hig The Swedis Its core bus the 1990s to tructure. Th US-based fi There ar Interdigital In the case Std. dev. Min Max 0.260 0.000 1.000 20.153 0.000 617.000 2.137 0.000 70.111 15308.510 1.000 53595.000 0.188 0.011 1.000 11.190 0.000 34.000 15.876 0.000 45.000 6764.148 0.650 25471.930 49172.590 0.000 171607.400 111.314 0.000 468.200 TO patents, including 752 patents that are claimed to be r W-CDMA. For each of these patents we have proxies aptured by the adjusted number of forward citations) strategic involvement of the patent owner in the stan- process as captured by voting weights within ETSI or ber of work items supported. Concerning the applicants, 50 firms in our data set, while the essential patents are 0 firms. tive statistics shows the descriptive statistics for the most important our data set. Essential is the binary variable that distin- ntial from non-essential patents. The mean is obviously w, because the data set contains roughly ten times as ssential patents as essential patents. Firms that are not mber have no voting weight within ETSI whereas the voting weight is 45. The maximum number of 3GPP supported by a firm is 34 (Ericsson), while the min- ro. Ericsson is also the company with the highest net st patents have zero forward citations, while USPTO 3,459 has the highest number of forward citations (617). al patent is owned by Qualcomm and the title is “System for Generating Signal Waveforms in a CDMA Cellular ystem”. A perusal of the patent reveals that this is likely y technology for CDMA. Qualcomm is also the firm with sential patents (208). The highest value for the fixed- tment of forward citations of an individual patent by th the yearly average is 70.1111 (variable adjusted for- ns). We have used two variables for characterising the folio of the patent owners. The first is the total number ts owned, whereas the second is the Herfindahl concen- x, indicating the degree of concentration of the firm’s folio over the US primary patent classes. Finally we have easuring the size of the companies (employees and sales nd the R&D expenditures. ows the 12 firms with the highest number of claimed -CDMA patents. The US-based firm Qualcomm owns ial W-CDMA patents, which is more than 25% of all lared to the standard. Qualcomm is the original devel- de Division Multiple Access (CDMA) technology for communications networks. The implementation of a ir interface in the W-CDMA standard accounts for Qual- h number of essential patents (Bekkers and West, 2009). h company Ericsson is in second place with 155 patents. iness moved from the production of mobile phones in the development and production of the cellular infras- e third place (94 patents) is held by Interdigital, another rm. e two aspects that set the companies Qualcomm and apart from the other large owners of essential patents. of Qualcomm, technology development and licensing is 1008 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 12 larg Table 4 Propensity to Firm Qualcomm Ericsson Interdigital Nokia Motorola Philips NEC Matsushita Siemens Samsung Toshiba Nortel netw a key eleme strong posi Interdigital firm and h licensing it industry. W guish them aim at comp mobile netw Table 4 lio of the c patents. It s twelve larg these firms standard (T two numbe as Qualcom by a relativ whereas th portfolios w company w Overall, the patents pla specialists” the associat 8 Qualcomm chipsets, the c nications func “MediaTek unFig. 1. Essential W-CDMA patent owners, claim patents as essential, 12 largest firms. USPTO patents claimed to be essential (our database) 208 155 94 48 44 22 22 18 18 16 14 orks 13 nt in its business model, although the firm also holds a tion in the market for components for mobile phones.8 is a non-producing entity (NPE) or ‘technology-only’ as a business model that is almost entirely based on s patented technologies to other firms in the telecom e call these two firms ‘technology specialists’, to distin- from firms that are vertically integrated and primarily eting in the product market for mobile telephones and ork infrastructure. examines the structure of the telecom patent portfo- ompanies owning the majority of W-CDMA essential hows the number of essential patents claimed by the est claimants, as well as the total number of patents of in the technological classes relevant for the W-CDMA able 1). A third column shows the ratio between these rs. The table reveals that “technology specialists” such m and Interdigital own patent portfolios characterised ely large share of claimed essential W-CDMA patents e (often larger) vertically integrated firms have patent ith smaller shares of claimed essential patents (the ith the largest share is Ericsson with a share of 5.11%). impression one gets from Table 4, is that essential y a key role in the technology strategies of “technology . This is understandable if one realises that patents – and ed royalty incomes – are key to the business models of has a 38% revenue share in the global market for so-called baseband ore component of a mobile phone in which the mobile telecommu- tionality is integrated. Source: The Free Library (March 15, 2010): seats TI as No.2 cellular baseband chip vendor”. these firms on the prod The glob in fourth p ing of pate number of an early pe CDMA, whi W-CDMA. Table 5 c folio of ess technologic computed a forward cit Table 4 con with the m the techno lio is higher these figure an average than 1 for m the only fir patents is lo fewer forw tial W-CDM might sugg egy for its value as es that Sieme Another ex CDMA pateest claimants. Total patent stock in relevant subclasses Ratio 795 26.16% 3033 5.11% 219 42.92% 1520 3.16% 4254 1.03% 1791 1.23% 3567 0.62% 1382 1.30% 1466 1.23% 1023 1.56% 1131 1.24% 1364 0.95%, whereas for integrated firms, the freedom to operate uct market is often more important. al market leader in producing cell phones, Nokia, is lace with 48 essential patents. Fig. 2 shows the tim- nt applications for the five firms owning the highest essential W-CDMA patents. Qualcomm’s series shows ak, which may be related to the development period of ch is the predecessor of the underlying air interface for ompares the technological importance of a firm’s port- ential patents with that of non-essential patents. The al importance of a firm’s patent portfolio has been s the average of the “fixed-effect” adjusted number of ations of the patents belonging to the firm’s portfolio. siders four different periods and contains the 14 firms ost essential W-CDMA patents. The expectation that logical importance of a firm’s essential patent portfo- than its non-essential patent portfolio is confirmed by s. Most of the average essential patent portfolios have citation ratio higher than 1, while this value is lower ost of the non-essential patent portfolios. Siemens is m where the technological importance of its essential wer than 1 (which means that its essential patents have ard citations than the yearly average). Siemens’ essen- A patents have an average citation ratio of 0.84. This est that Siemens has used a suspicious patenting strat- W-CDMA patents, claiming patents of below average sential. However, we should note, as Table 4 indicates, ns’ non-essential patents are also of less importance. planation for Siemens’ low importance of essential W- nts might be that Siemens was one of the firms that R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 1009 120 19 93 Yea All firms om m Table 5 Average techn Company na Qualcomm Nokia Ericsson Interdigital Siemens Motorola Samsung Philips NEC Alcatel NTT Nortel Matsushita Toshiba focused on Alcatel, trie lying air int the compet Siemens’ lo Table 6 between th Since our v citations) a parametric number of f tance of the within ETSI standardisa Interestingl that essent number of also owned ing more w evidence su patent and determinan 4. Multiva In this se of essential0 20 40 60 80 100 19 81 19 82 19 83 19 84 19 85 19 86 19 87 19 88 19 89 19 90 19 91 19 92 USPTO Applicaon Es se n al P at en ts Nokia Ericsson Interdigital Qualcomm Fig. 2. Timing of essential patents by telec ological importance of firms’ patent portfolio (USPTO patents). me Patents claimed as essential 1986–1990 1991–1995 1996–2001 All 10.27 3.01 2.98 3.21 3.14 1.15 1.28 1.21 1.64 1.93 2.28 2.14 4.44 1.77 1.35 1.46 – 0.67 0.84 0.84 4.25 1.62 1.29 2.25 – 0.79 1.80 1.74 3.96 0.41 0.51 1.62 3.90 0.43 2.76 2.31 – 1.18 1.14 1.15 – – – – – 0.76 3.10 2.71 – 1.36 0.98 1.09 0.89 0.56 0.00 0.71 the development of TD-SCDMA. Siemens, together with d to convince ETSI to choose TD-SCDMA as the under- erface for the W-CDMA standard. So ETSI’s choice for ing W-CDMA technology in 1999 might be a reason for w importance of essential W-CDMA patents. reports a preliminary assessment of the differences e samples of essential and non-essential patents. ariables of interest (in particular the adjusted forward re skewed and not normally distributed, we use a non- Mann–Whitney test. Table 5 compares the adjusted orward citations, our proxy for the technological impor- patent and work items supported and voting weight , our proxies for the influence of a specific firm in the tion process for essential and non-essential patents. y enough, the Mann–Whitney tests in Table 5 indicate ial patents are characterised by a significantly higher adjusted forward citations, and at the same time, are by companies with higher voting weights or support- ork items within ETSI. In other words, this preliminary ggests that both the intrinsic technological value of the strategic considerations by companies do play a role as ts of essential patents. riate analysis ction we probe further into the possible determinants patents estimating a number of logit regression mod- els. Our dep ‘1′ if the pa not claimed models. In factors as d essential is we use the ing the role the variable comparing factors, we by normalis and standa note that in for controll that, as we ogy special by relativel B presents t Table 7 p In this set, are the key as control v tions and w level in all differences ables, overa19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 r arket leaders. Patents not claimed as essential (sampled control group) 1986–1990 1991–1995 1996–2001 All – 1.36 1.38 1.32 1.74 1.00 1.05 1.05 1.45 1.28 1.00 1.05 – 0.79 0.46 0.59 0.33 0.52 0.60 0.56 1.10 0.79 0.95 0.94 0.27 0.83 0.56 0.57 0.71 0.49 0.54 0.55 0.66 0.59 0.43 0.56 0.67 0.62 0.69 0.66 1.09 0.96 1.30 1.11 1.09 1.15 1.12 1.12 0.77 0.67 0.63 0.65 0.54 0.77 0.96 0.80 endent variable is the dummy essential which equals tent is claimed essential and equals ‘0′ if the patent is as being essential. We estimate two sets of regression the first set the variable capturing the role of strategic eterminant of the probability of a patent being claimed the number of work items supported. In the second set voting weight instead. In both cases the variable assess- of the intrinsic value of the patent is measured using “adjusted forward citations”. Since we are interested in the relative effect of patent value with that of strategic estimate standardised coefficients which are obtained ing all the co-variates so that they have a mean of zero rd deviation of one (Pampel, 2000). It is important to all specifications, we introduce two dummy variables ing for patents owned by Qualcomm and Interdigital have seen in the previous section, are the two “technol- ists” companies owning patent portfolios characterised y large shares of W-CDMA essential patents. Appendix he correlations between our variables. resents the results of the first set of regression models. adjusted forward citations and work items supported variables of interest. The other co-variates are included ariables. The coefficients for both adjusted forward cita- ork items supported are positive and significant at 1% regression models. Although the results show some in the estimated coefficients of these explanatory vari- ll they coefficients remain relatively stable across the 1010 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 Table 6 Descriptive statistics for the sample of claimed essential W-CDMA patents and the patents not claimed as essential (as present in the sampled control group). Note: In all three cases the Mann–Whitney test rejects the hypothesis of equal populations (p < 1%). Variable Sample/group Observations Mean Median Std. dev. Min Max Adjusted forward citations Claimed essential 691 2.242 1.394 3.098 0 35.055 Not claimed essential 8859 0.907 0.464 1.885 0 35.055 Work items supported Claimed essential 691 16.214 15.000 13.211 0 34.000 Not claimed essential 8859 7.279 0.000 10.975 0 34.000 Voting weight Claimed essential 691 27.852 30.000 16.095 0 45.000 Not claimed essential 8859 12.046 6.000 15.485 0 45.000 different sp regression ported is hi This finding dards body of the pate essential. It is wor control var ables contr significant. puted over here as a p significant essential p patent port to non-esse tive and sig by Asian an being claim result is lik the technol The coef cant and po two variabl employees Again this r the patents in telecom, rather than Table 8 explanatory voting weig involvemen larly intere whereas th reflecting al weight pro malised ins process. Th line with th also has a s all the mod higher than stronger ro 9 We obtain also estimatin specialists” co 10 The first tw tures extracted all firms in our observations. t of th at th than e 7. T with once The s ow ial. Fi affe ial, w les tu in al onfi d ess proc ial be clus que senti f the o the vene oul pen ade a tion ts for pate our o chni is wo tually eatio barr rmor rd fro sed b erved stud the W nsic ecifications.9 Furthermore, our results show that in all models the coefficient of the variable work items sup- gher than the coefficient of adjusted forward citations. indicates that the influence of companies in the stan- exerts a greater impact than the technological value nt on the probability of the patent being claimed as th discussing briefly the estimated coefficients of the iables. As expected the sign of the two dummy vari- olling for Qualcomm and Interdigital are positive and The coefficient of the Herfindahl Index (which is com- the technological classes of the patent owner) is used roxy for the technological diversity of firms and has a and positive influence on essential. This indicates that atents are much more often owned by firms whose folio is concentrated on one technology field, compared ntial patents. The dummy for Asia and Europe are posi- nificant in all regressions indicating that patents owned d European companies have a higher probability of ed as essential for W-CDMA. Note, however, that this ely to be affected by the two dummies that control for ogy specialists companies. ficient of the variable for R&D expenditures is signifi- sitive in the model of column 2.10 Note instead that the es that we use to control for the size of the companies and sales turnover are significant with a negative sign. esult may be probably interpreted as an indication that owned by companies with specialised patent portfolios have higher probability to be claimed as essentials than patents owned by large and diversified companies. presents six regression models where the two key variables of interest are adjusted forward citations and ht. We use voting weight as a proxy of dimension of t in the standardisation process and we are particu- sted in this set of regression, because as noted above, e variable work items supported may be thought as so some technological competences, the variable voting bably captures more adequately the effect of the for- titutional power of companies in the standardisation e coefficients of adjusted forward citations are fully in ose estimated in the models in Table 7. Voting weight ignificant and positive influence on essential. Again, in els estimated, the size of the standardised coefficient is adjusted forward citations and this finding confirms the le of the involvement with standards bodies as a deter- minan also th higher in Tabl sistent more c effect. patent essent ditures essent and sa All 2 are c claime isation essent 5. Con The are es result o value t effecti firms w the com ogy m innova marke ever, if behavi uine te that th and ac dard cr create Furthe standa expres he obs Our that in of intri the same results concerning the different size of the two coefficients g a similar set of models that do not control for the two “technology mpanies. o columns contain control variables for firm size and R&D expendi- from the Compustat database. As this data could not be retrieved for sample, these models are estimated on somewhat lower numbers of 11 Again, we ficients also e “technology sp 12 “[There is nology into t standards rel EC/EPO Work dardisation: r http://ec.euroe probability of a patent being claimed essential.11 Note e coefficient of the variable voting weight is in general the corresponding coefficient for work items supported he results for the other control variables are fully con- those of Table 7. Also in this set of models, a smaller and ntrated patent portfolio has a significant and positive regional dummies again reveal a higher probability of ned by European and Asian companies of being claimed nally, also in this set models, the variable for R&D expen- cts positively the probability that a patent is claimed as hereas the variables controlling for size (employment rnover) have a negative impact. l, in both of our models, Hypothesis 1 and Hypothesis rmed, and we show that the second one (patents are ential because their owners participate in the standard- ess) is stronger than the first one (patents are claimed cause of their high intrinsic value). ion and discussion stion of what makes companies claim that their patents al for a technical standard is fascinating. If this is a technical value of the patented technology (attributing standard, such as increased performance, better cost- ss, etc.), there should be no concern. If non-producing d be licensing these patents for monetary return, and sation is in line with the technical value of the technol- vailable, then it would be a good example of the open model as described by Chesbrough et al. (2008) and the technology as suggested by Arora et al. (2004). How- nts are claimed to be essential as a result of the strategic f the participants in standards bodies, but have no gen- cal value, we have a different story. It can be argued uld be undesirable from a public welfare point of view incompatible with the fundamental principles of stan- n as it would unnecessarily restrict access to standards, iers to their use and adoption, and increase their costs. e, such behaviour could unnecessarily complicate the m a technical point of view, a concern that was recently y the chairman of ETSI’s IPR Special Committee when an increase of what he called ‘marginal patents’.12 y on the determinants of essential patent claims shows -CDMA case, essential patent claims are both the result technical value (‘merit’) and of the involvement of their obtain the same results concerning the different size of the two coef- stimating a similar set of models that do not control for the two ecialists” companies. a] risk of complicating the solutions just for getting patented tech- he standard rather than to improve the standard.” Dirk Weiler, ated patent Quality A view from a standardisation organization, shop on “Tensions between intellectual property rights and stan- easons and remedies”, Brussels, 22 November 2010. Available from pa.eu/enterprise/sectors/ict/files/ict-policies/5 weiler en.pdf. R . Bekkers et al. / R esearch Policy 40 (2011) 1001– 1015 1011 Table 7 Logistic regressions (standardised coefficients). Dependent variable is “Essential”. (1) (2) (3) (4) (5) (6) Adjusted forward citations 0.294*** (0.0415) 0.320*** (0.0373) 0.299*** (0.0319) 0.283*** (0.0310) 0.278*** (0.0310) 0.278*** (0.0309) Work items supported 0.567*** (0.125) 0.661*** (0.0817) 0.730*** (0.0577) 0.912*** (0.0430) 0.908*** (0.0432) 0.908*** (0.0427) Application year −0.0125 (0.0151) −0.0114 (0.0129) −0.00332 (0.0119) 0.000533 (0.0118) −0.000113 (0.0116) Total patents 0.568*** (0.169) 0.515*** (0.126) 0.141** (0.0692) 0.225*** (0.0682) Technological concentration (Herf.) 0.747*** (0.235) 0.766*** (0.203) 0.296** (0.122) 0.391*** (0.106) Europe dummy 0.743*** (0.190) 0.977*** (0.188) 1.008*** (0.158) Asia dummy 0.789** (0.375) 0.156 (0.227) 0.696*** (0.170) R&D expenditures 1.070 (0.769) 0.881*** (0.263) Employees −0.552*** (0.207) Sales turnover −0.943*** (0.330) Qualcomm dummy 4.656*** (0.228) 4.660*** (0.205) 4.807*** (0.179) 4.444*** (0.162) 4.372*** (0.153) 4.372*** (0.151) Interdigital dummy 4.032*** (0.412) 3.912*** (0.360) 4.362*** (0.301) 3.758*** (0.277) 3.999*** (0.255) 3.999*** (0.254) Constant 21.25 (30.09) 18.86 (25.65) 2.638 (23.75) −4.603 (23.58) −3.345 (23.08) −3.571*** (0.0721) Observations 5480 7363 9550 9550 9550 9550 Log-likelihood −994.0 −1441 −1657 −1681 −1688 −1688 Pseudo R2 0.379 0.342 0.332 0.322 0.319 0.319 Standard errors in parenthesis. *Significant at 10%. ** Significant at 5%. *** Significant at 1%. Table 8 Logistic regressions (standardised coefficients). Dependent variable is “Essential”. (1) (2) (3) (4) (5) (6) Adjusted forward citations 0.275*** (0.0425) 0.308*** (0.0390) 0.296*** (0.0332) 0.284*** (0.0323) 0.280*** (0.0321) 0.281*** (0.0321) Voting weight 0.524*** (0.138) 0.809*** (0.111) 0.949*** (0.0860) 0.956*** (0.0468) 0.939*** (0.0461) 0.940*** (0.0457) Application year −0.0172 (0.0151) −0.0126 (0.0129) −0.000742 (0.0120) 0.00235 (0.0119) 0.00208 (0.0117) Total patents 0.466*** (0.163) 0.363*** (0.110) 0.177*** (0.0666) 0.234*** (0.0636) Technological concentration (Herf.) 0.868*** (0.229) 0.839*** (0.199) 0.456*** (0.113) 0.409*** (0.106) Europe dummy 0.450* (0.246) 0.189 (0.230) 0.455** (0.191) Asia dummy 1.486*** (0.339) 0.608*** (0.228) 0.886*** (0.176) R&D expenditures 2.525*** (0.650) 0.428** (0.208) Employees −0.687*** (0.217) Sales turnover −0.194 (0.241) Qualcomm dummy 4.178*** (0.283) 3.650*** (0.213) 3.502*** (0.192) 3.171*** (0.149) 3.102*** (0.140) 3.104*** (0.140) Interdigital dummy 4.916*** (0.400) 4.853*** (0.366) 5.247*** (0.296) 4.971*** (0.287) 5.196*** (0.266) 5.201*** (0.265) Constant 30.99 (30.12) 21.43 (25.79) −2.425 (23.86) −8.222 (23.73) −7.690 (23.30) −3.546*** (0.0718) Observations 5,480 7,363 9,550 9,550 9,550 9,550 Log likelihood −997.3 −1446 −1677 −1691 −1700 −1700 Pseudo R2 0.377 0.339 0.324 0.318 0.315 0.315 Standard errors in parenthesis. * Significant at 10%. ** Significant at 5%. *** Significant at 1%. 1012 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 holders in the standardisation process (creating room for strategic conduct). We observe, however, that the involvement in the process is a stronger determinant than the technical value of the patent. While our findings are in line with several other studies providing evidence of the existence of both effects, this study – to the best of our knowledge – is the first one that has considered both effects simultaneously. Assuming that forward citations do reveal patent value, we con- clude that technical value is a significant determinant of claims of essentiality. We find a positive, significant and robust relationship, and looking a higher th was not ac at the time ously becam most of the relied on th account.) If we see tha on essentia firms show are of high behaving st the involve the voting w involvemen tered ‘work involvemen clude that s of claims fo more essen is found to of the paten We will Although pr dards bodie performanc patent own suggest suc dards bodie changes in m in the mark three posit that were h the numbe been captu ities were n Motion (ma new entran ever these c patents is n reliable dat hard to fin up to 30% f given by a re (Interplay, ing fees for 13 For GSM, Telecommunic fees make up 2 claim that “Es 30% and the m payments up MobileMonda 12% of the wholesale (‘ex-works’) value of these devices. Firms that have a reasonably strong patent portfolio themselves can negoti- ate cross-licenses and would typically be paying around 8% (which already signals the value of owning such patents). Also the high number of high-profile patent cases in this field signals the eco- nomic importance of these patents. These numbers illustrate the substantial sums of money involved and the possible distortion of the market if these sums are paid for patents that do not have actual technical v strategic va ger a ities infri e U. in bo t Qua es (a law es o 15 ing n em tha ial pa ly th here . Thi rld. A ropea , and laim rds i ries. ve (m rds w con odin f the ird ca ecific tand rds h exam egor or w tand , man D po larly sett es an ding tego shed ools s are luat sent ding loom 009. Financ he Ec t amo at the data, we indeed observe sets of patents with an average value even though the owner (Qualcomm) tively participating in the standards body in question the technology was determined. These patents obvi- e essential by virtue of their technical value. (In fact, different 3G candidate technologies put forward in 1999 at same technology; see Bekkers (2001) for a detailed we remove the Qualcomm patents from our database, t the influence of the technical importance of patents lity is reduced, while at the same time the remaining an even stronger behaviour. So Qualcomm’s patents technical importance, and this firm was probably not rategically in the standard setting process (in terms of ment we are capturing in our study). Assuming that eight a firm has in the standards body, and the actual t in the standards drafting process (captured by regis- item support’), are both proxies of the opportunities for t and influencing the content of the standard, we con- trategic involvement is also a significant determinant r essential items. Being an active member helps to get tial patents. Furthermore the status of active member be a stronger determinant than the technological value t as captured by forward citations. now further discuss the implications of these findings. evious studies on the era before the emergence of stan- s have shown cases in which market access and firm e was highly dictated by patent strategies of essential ers (Bekkers et al., 2002), it would be unreasonable to h an effect in the current context where formal stan- s have (F)RAND regimes. Also, we do observe significant arket share over time in the market we examine here: et for mobile terminals, the current number two and ion are for Samsung and LG, two South-Korean firms ardly active on the global market a decade ago, and r two position in mobile infrastructure has recently red by a Chinese supplier, Huawei, whose global activ- egligible even shorter ago. Also Apple and Research in ker of the BlackBerry phone) are examples of successful ts, outside the set of ‘incumbent champions’. How- onsiderations do not mean that ownership of essential ot extremely important in the telecom sector. While a on actual licensing fees for specific technologies are d, some parties have been claiming licensing fees of or mobile devices,13 much more realistic estimates are cent study commissioned by the European Commission 2010). Here, it is estimated that the aggregate licens- firms implementing W-CDMA phones is approximately the predecessor of W CDMA, the director of the European public ations Network Operators’ association (ETNO) revealed that royalty 9% of the costs of GSM handsets (Taaffe, 2000). Some industry analysts timates for cumulative royalties for W CDMA are between 25% and obile industry could spend US$80-100 billion on W CDMA-IP-royalty to 2017′′ (Poropudas, T., 2006. ETSI moves to exclude Qualcomm. y Global: Industry News, 1 March 2006). to trig author patent with th digital agains royalti filing a infring Apple. Hav questio believe essent certain tings w applies the wo the Eu bodies were c standa catego that ha standa ‘IT and video c most o The th are sp RFID s standa esting the cat dards f these s bodies (F)RAN particu in such nologi our fin cific ca establi such p patent nal eva non-es our fin 14 See B ber 24, 2 15 See well as T legal spaalue. In fact, it is likely that the pivotal monetary and lue associated with essential patents have contributed number of high-profile legal cases and complaints to , such as Qualcomm suing both Nokia and Broadcom for ngement in the US and the UK (including a complaint S. International Trade Commission), Nokia suing Inter- th the US and in the UK, six large firms filing a complaint lcomm with the European Commission for excessive case that was dropped later on),14 and, recently, Nokia suit against Apple in the US, arguing that the iPhone n its essential patents, followed by a countersuit by performed this analysis for one single standard, the erges how generalizable our results are. In general, we t our findings will hold for all technologies for which tents are claimed by their respective owners. This is e case for technologies that are standardised in set- a (F)RAND policy with declarations of essential patents s includes virtually all formal standards bodies around recent study on patents in standards commissioned by n Commission analysed eleven major formal standards identified more than 200 standards for which patents ed (Interplay, 2010). The number of claimed patents for s very skewed, however. The report distinguishes four The first, ‘Telecom standards’ includes seven standards uch) more than 100 patents claimed, and another 20 ith less than 100 claimed patents. The second category, sumer electronics’, includes many picture, audio and g standards, and computer interfaces (e.g. FireWire). For se standards, between 20 and 100 patents are claimed. tegory, ‘Enabling technologies’, includes standards that ally designed for certain application areas. Here, the ard has more than 100 claimed patents, many other ave typically one to 20 patents claimed (one inter- ple is ‘agricultural electronics identification’. Finally, y of ‘Genuine non-ICT related standards’ has few stan- hich patents are claimed, and the number of patents for ards is generally low. In addition to the formal standards y standard-developing fora and consortia have similar licies with declaration rules as well. Many technologies, consumer electronics standards, have been developed ings. (Well-known pools are the two pools for DVD tech- d the pool for MPEG coding standards.) We expect that will hold for such type of organisations as well. A spe- ry would be those standards for which patent pools are . In order to comply with antitrust/competition laws, virtually always include a procedure where candidate thoroughly tested for essentiality, usually by an exter- or. Because this limits the possibility of parties claiming ial patents to be essential we expect that, compared to s, technological value will prevail over strategic involve- berg. Qualcomm Antitrust Probe Is Dropped by EU Regulators. Novem- ial Times, 22 October 2009, “Nokia takes on iPhone in legal fight”, as onomist (online edition, 23 October 2009), “Dial L for lawyer: A nasty ng tech giants pits Nokia against Apple”. R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 1013 ment as a determinant in such a setting. Still, patent pools do not prevent that patents are included that have low technical value yet are indeed essential in a literal sense of the definition. A recent paper by Baron and Decamp showed that firms that are already member of the pool are indeed more able to include lower value patents than ‘outsiders’ (Baron and Delcamp, 2010). Finally, it should be stressed that any patent owner is free whether or not to join a pool, while such a freedom does not exist for participants in standardisation bodies when it comes to the IPR rules, which is something that might also affect the findings. Having indicated in which field we believe our findings would hold, we still believe it would be an obvious direction of further research to test other standards or technological fields. Obviously, our study has some limitations. As already sug- gested above, we can only make claims about the specific case we discussed (although we indicated in which other situations the findings would hold). Also, patent citations are no perfect value indica is claimed ing incomin frozen). Th mention he essential p the process erately clai are aware ally determ requires the is held agai lenging suc that include when they erful player patent own risking losin could expla believe that this. Our wor be an ungro of strategic content of t nologies, va world wher that on the process wit of certain ( more transp the inclusio small techn able) that r include a trivial patent, I will allow you to do likewise. Standards bodies, as well as other stakeholders (such as the public authori- ties that recognise these bodies) would do well to study in detail the dynamics of these technical inclusion processes and consider possible reforms, for instance by introducing new incentives and safeguards. Other help could come from current efforts to allow the patent offices access to – now often confidential – documents at the time of the standardisation effort, improving their ability to deter- mine the prior art and dismiss patent applications on inventions that are not novel.16 Finally, voluntary ex-ante licensing, a process whereby patent holders can choose to reveal their licensing con- ditions before the technology gets included, may help to promote more conscious technological inclusion discussions. While the suc- cessful introduction of such processes will certainly not be easy, and the discussions on this topic are surrounded by mutual incompre- hension, the recent adoption of such procedures by the IEEE shows promise. A more far-reaching policy would be that for each known to be included, a formal assessment is performed about the tives ed al wled aut ailab . We aluab umm ence (SIIT rese dix A rce: E nics c turn user a nal m -Enter rsities t-for- 135 m 00 mi 50 mi 00 mi 350 m 2000 3500 5000 8000 8000 effor sion a Right on 22 tors, and citations might be attracted because a patent as essential (we have limited that effect by truncat- g citations after the first release of the standard was ere is one more specific limitation we would like to re: there is one more mechanism that might lead to atent claims that is not studied in this paper. Since we study involves self-declarations, firms may delib- m certain patents as being essential even when they this is not the case. Because it is quite hard to actu- ine whether a certain patent is essential or not (this development of a chart where each claim in the patent nst each element of the standard), and because chal- h patents is an expensive and time-consuming activity s, some implementers rather just pay for a license, even have some doubts. At the same time, some more pow- s might refrain from paying a license, knowing that the er rather enjoys the income from other licenses than g an infringement case. While this mechanism certainly in a specific category of essential patent claims, we there is no reliable data that would allow us to test for k has also some policy implications. Although it would unded assertion that patent claims are solely the result gaming, participants still systematically influence the he standard in the direction of their own patented tech- luable or not. Although it would be hard to imagine a e such things do not take place at all, it may be argued basis of welfare effects, society would benefit from a h more feedback loops, where the technical inclusion patented) technologies in standards would result in a arent and open process. Current decisions concerning n of specific patents in a standard are made in relatively ical groups, where it is likely (and perhaps unavoid- eciprocal favours are granted—e.g. if you allow me to patent alterna patent Ackno The data av groups their v 2009 S confer nology were p Appen Sou Electro related SMEs, additio Micro Unive and no Up to 136–2 201–4 451–7 701–1 1351– 2001– 3501– 5001– Above 16 Such Commis Property Brussels and their relative costs and benefits (including non- ternatives). gements hors would like to thank Aija Leiponen for making le on the participation of companies in 3GPP working would like to thank the two anonymous referees for le comments, as well as the participants of the DRUID er conference in Copenhagen and the 6th international on Standardization and Innovation in Information Tech- 2009) in Tokyo, where earlier versions of this paper nted. . ETSI annual contribution fees TSI website, 2008. ommunications over (ECRT) in Euro Voting weight (‘units’) Annual contribution fee (Euro) nd trade associations, embership 1 6,000 prises 1 3,000 , public research bodies profit user associations 1 2,000 illion 2 9,380 llion 3 12,760 llion 6 22,900 llion 9 33,040 illion 13 46,560 million 18 63,460 million 24 83,740 million 30 104020 million 37 127,680 million 45 154,720 ts were extensively discussed at a recent workshop of the European nd the European Patent Office (EPO): “Tensions between Intellectual s and the ICT standardisation process: reasons and remedies”, held in November 2010. 1014 R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 Appendix B. Correlations Essential Adjusted forward citations Voting weig Work items supported Total patent Technologic concentratio (Herf.) 1 Employees 0.50* Net income 0.10* R&D expenditure 0.15* Dummy qualcomm 0.07* Dummy interdigital 0.17* * Significant ** Significant *** Significant References Albert, M.B., e trially imp Andersen, P., 2 to Open St Arora, A., Fosfu of Innovat Bar, T., Leipon Setting. Pa June 17–2 Baron, J., Delca peter Soci Bekkers, R., 20 ERMES. Ar Bekkers, R., W strategic p Bekkers, R., Du technolog 31, 1141–1 Carpenter, M.P patents. W Chesbrough, H New Parad Chiao, B., Lern empirical Criscuolo, P., from? Inve 1892–190 Farrell, J., 200 603–670. Fontana, R., N as patent Economics Funk, J.L., 200 Mobile Ph Funk, J.L., Me creation a municatio Garrard, G.A., Artech Ho Geradin, D., La standard s Journal of Goodman, D.J Wireless C Granstrand, O D.C, M., Ne sity Press, Hall, B.H., Jaff lessons, in Hegde, D., Mo builders: w 13153. 2007. rt 20 sels. Resea ew: A tion, N y cons al pr missio orga . E., 19 ersial dardis er, K., er, K., ual H 4. arrar, nting nce on , et al., 2009. Technical innovation and 3.5 mobile phone generation: lessons Korea. Telecommunications Policy 33, 296–308. M.A., 2002. Intellectual property rights and standard-setting organizations. ornia Law Review 90, 1889–1980. M.A., Shapiro, C., 2006. Patent holdup and royalty stacking. Texas Law ew 85, 1991–2049. F.C., 2000. Logistic Regression. A Primer. Sage, Thousands Oaks. M., Henkel, J., Heath, C., 2007. On sharks, trolls, and their patent —unrealistic damage awards and firms’ strategies of “being infringed”. arch Policy 36, 134–154. , M., Simcoe, T., Heath, C., 2008. Patents and the performance of voluntary dard-setting organizations. Management Science 54, 1920–1934. B.N., Ziedonis, A.A., 2004. Patent citations and the economic value of patents. lanzel, F.M.H., Schmoch, U. (Eds.), Handbook of Quantitative Science and nology Research. Kluwer, Dordrecht, the Netherlands, pp. 277–299. , S.K., Werle, R., 1998. Coordinating Technology: Studies in the International dardization of Telecommunications. The MIT Press, Cambridge. C., 2001. Navigating the patent thicket: Cross licenses, patent pools, and dard setting. In: Jaffe, F.A., Lerner, J., Stern, S. (Eds.), Innovation policy the economy. National Bureau of Economic Research, Cambridge, MA, . r, M., Lea, G., 1995. Telecommunications standardisation and intellectual erty rights: a fundamental dilemma? In: Kahin, T.I.B., Abbate, J. (Eds.), Stan- s Policy for Information Structure. MIT press, Cambridge, MA, USA. G., 2009. Patent Holdup and Oligopsonistic Collusion in Standard-Setting nizations. Journal of Competition Law and Economics 5, 123. T.S., Graham, S.J.H., Feldman, M.P., 2009. Competing on standards? epreneurship, intellectual property, and platform technologies. Journal of omics & Management Strategy 18, 775–816. ., 2000. 3G patents initiative devised to avoid ‘Qualcomm-type’ disputes. municationsWeek International, June 19, 2000.1 2 3 4 5 6 1 0.15*** 1 ht 0.25*** 0.04*** 1 0.21*** 0.02** 0.78*** 1 s −0.07*** −0.06*** 0.13*** 0.11*** 1 al n 0.08*** 0.05*** −0.11*** −0.09*** −0.43*** −0.17*** −0.07*** 0.03* 0.13*** 0.77*** − 0.10*** 0.04*** 0.45*** 0.48*** 0.08*** s 0.14*** 0.04*** 0.61*** 0.69*** −0.08*** 0.41*** 0.13*** 0.19*** −0.04*** −0.14*** 0.30*** 0.01 −0.07*** 0.07*** −0.10*** at 10%. at 5%. at 1%. t al., 1991. Direct validation of citation counts as indicators of indus- ortant patents. Research Policy 20, 251–259. 008. Evaluation of Ten Standard Setting Organizations with Regard andards. IDC, Copenhagen, Denmark. ri, A., Gambardella, A., 2004. Markets for Technology: The Economics ion and Corporate Strategy. The MIT Press, Cambridge. en, A., 2008. Collaboration and Networking in Cooperative Standard per presented at the 25th DRUID Celebration Conference, Copenhagen 0, 2008. mp, H., 2010. Strategic inputs to patent pools. International Schum- ety Conference, 2010. Aalborg, June 21–24. 01. Mobile Telecommunications Standards: GSM, UMTS, TETRA and tech House, Boston, MA, USA. est, J., 2009. The limits to IPR standardization policies as evidenced by atenting in UMTS. Telecommunications Policy 33, 80–97. ysters, G., Verspagen, B., 2002. Intellectual property rights, strategic y agreements and market structure—the case of GSM. Research Policy 161. ., Narin, F., Woolf, P., 1981. Citation rates to technologically important orld Patent Information 3, 160–163. ., Vanhaverbeke, W., West, J., 2008. Open Innovation: Researching a igm. Oxford University Press, Oxford, UK. er, J., Tirole, J., 2007. The rules of standard-setting organizations: an analysis. RAND Journal of Economics 38, 905–930. Verspagen, B., 2008. Does it matter where patent citations come ntor vs. examiner citations in European patents. Research Policy 37, 8. 7. Standard setting, patents, and hold-up. Antitrust Law Journal 74, uvolari, A., Verspagen, B., 2009. Mapping technological trajectories citation networks. An application to data communication standards. of Innovation and New Technology 18, 311–336. 2. Global Competition between and within Standards: The Case of ones. Palgrave, Hampshire, UK. the, D.T., 2001. Market-and committee-based mechanisms in the nd diffusion of global industry standards: the case of mobile com- n. Research Policy 30, 589–610. 1998. Cellular Communications: World-Wide Market Development. use, Norwood, MA, USA. yne-Farrar, A., Padilla, A.J., 2008. The complements problem within etting: assessing the evidence on royalty stacking. Boston University Science & Technology 14 (2), 144–167. ., Myers, R.A., 2005. 3G Cellular standards and patents, 2005. IEEE ommunications and Networking Conference, pp. 415–420. ., 2005. Innovation and intellectual property rights. In: Fagerberg, J., lson, R.R. (Eds.), The Oxford Handbook of Innovation. Oxford Univer- Oxford, UK. e, A.B., Trajtenberg, M., 2001. The NBER patent citations data file: sights and methodological tools. NBER Working Paper No. 8498. wery, D.C., Graham, S.J.H., 2007. Pioneers, submariners, or thicket- hich firms use continuations in patenting? NBER Working Paper No. IDATE, Repo Brus Insight Revi pora Interpla lectu com shop 2010 Iversen, trov Stan Krechm 1–6. Krechm Ann p. 20 Layne-F Pate fere Lee, B.G. from Lemley, Calif Lemley, Revi Pampel, Reitzig, prey Rese Rysman stan Sampat, In: G Tech Schmidt Stan Shapiro, stan and USA Shurme prop dard Sidak, J. Orga Simcoe, Entr Econ Taaffe, J Com7 8 9 10 11 ** 1 ** 0.47*** 1 ** 0.63*** 0.81*** 1 ** −0.24*** −0.11*** −0.10*** 1 ** −0.16*** −0.07*** −0.07*** −0.02** 1 Development of Broadband Access in Europe: Methodological 07 Survey Data as of 31 December 2006. DG Information Society, rch Corporation, 2009. The 2009 Telecommunications Industry n Anthology of Markets Facts and Forecasts. Insight Research Cor- ew Jersey. ortium, 2010. Study on the interplay between standards and intel- operty rights (IPRs); Preliminary Results. Interim results of a study ned by the European Commission. Presented at the Open Work- nized by DG Enterprise of the European Commission, November 23rd 99. Standardisation and Intellectual Property Rights: ETSI’s con- search for new IPR-procedures. In: SIIT’99—IEEE Conference on ation and Innovation , Aachen, Germany, pp. 15–17. 1998. The principles of open standards. Standards Engineering 50 (6), 2005. The meaning of open standards. In: Proceedings of the awaii International Conference on System Sciences , Hawaii, A. 2008. Innovative or Indefensible? An Empirical Assessment of within Standard Setting, Presented at the CEPR-Bank of Finland con- Innovation and Intellectual Property in Finance. R. Bekkers et al. / Research Policy 40 (2011) 1001– 1015 1015 Tilson, D., Lyytinen, K., 2006. The 3G transition: changes in the US wireless industry. Telecommunications policy 30, 569–586. Trajtenberg, M., 1990. A Penny for your Quotes: Patent Citations and the Value of Innovations. RAND Journal of Economics 21, 172–187. Van Zeebroeck, N., 2011. The puzzle of patent value indicators. Economics of Inno- vation and New Technology 20, 33–62. Wilkinson, S.L., 1991. They’re stealing our diamonds: the standards assault on patents. Revue Canadienne de propriété intellectuelle 8, 193–205.