Transaction Costs and Trolls: Individual Inventors, Small Firms and Entrepreneurs in Patent Litigation

Transcription

1 Transaction Costs and Trolls: Individual Inventors, Small Firms and Entrepreneurs in Patent Litigation Gwendolyn G. Ball Research Fellow Business, Economics and Law Group Institute for Genomic Biology University of Illinois Jay P. Kesan Professor and Mildred Van.Voorhis Jones Faculty Scholar College of Law Business, Economics and Law Group Institute of Genomic Biology University of Illinois January 15,

2 Ball and Kesan, Preliminary Draft: Do not Quote, Cite or Distribute 2 Abstract Using original court documents to accurately identify the parties, outcomes and disputed patents filed in 2000, we explore how the resolution of patent cases relates to the nature of the parties. We find that small parties are quite active in our cohort of cases, constituting nearly half of all plaintiffs. However, only a small number of these parties fall into controversial categories such as universities and patent licensing firms. Nonetheless, the distribution of number of parties sued among licensing firms is highly skewed, with one organization suing over 500 defendants. We also find that the vast majority of small parties and entrepreneurs are able to overcome transaction costs and seek judgments against large firms. Small parties suing large firms are more likely to seek a judgment and pursue the case to high-cost stages of litigation than are large firms suing small parties. Thus, the potential surplus available in the case is the best indicator of its resolution; institutional arrangements, such as contingency fee lawyers, may be able to overcome the barriers to protection of property rights like the high cost of litigation faced by small firms. Keywords: Intellectual Property; Strategy, Technology and Innovation; Transaction Cost Economics

3 Ball and Kesan, Preliminary Draft: Do not Quote, Cite or Distribute 3 Transaction Costs and Trolls the Behavior of Individual Inventors, Small Firms and Entrepreneurs in Patent Litigation The patent system is a cornerstone of American innovation policy. Specialized areas of research may receive government research funds, and collaboration between public universities and private firms can provide new sources of innovation, but much of the research and development in the United States is financed and conducted by private individuals and firms, with the market determining where inventors decide to invest their resources and which inventions will succeed. However, the patent system is a two-stage process. 1 In the first stage, the U.S. Patent and Trademark Office grants property rights to inventors. In the second stage, inventors can protect those rights through patent infringement suits in the courts and alleged infringers have the right to challenge improvidently granted patents and have them declared invalid. 2 Both these functions are equally important to a successful innovation policy. If inventors cannot protect their rights, the incentive to innovate is lessened. If overbroad or invalid patents cannot be revoked, improvidently granted property rights will foreclose downstream innovation, create wealth transfers not contemplated by the patent system, and encourage investment solely to acquire patents without innovation. Unfortunately, there are fundamental problems with using the courts as a mechanism for revoking wrongly granted patent rights. There is general agreement that the transaction costs associated with pursuing a patent lawsuit are high. 3 Previous authors have cited legal costs of patent litigation running from half a million dollars to three million dollars per suit or $500,000 per claim at issue per side. 4 These costs create incentives for the parties to settle their dispute rather than seek a final judgment on the merits. 5 However, society may have an economic interest in seeing these disputes decided through a formal judgment, which neither the court nor the parties take into account. The parties may settle when both decide that the benefits to doing so exceed their private costs of continuing litigation. In short, as other authors have pointed out, pursuit of patent invalidation suffers from a free rider problem., One firm may incur the court costs, but firms incurring no costs will benefit from an invalidation ruling, too. 6 Thus, any analysis of reform to the patent system requires a better understanding of the effectiveness and cost of adjudicating the validity, infringement, and enforceability of issued patents. It is important to know exactly how patent cases are resolved-do they settle or do they proceed to some form of adjudication on the merits? It is important to understand the 1 See Kesan [9], Kesan and Banik [11]. 2 The PTO currently has two mechanisms for re-examination of previously granted patents. However, they are only used on a very limited basis compared to reexamination procedures in Europe. See Graham, et. al.[5] 3 See Barton, p See Hall. [7] 5 See Gross and Syverud [6], p The free rider problem in patent litigation is discussed in more detail in Farrell and Merges.[4]

4 Ball and Kesan, Preliminary Draft: Do not Quote, Cite or Distribute 4 costs of such cases and the magnitude of the incentives to settle, and how these incentives vary across the types of parties, technologies, etc. And it is crucial to know how these factors relate to the courts ability to correct or tailor the scope of patent rights. There is an enormous theoretical literature on how the parties in civil litigation choose whether to settle their case or resolve it through a trial. 7 Within this literature, there is a general sense that the outcome of the case will depend in the nature of the parties, i.e., their ability to overcome transaction costs, their degree of risk aversion, in sum, their relative ability to bargain for a settlement agreement or continue the case to judgment. Thus, individual inventors, small firms and entrepreneurs may be at a disadvantage when defending their intellectual property rights in the courts. However, models are frequently constructed with no reference to the actual bargaining strength of parties. Empirical studies tend to deal with this issue more explicitly. In an exhaustive review of empirical work on civil litigation, Kessler and Rubinfeld, 8 summarize the empirical literature on civil litigation, including of the general propensity to settle versus go to trial and how that relates to the parties in the case. They note that difficulty in discussing this literature is that the institutional and legal environments vary substantially across types of civil litigation. For example, there may be evidence that the relative size of parties is important in the litigation of medical malpractice disputes. 9 However, it is difficult to translate this result to other areas of civil litigation. The authors conclude: The particulars of settlement behavior depend on the nature of the parties (firms or individuals, risk neutral or risk-averse, etc.) on the nature of the cases (large stakes, small stakes, reputation effects, etc.) and more generally on the institutional characteristics associated with the subject matter at issue 10 Patent litigation provides an interesting case in this context, since it operates in a different institutional environment from other forms of civil litigation. While malpractice and tort suits are largely a matter of state law, all patent disputes are litigated in the Federal District Court system. Moreover, all the appeals of patent cases are handled by one appeals court, the Federal Circuit, rather than the geographically defined circuits that correspond to all other federal courts. Patent litigation is generally considered to be a form of complex litigation. And it is less likely to be tried by a jury than other civil cases. But perhaps most importantly, the issue of how the nature of the parties particularly their financial size and bargaining power relates to outcomes is of particular philosophical 7 See, among many others, Bebchuk [1], Nalebuff [14], Png [15], Priest and Klein [16], Reinganum and Wilder [17], Shavell [19], and Spier [20]. An excellent summary of the early literature can be found in Cooter and Rubinfeld [2]. More recent work, particularly that involving multiple parties, is summarized in Daugherty and Reinganum.[3]. 8 See [12] pp Kesslar and Rubinfeld.[?], p See Kessler and Rubinfeld.[12] p. 34

5 Ball and Kesan, Preliminary Draft: Do not Quote, Cite or Distribute 5 and policy importance in patent cases. Fear that small parties may not be able to defend their rights in the courts has led to a call for institutions with lower transactions costs, such as alternative dispute resulution and enforced arbitration. However, there is also a school of thought which fears that the role of small parties may be less benign. The publicity accorded to cases like the NTP v. RIM Blackberry case 11 has created concern about the existence of patent trolls, small firms, inventors and the licensing firms they hire, who are unable or unwilling to develop their inventions but use them to extort large firms into payments to avoid litigation.footnotefor a discussion of the debate on patent hold ups and trolls, see Shapiro.[18] Despite the contradictory views of small inventors as victims or vicimizers in the courts, there is little research on how the nature of the parties influence cases. An exception is the work of Lanjouw and Schankerman,[13] who explore whether small parties are at a disadvantage in patent cases. They were unable to come to any statistically significant conclusion on this issue. 12 To some extent, their inability to find any concrete results on this issue may result from their exclusive focus on trials and exclusion of pre-trial rulings. But there is also a fundamental problem in most of the empirical work on patent litigation: the major data sources are centered around the patents and not around cases. Most work in this area tries to combine the various data sources identifying patent cases 13 with data about the patents and patent holders available from the exhaustive NBER database. 14 Above and beyond any concerns about the age of the financial information in the NBER statistics 15 this information can only describe the plaintiff in the case. But the choice to accept a settlement as opposed to continuing litigation is the result of, among other factors, the relative financial resources and bargaining power of both the plaintiff and the defendant.without information on nature of both parties in the litigation, it is impossible to understand how the size of the party and their relative bargaining power influence case outcomes, let along whether the patent system and the courts favor small parties, large parties or no one. This work attempts to fill this gap. We start from a previous study which describes the outcomes of patent cases including settlements, pre-trial rulings, and trials. 16 In Section 1 of this paper we summarize the results of that work, describing the outcome of patent infringement cases filed in three recent years: 1995, 1997 and Then, in Section 2 we report detailed information on the parties to the disputes filed in 2000 and the technologies of the patents under dispute. In Sections 3 and 4 we present results on how the nature of these 11 a search of the Lexis/Nexis business press database found almost 400 articles discussing the case. 12 Though they were more successful in examining how the nature of the parties influence the propensity of patents to be involved in cases. 13 Derwant Litelert and the database created by the Administrative Office of the District Courts, to be described later. 14 See Hall, [8] etc. 15 which is drawn from Compustat data from See Kesan and Ball.[10]

6 1 HOW ARE PATENT DISPUTES RESOLVED: THE 1995, 1997 AND 2000 COHORTS6 parties, in terms of relative bargaining power, can help explain the resolution of patent cases. 1 How are Patent Disputes Resolved: The 1995, 1997 and 2000 Cohorts Previous studies of patent litigation have suffered from a major deficiency in the data: while the concept of a settlement is clear conceptually, it is not an official legal term. Two parties may reach a mutually acceptable agreement, or settle, their dispute, but the resolution of the case will be legally recorded in some other category. 17 As a consequence, the official database produced by the Administrative Office of the District Courts 18 recording the outcome of U.S. civil cases does not give a complete picture of how many cases are resolved though a judgment 19 and how many are resolved through a settlement between the two parties. To counter this problem, we relied on the official record of case progression and resolution, the docket report. This document, which is now available on line 20, provides a record of all hearing, motions, and documents filed in the case as well as a description of the judge, parties, and lawyers involved. It also describes the manner in which the case was officially resolved. Using the list of patent disputes provided by the Administrative Office of the District Courts (AO) database, we examined the dockets for patent cases filed in three recent years: 1995, 1997, and We then classified the cases according to the way they were resolved. The results are given in Table INSERT TABLE 1.A HERE INSERT TABLE 1.B HERE Much of the detail in Table 1 is of interest primarily to legal scholars. However, in general terms, most of the items in the first column 22 represent cases in which either the plaintiff 17 Such as a dismissal, a consent judgment, etc. 18 And available through ICPSR. 19 Strictly speaking, a judgment on the merits of the case. 20 through the courts PACER system, at pacer.psc.uscourts.gov. 21 Further detail on the methodology used to classify cases and additional results can be found in Kesan and Ball.[10] 22 The exception being cases terminated on a ruling of lack of jurisdiction, which pertains to the legal

7 1 HOW ARE PATENT DISPUTES RESOLVED: THE 1995, 1997 AND 2000 COHORTS7 chose not to pursue the case 23 or the defendant never responded to the plaintiffs complaint. When these conditions are true, the case may have just disappeared for procedural reasons, or it may have been unofficially settled. The second column represents cases that are officially designated in the docket as having settled, or the docket declares them resolved through a mechanism which generally indicates a settlement. The final column indicates cases which were resolved through a final ruling. Overall, about 4.3% of cases are resolved through a trial; 24 this figure is largely consistent with previous estimates of the number of cases ending in trials. However, almost twice as many cases cases were decided through a pretrial ruling, either on a defendant s motion to dismiss or a motion for a summary judgment for the plaintiff 25. Thus, rulings occurring before a trial must also be included in the analysis. For the remainder of the paper, we will be focusing solely on the cases filed in the year For comparison, the outcome of patent cases filed in that year are included in Table 1.b. As can be seen, there is little difference between that single year and the other years included in Table 1. Thus, the third column of Tables 1 and 1.b represent the cases resolved with a final determination by the court that the patent was valid or invalid and/or that it was infringed. However, there is an important distinction between the pre-trial (dismissals with prejudice and, more importantly, summary judgments) and the trial rulings. Filing and arguing a motion for summary judgment is an expensive procedure; a significant amount of time must be dedicated by the legal team to writing and filing briefs and motions and arguing the position in hearings. Nonetheless, the expense of going to trial is much greater. In general, getting any ruling is expensive, but getting a ruling through a trial is extremely expensive. However, these results do not explain why the patent holders and alleged infringers settle some cases and pursue a ruling in others. The remainder of this paper will explore one hypothesis explaining case outcomes: the resolution of the case depends on the relative financial and legal resources of the two parties. This hypothesis has considerable policy implications; it would imply that the transactions costs associated with patent cases, and their differential impacts on the parties with different levels of resources, explain whether or not there is a ruling in the case. In other words, it would support the conclusion that the operation of the courts and the economics of litigation might lead to socially non-optimal solutions in patent cases. rules of case prosecution. 23 Want of Prosecution or Voluntary Dismissal. 24 which may be a jury trial or a bench trial in which a judge makes the final ruling. Rulings as a matter of law indicate cases where a jury renders a verdict which is adjusted by the judge if it is not legally accurate. 25 Legally, a summary judgment is granted if the parties are in agreement about the facts of a case and one party persuades the judge that the law is in their favor.

8 2 THE NATURE OF THE PARTIES AND THE RESOLUTION IN PATENT INFRINGEMENT CASES8 2 The Nature of the Parties and the Resolution in Patent Infringement Cases While there is a rich and extensive collection of data on patents and patent holders 26, detailed information on the parties in patent cases is much more rare. Obtaining such information is particularly difficult since patent cases may have multiple plaintiffs and defendants. Unfortunately, once again, we must use the dockets as our source on the parties to the case. The data provided by the AO lists only the first plaintiff and first defendant, and the first party listed is often decided solely on alphabetical order. There may be multiple plaintiffs for a number of reasons relating to the issue of who has standing in the case. Only the actual patent holder has the right to sue for infringement. However, a licensee may want to initiate the suit. Therefore the inventor or assignee of the patent must also be listed as a plaintiff. The AO data typically lists only the first plaintiff and defendant listed on the docket for each case. Fortunately, this situation creates relatively few analytical problems in the analysis. All the parties should be making decisions in concert about whether to settle the dispute at a given moment, etc. Thus, all plaintiffs can be consolidated as a single plaintiff, using the characteristics of the firm with the biggest pockets as covariates describing the nature of the plaintiff. The situation with multiple defendants is more problematic. A patent holder may be unclear as to the exact name of the firm and sue multiple variations as an insurance mechanism. He may sue the subsidiary, or multiple subsidiaries, of a firm whom he believes has actually infringed, and the parent company as well. He may sue the U.S. and international subsidiaries. He may sue the firm and the owner of the firm. He may sue the firm and certain employees of the firm. He may sue the firm manufacturing the product and the firm who sells it. In a patent infringement suits, none of these defendants is officially the lead defendant in the case. In purely legal terms, all have the power to independently decide whether to pursue a judgement or settle the case. Yet, it is obvious that not all defendants are independent economic decision makers. To weed out these extraneous parties, we devised a set of rules for consolidating such dependent parties into those who are the actual decision makers. Thus, our analysis of the impact of the relative financial resources and bargain of the parties began with compiling a list of all the parties in cases filed in We also recorded how each defendant s participation in the case was resolved i.e., was there a settlement agreement between that defendant and the plaintiff, or a final ruling for or against that defendant. We then used a variety of data sources to construct a profile of the characteristics of each plaintiff and defendant. Once these profiles were known, we could consolidate plaintiffs and eliminate defendants to compile the final list of plaintiff/defendant negotiating pairs which 26 See Hall, et al, for an explanation of the NBER database on patents.

9 2 THE NATURE OF THE PARTIES AND THE RESOLUTION IN PATENT INFRINGEMENT CASES9 actually decide the outcome of the case. 2.1 Characteristics of the Parties in Patent Cases Filed in 2000 Data was collected on every party both plaintiff and defendant in all suits filed during Information about each party was sought in the Corporate Affiliations database available though Lexis/Nexis which provides world-wide coverage of public corporations, U.S private corporations with sales in excess of $10 million and international private corporations with sales in excess of $50 million. This information was supplemented by private industry analysis, such as the Forbes lists of the the 500 Largest Private U.S. Firms and 2000 Largest Global Firms. Results for U.S. public firms and their subsidiaries were verified in the Securities and Exchange Commission s Edgar database of SEC filings. If no information could be found from any of these sources, we searched on the internet for any applicable information. If we still found no information on a firm, it was assumed that a company that had left absolutely no footprint must have been small, and the firm was classified as such. A search was also conducted for individuals listed as plaintiff and defendants to see if they had any affiliation with other parties or there was any other relevant information. Using these same data sources, we determined whether the party was an individual, firm, university, hospital or other non-profit. We determined the line of business of firms, as defined by SIC codes, NAICS codes, or descriptions, the location or state of incorporation of the firm, whether the firm was public or private, the stock exchange on which it was listed (if relevant), and whether it was an independent corporation or a subsidiary of a larger corporation. But most importantly for the analysis in this paper, we determined the firm s annual sales for the year 2000, or the latest year available. We used this information to classify firms into three categories: small firms, with sales of less than $10 million per year; 27 medium firms with sales between $10 million and $100 million; and large firms with sales over $100 million. We used this information both for purposes of combining plaintiffs according to the characteristic of the consolidated plaintiff, eliminating defendants who are not independent decision makers, and constructing multiple plaintiff/defendant pairs for each case representing the relative strength of the independent defendants against the consolidated plaintiff. Table 2 gives a breakdown of the parties according to some of these criteria Firms for which no financial information could be found in any source were also classified as small, since these firms were usually producers of a limited range of products or small retailers. We also believe that a firm which left no footprint in the paper or virtual media was unlikely to have been a major financial player, and therefore is or was likely to have been small. 28 Cases filed by the Lemelson Foundation are not included in these tallies, or in the anlaysis to follow. These 5 cases, each with between 80 and 150 large-firm defendants would have dominated the statistical results, biasing the analysis to reflect how defendants react to being sued by Lemelson rather than how they

10 2 THE NATURE OF THE PARTIES AND THE RESOLUTION IN PATENT INFRINGEMENT CASES1 INSERT TABLE 2 HERE The data suggest that, based solely on a cursory examination of the dockets, individuals and small firms are very active in patent litigation both as plaintiffs and defendants. However, most plaintiffs act as one decision-making unit, so in reality they can be consolidated as a single party. And many defendants list in the docket are not truly involved in any decision regarding the pursuit of the case. These extraneous parties must be consolidated or eliminated to truly analyze the breakdown of relative bargaining power. Our data revealed several interesting facts about different kinds of parties that have been sources of controversy. For example, universities were not highly active in litigation; only twenty-six cases had a university as a plaintiff and they defendants in only a handful of cases. Thus, our results show that universities were still not active in litigation, even as late as Moreover, they were typically accompanied by other parties. Moreover, universities were never the sole plaintiff in the case. In 17 cases they were accompanied by a large firm as a co-plaintiff; in four cases they had a medium sized firm as a co-plaintiff; in 7 cases they had a small firm as a co-plaintiff; in 1 cases another university and a small research firm were co-plaintiffs. Finally, among the most controversial firms, are the patent licensing firms. Again, our one year picture of patent litigation is insufficient to adequately describe the role of such firms in the courts. Only nineteen firms filing 49 cases were identified as the type of patent licensing firm commonly given the troll moniker. Most licensing firms sued 2 to 10 firms during the year. However, a small number took legal action against a large number of defendants; one organization filed 5 suits, each with 80 to 140 defendants. Unfortunately, the small number of firms falling into this category within a single year makes statistical analysis of their performance difficult if not impossible 2.2 Consolidating Plaintiffs and Eliminating Irrelevant Defendants: The Plaintiff/Defendant Pairs To classify the actual nature of both plaintiffs and defendants, we tried to eliminate any parties that were not true decision makers. This was relatively easy for plaintiffs. It was assumed that all the plaintiffs would be pursuing the same issue, and the plaintiff with the react to being sued by a large organization or a patent licening company.

11 2 THE NATURE OF THE PARTIES AND THE RESOLUTION IN PATENT INFRINGEMENT CASES1 deepest pockets would have the greatest influence on the prosecution of the case. Therefore, if any plaintiff could be classified as large, the case was assumed to have a large plaintiff. If the party with the greatest financial resources was medium sized, then the case had a medium plaintiff. If there was no large or medium plaintiff so the party was either a small firm or an individual the plaintiff was considered small. Classification of the defendants was somewhat more complicated. In many cases, the multiple defendants were simply variations on the firm s name used by the plaintiff in order to insure that they got it right. But in other cases, we were not consolidating parties but eliminating parties that were not independent decision makers. We started from the default position that all parties in the case were independent decision makers such as subsidiaries or the individual who owned the firm. Details on the rules employed for consolidation are given in Appendix 1. After consolidating plaintiffs and eliminating irrelevant defendants, we were able to define the multiple plaintiff/defendant pairs within each case which could potentially negotiate a settlement agreement. These are the true decision making units in patent litigation. Each pair will decide, given their particular liability in the case, ability to invest resources in the pursuit of the case, and expectations for financial gains from damages, how to pursue the case. Table 3 shows the number of defendants per case after eliminating irrelevant defendants as well as the number of plaintiff/defendant pairs. 29 From this data it is apparent that patent cases do not typically involve large numbers of effective defendants. 30 Half of all cases have only one defendant; two thirds of all cases have two defendants or less. Thus, after removing extraneous parties named in the dockets, patent litigation usually is a one-on-one conflict. Table 4 shows the final breakdown of plaintiff/defendant pairs by the size of the parties. The nine classifications of plaintiff/defendant pairs demonstrated in this table (i.e., small plaintiff, small defendant; small plaintiff, medium defendant; etc.) allow us to demonstrate the relative financial resources available to each party in the negotiating pair. Again, we observe that small parties are quite active in patent litigation, both as plaintiffs and as defendants. In fact, the largest number of cases approximately 30% of all pairs fall into the small plaintiff/small defendant pairs. This result suggests that the vast number of patent cases are resolving disputes between small firms/inventors. Nontheless, there are a substantial number of cases in the pair categories that represent radically different financial resources. Just under a third of all cases fall into the small plaintiff, large defendant; or large plaintiff, small defendant categories. These pairs may 29 Some cases were eliminated from the database because they were declaratory judgment actions or suits filed in unusual circumstances which would make them extreme outliers. 30 We should note that we dropped four cases filed by the Lemelson Foundation, each of which had approximately 150 defendants, since this was an extreme outlier.

12 2 THE NATURE OF THE PARTIES AND THE RESOLUTION IN PATENT INFRINGEMENT CASES1 be the most interesting for our analysis of the impact of the nature of the parties on patent litigation. However, it is also worth noting that Table 4 suggests that any concern of small firms holding up large firms through litigation is unwarranted. There are a significant number (about 15%) of cases in the small plaintiff/large defendant category. But this statistic, in and of itself, does not seem sufficiently large to suggest a source of problems. INSERT TABLE 3 HERE INSERT TABLE 4 HERE 2.3 Patent Technologies While the focus of this study is on the parties particularly small parties in patent cases, information on patents is essential to the analysis. Without information on the patents and the technologies they embody, our results might be biased by correlation between party size and technology. To avoid this problem, we collected data on the patents at issue in our cohort of cases from two sources: Derwant LitAlert and information on the patents at issue in the case dockets and accompanying documents. Both these sources were equally important. LitAlert draws its data primarily from information about actions regarding a patent or trademark which courts are required to report to the U.S. Patent and Trademark Office (PTO). However, despite the legal requirement, reporting is incomplete and there is essentially no enforcement. 31 At the same time, there is no requirement that the patent be listed in the case dockets. Patents are listed in the complaint filed by the plaintiff, and as more districts link to the complaints on line data on litigated patents should become more complete. We obtained the numbers of the patents for 1086 cases, about 55% of the total, from LitAlert, so the data was obtained nearly equally from the two sources. Table 5 lists the data we were able to obtain on litigated patents. We have employed the technology categories developed in conjunction with the National Bureau of Economic Research (NBER) patent database, which condense the intricately detailed PTO classification 31 This situation is particularly problematic since some districts do almost no reporting while others are highly compliant.

13 3 HOW DOES THE NATURE OF THE PARTIES INFLUENCE CASE OUTCOMES 13 system into six categories. 32 We have also added a category zero for design patents. Table 5 reports the breakdown across these technological categories, both overall and across the size of the consolidated plaintiff. As can be seen, there does not seem to be any overwhelming trend across the size categories. Approximately one third of the patents are missing for all categories, though a slightly higher percentage is available for large plaintiffs. There is also very little variation in the percentage of each technological category, although there is a somewhat lower percentage of large firms defending patents which fall in the catch-all other category. 33 Thus, there appears to be little concern that strong correlations between party size and technology will affect our results. However, as demonstrated in Table 6, there is much wider variation when technologies are broken down by our unit of analysis: the nature of plaintiff/defendant pairs. For example, a high percentage of the pairs where a small plaintiff is suing a large defendant fall in the computers and communications category. Thus, controlling for technology will be necessary in the statistical analysis. INSERT TABLE 5 HERE INSERT TABLE 6 HERE 3 How does the Nature of the Parties Influence Case Outcomes The data described in the preceding section forms the basis for analyzing how the resolution of patent infringement cases is related to the relative size of the parties. We conducted this analysis using two measures. First, we analyzed whether or not there was a final ruling on the 32 It is common for there to be more than one patent in a case, but there is seldom any difference in the technology of said patents at the level of the NBER classification. When there were both design patents and utility patents at issue in the same case, we used the utility patent. 33 This probably reflects the higher level of activity of small firms in manufacturing items like sporting goods, clothing, novelties, etc., which are grouped in this category.

14 3 HOW DOES THE NATURE OF THE PARTIES INFLUENCE CASE OUTCOMES 14 merits of the case. 34 As described above, it is a generally accepted principle, that judgments in patent cases are expensive. Trials involve a considerable expenditure of resources; even pre-trial judgments have apparently become quite expensive. 35 Thus, parties may choose to settle the case to avoid these considerable transactions costs. If there is a difference between the rate at which small parties pursue judgments after controlling for technology, that fact would indicate that entrepreneurs and individual inventors are at a disadvantage in the protection of intellectual property rights. 36 The results of our estimation are summarized in Tables 7.a, 7.b, and 7.c. The full estimation results are given in Appendix Table 1. Since we could be sure that all the observations terminated through so-called non-merit resolutions the first column in Table 1.b were procedural or settlements, we estimated three models. Model 1 employs all the cases except those which terminated through a dismissal for lack of jurisdiction. 37 Model 1 was estimated using a logistic regression across plaintiff/defendant size categories while controlling for the technology of the patent as measured by the NBER categories. We found that all the technology categories with the exception of NBER category 1 (Chemical) were significant in explaining whether the case terminated in a judgement. Among the plaintiff/defendant pairs, small plaintiff/large defendant, medium plaintiff/small defendant and large plaintiff/small defendant were all significant. Table 7.a summarizes these results by reporting the probability of the case terminating in a judgment for all significant pairs across all significant technology categories. The table also reports the estimated probability of termination through a judgment for all remaining categories. 38 Interestingly the small plaintiff/large defendant pair is the most likely to litigate it s dispute to a judgment; the probability that such a pair will seek a judgment is twice that of the remaining, non-significant categories and four times that of the large plaintiff/small defendant pair. This result suggest that the high transaction costs associated with litigation do not appear to deter small inventors from vigorously protecting their property rights in patent court. Tables 7.b and 7.c summarize the results of two additional models, each discarding another group of cases which may have determined through neither a settlement nor a judgment. Model 2 (Table 7.b) drops cases dismissed for lack of jurisdcation and those terminated through a default judgment. 39 Table 7.c drops all non-merit resolutions. 34 That is, whether or not the final resolution for the pair would have fallen in the third column in Table See Kesan and Ball, p. xxx 36 See Lanjouw and Schankerman 37 I.e., the judge ruled that the case was filed in the wrong geographical jurisdiction or that the plaintiff did not actually have standing with respect to the patent in dispute. 38 That is, those which have a value of zero for all the significant pairs. 39 I.e., cases where the defendant never responded to the complaint and a judgment was awarded on default.

15 4 HOW DOES ACCESS TO RESOURCES INFLUENCE THE INVESTMENT OF THE PARTIES IN TH In general, the results of these two models do not differ substantially from those of MOdel 1: the small plaintiff/large defendant pair is still the most likely to litigate to a judgment. It appears that the size of the potential surplus gained from obtaining a judgment is more important than the cost of litigation in explaining the behavior of parties. Small plaintiffs suing large defendants could potentially earn a large award. Large plaintiffs suing small defendants, however, are unlikely to earn sufficient damages to make pursuit of a judgment worthwhile. And, other institutional arrangements, like the availability of contingency fee lawyers, allow small plaintiffs to overcome the transaction costs in the case. INSERT TABLE 7.A HERE INSERT TABLE 7.B HERE INSERT TABLE 7.C HERE 4 How Does Access to Resources Influence the Investment of the Parties in the Case A second measure of the relationship between the bargaining power of the parties and case outcome is the level of (transaction) costs the parties are willing to expend in pursuing the case. The willingness of different defendant/plaintiff pairs to settle the case versus seeking a judgment on the merits which we explored previously was the result of the benefit/cost calculations of the bargaining pairs. In interpreting these results, we argued that judgments were expensive, so the gains from a judgment should explain why so much was spent on the case. Here, we examine only the level of expenditure on the case. Thus, while the results in Table 5 suggest that large plaintiffs are more willing to settle if they are suing a small defendant because they have little to gain from a judgment and want to avoid costs, and the small defendant would also like to avoid costs. A hypothesis related to expenditure would

16 4 HOW DOES ACCESS TO RESOURCES INFLUENCE THE INVESTMENT OF THE PARTIES IN TH suggest that among large plaintiff/small defendant pairs the case would be resolved quickly, with low expenditure. However, since small defendants, and their contingency fee lawyers, are more likely to seek a decision on the merits, we might also expect that defendant/plaintiff profile to be willing to expend more on the prosecution of the case. To analyze the willingness of different types of defendant/palintiff pairs parties to invest in the case, the progress of the case was examined to see in which of three stages it ended: at a low level of expenditure (before the filing of summary judgment motions); at an high level of expenditure (after the filing of such motions but without going to trial); or at a very high expenditure (resolved through a trial). It should be noted that this ranking does not directly reflect the actual outcome of the case. Cases which terminate before filing of summary judgment motions are nearly always settlements. Cases which go to trial are always decided through rulings. 40 However, after the filing of summary judgment motions the case may be terminate through a ruling on that motion or it may settle without a ruling. But if a case does settle during that phase, it is a more expensive settlement than if it had settled in the first phase. INSERT FIGURE 1 HERE Table 8 gives the breakdown of the level of expenditure by low level of expenditure (resolved before the filing of a motion for summary judgment), high level of expenditure (a summary judgment motion filed, but resolved before a trial), and very high expenditure (case resolved through a trial). In Table 8.a the level of expenditure is given for each possible combination of plaintiff and defendant. The top row of the table represents small plaintiffs; the second row represents middle sized plaintiffs, and the third row represents large plaintiffs. The first column of the table represents small defendants; the second column represents middle sized defendants, etc. Thus, Table 8.a can be read as a three-by-three matrix, with the top left-hand corner giving the breakdown of the level of investment for the small plaintiff/small defendant pair and the bottom right hand corner breaking down the level of investment for the large plaintiff/large defendant pair. Table 8.b gives the breakdown of level of investment for the total data set for comparison. These results demonstrate an apparent pattern in the level of expenditure in patent infringement cases depending on the nature of the plaintiff/defendant pairs. Moving across the first row that is, comparing the investment levels for small plaintiffs against small, medium 40 In the 2000 cohort, once a trial had started it always proceeded to a final ruling; there were no cases which settled during the course of a trial.

17 4 HOW DOES ACCESS TO RESOURCES INFLUENCE THE INVESTMENT OF THE PARTIES IN TH and large defendants there seems to be a decline in the number of cases resolved with a low level of expenditure. 41 In a similar manner, moving down the the first column and thereby comparing cases with small, medium and large plaintiffs and small defendants there is an increase in the number of cases resolved with low levels of expenditure. Similarly, we can examine the levels of investment moving across the bottom row of Table 8.a, comparing cases with large plaintiffs and the three sizes of defendants. Once again, a higher proportion of cases are with a small defendant are resolved with low expenditure than with a large defendant. Moving down the final column, we can compare cases with the three sizes of plaintiffs and large defendants. Here the proportion of cases resolved with little investment is higher when a small plaintiff faces a large defendant than when a large plaintiff faces a large defendant. Thus, the two squares of the matrix represented in Table 8.a which represent the maximal difference in the resources and bargaining power give a strikingly different picture. In the lower left square (large plaintiff/small defendant) there is a relatively high probability that the case will be resolved with low expenditure. The large plaintiff has little to gain by pursuing the case to a judgment; a small defendant will be able to pay damages that are insignificant to a major firm. However, when a small defendant sues a large plaintiff (upper right corner) the story is quite different. The small plaintiff may be able to win large damages from the large defendant, a fact that will be attractive to attorneys working on a contingency basis. Meanwhile, the large firm can afford to drag out the case through summary judgment motions, etc. 42 Thus, Table 8.a demonstrates two facts. The relative level of resources available to the parties in a patent infringement case does explain the way the case is resolved. And this effect is asymmetric; it makes a difference if the firm with greater resources is the plaintiff or the defendant. 41 A cursory calculation found that the was percentage of cases resolved with a large defendant was less than that with a small defendant at a 5% level of significance. 42 And the large defendant may wish to do so to establish reputation effects against future potential litigants.

18 4 HOW DOES ACCESS TO RESOURCES INFLUENCE THE INVESTMENT OF THE PARTIES IN TH The previous section explored the the willingness of parties to overcome transactions costs and pursue a judgment in a patent cases. In formulating that analysis, we argued that the willingness to pursue a judgment was reflective of the willingness to expend transaction costs because judgments are, on the whole, expensive. However, it is possible for a case to settle and still involve considerable expenditure of resources. The parties may litigate a case, filing motions, pursuing discovery, and even taking the case to trial before choosing to settle. Thus, the procedural steps taken in a case, no matter how it is resolved, may be a better indicator of costs than termination in a judgment. To explore this approach to measuring transactions costs, we tracked the occurrence of several benchmarks in the history of each case in our cohort. These procedural indicators are described in Figure 1. Cases begin with a complaint being filed by the plaintiff and served on the defendant. However, litigation begins in earnest when the parties begin filing motions for summary judgment. 43 At this stage costs begin to rise, as the parties conduct discovery, collect dispositions, consult experts and file motions. The rate at which costs accrue further escalates if the case goes to trial. Thus, a case passes through three procedural stages: the low cost pre summary judgment phase; the high cost summary judgment motions filed stage; and the very high cost trial stage. Even if the case is not terminated with a judgment, it may settle in any one of these stages. Since not only costs but the rate at which costs accrue accelerates with each stage, we were able to use an ordered logit model to analyze the relationship between termination in a given stage and the relative size of the parties involved. Once again, we controll for technology and estimated three models to account for the ambiguity in determining whether some outcomes were merely procedural terminations or settlements. The results of this estimation is summarized in Tables 8.a, 8.b, and 8.c. Full estimation results are given in Appendix Table 2. Table 8.a summarizes the results for Model 1, as described in the previous section. In the ordered logit analysis, technology did not seem to have as great an impact; only NBER category 3 (Drugs and Medical) was significant. However, the ordered logit results in many ways support the results of the logistic regressions. Once again, identification as a small plaintiff/large defendant pair is significant. Moreover, such pairs are more likely to pursue the case to a high cost procedural stage than are other pairs in general and large plaintiff/small defendant pairs in particular. Nonetheless, these results are not as strong as those found in the analysis of the propensity to pursue a judgment. In fact, the small plaintiff/large defendants pairs are no longer significant in Model 3. However, Table 8 demonstrates that large firms suing other large firms are as likely to pursue the case to a high cost stage as are small plaintiffs suing large defendants. And these are some of the most statistically significant results in the estimation. To some extent, 43 I.e., motions for a pre-trial ruling on some or all of the issues in dispute.

19 5 CONCLUSION 19 this result may reflect the financial strength of the two parties as they try to improve their bargain position before agreeing to a settlement. However, large firms usually have multiple overlapping technologies and patent disputes and can bargain across several at once without resorting to litigation. We suspect that the longer duration cases may involve other factors, such as asymmetric stakes. We believe that studying these cases further may be a fruitful area for further research. Finally, we regret that it was not possible to incorporate a fuller treatment of patent licensing firms into the formal statistical analysis. The highly skewed distribution of cases among the small number of licensing firms observed in a single year would have made our results difficult to interpret; were we measuring the impact of the plaintiff being a licensing firm, or the impact of a particular licensing firm on the willingness to pursue the case. In future research, with additional years of data, we may be able to explore this issue more thoroughly. INSERT TABLE 8.A INSERT TABLE 8.B INSERT TABLE 8.C 5 Conclusion The role of individual inventors, small firms and entrepreneurs in the patent courts has become controversial for two, somewhat contradictory reasons. First, there is the view that small parties may be at a serious disadvantage in defending their intellectual property rights in the courts since they do not have the financial resources to overcome the transactions costs of litigation. However, there is also a fear that some small inventors and licensing firms

20 5 CONCLUSION 20 may be using the courts as a mechanism to extract economic rents from large innovative companies. Using detailed data on the parties, case outcomes, and disputed patents of cases filed in a single year (2000), we attempted to explore these issues. While we were not able to completely explore how licensing firms may be behaving differently than other firms, we did find that very few such firms were active in the courts in 2000, though the distribution of the number of parties they sued was highly skewed. We intend to explore this issue further with more data. However, we did find that the fear inventors do seem to be able to aggressively protect their property rights in the courts. When faced with a large alleged infringer and the potential for a significant damage award, small parties are able to overcome transactions costs and are the most likely to litigate to a judgment or to a high cost procedural stage of litigation. On the other hand, large firms suing small alleged infringers are less likely to aggressively pursue the case, preferring to settle rather than expend resources when there is little chance of a high damage award.