Authors: Banks Miller and Brett Curry
Publication: Journal of Law, Economics, & Organization, published online 2026
Original source: jleo_ideology_judicial_discretion_2026.pdf
Conversion note: this Markdown was generated by text extraction from the hosted file. Review equations, tables, notes, and citations before treating it as a final web copy.
The Journal of Law, Economics, and Organization, 2026, 1–25 https://doi.org/10.1093/jleo/ewag028 Article Ideology and the scope of judicial discretion: evidence from the federal district courts Banks Miller1 and Brett Curry2,* 1University of Texas at Dallas 2Georgia Southern University *Corresponding author: Department of Political Science and International Studies, Georgia Southern University, P.O. Box 8101, Statesboro, GA 30460-8101, USA. Email: bcurry@georgiasouthern.edu. This study examines how changes in judicial discretion influence ideological decision-making in the federal courts. Leveraging a pair of unanimous Supreme Court decisions that expanded district judges’ discretion to award attorney’s fees in patent litigation, we investigate whether this procedural reform led to substantive shifts in outcomes. Our differences-in-differences design reveals that conservative judges became significantly more likely to award fees after 2014 while more liberal ones did not. We observe no similar pattern in copyright cases, which experienced no comparable expansion of discretion. This patent-specific effect is driven by accused infringers seeking fees after defeating patent claims, consistent with conservative preferences for deterring frivolous litigation. Even unanimous, procedurally focused appellate rulings can lead to asymmetric ideological effects. These findings have implications for understanding how changes in discretionary authority— including the Court’s recent abandonment of Chevron deference—may systematically empower certain ideological preferences in the federal judiciary. (JEL C21, H11, K4, O34) Introduction Judicial discretion is a fundamental, although hardly immutable, feature of the American legal system. While Congress possesses legislative authority to adjust the boundaries of that discretion (e.g., Stith and Cabranes 1998; Johnson, Nelson, and Pritchard 2007; Ryan 2020), in many ways, the Supreme Court is the primary architect of its scope in the federal system. Indeed, in the last several decades, the Court has both contracted and expanded © The Author(s) 2026. Published by Oxford University Press on behalf of Yale University. All rights reserved. For commercial re-use, please contact reprints@oup.com for reprints and translation rights for reprints. All other permissions can be obtained through our RightsLink service via the Permissions link on the article page on our site—for further information please contact journals.permissions@oup.com.
2 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 the discretion of lower court judges in numerous legal domains. Examples of the former include decisions circumscribing the ability of bankruptcy judges to utilize certain equitable powers (see Law v. Siegel 2014) and ones making it harder for district judges to deny motions to dismiss (e.g., Bell Atlantic v. Twombly 2007; Ashcroft v. Iqbal 2009). On the other hand, the Court’s rendering of the US sentencing guidelines as advisory (U.S. v. Booker 2005) is widely viewed as a catalyst for increased district court discretion. And its recent abandonment of Chevron deference (Loper Bright Enterprises v. Raimondo 2024) represents perhaps the most significant readjustment of deference in decades. Such judicial recalibrations of discretion raise important questions about its effects; perhaps most prominently, to what extent does expanded discretion serve as a neutral mechanism allowing judges to tailor decisions to case-specific circumstances, or how might it selectively empower certain ideological preferences—perhaps those preferences previously constrained by former legal rules—and operate asymmetrically? The Court’s recent abandonment of Chevron deference has brought renewed attention to this question. Observers typically mention Loper Bright’s 6-3 ideological split, anticipating that expanded judicial discretion will systematically advantage conservative policy preferences in administrative law. To understand the substantive consequences of judicially induced adjustments to discretion, we require a case where discretionary expansion occurred through a more anodyne, procedurally focused decision. The Supreme Court’s unanimous 2014 decisions in Octane Fitness, LLC v. Icon Health & Fitness and Highmark Inc. v. Allcare Health Management System, Inc., which expanded the discretion of district court judges to award attorney’s fees in patent cases, afford the opportunity to examine these questions. The Court framed these decisions solely in terms of institutional expertise and flexibility, with no apparent ideological divisions—making them an ideal least likely case: if discretion-expanding reforms produce asymmetric ideological effects in this context, it seems likely that such effects are a more general consequence of expanding discretion rather than artifacts of more overtly ideological Supreme Court decisions. Leveraging this intervention as a natural experiment, we demonstrate that these unanimous rulings on a seemingly technical procedural reform led to substantive shifts in judicial policy: conservative district judges grew much more likely to award attorney’s fees after the Court’s recalibration of the legal standard, while more liberal judges did not. Utilizing a differences-in-differences research design, we find confirmation of the domainspecific nature of this effect by analyzing copyright fee awards—an area theoretically unaffected by Octane Fitness and Highmark—with similar fee-shifting provisions where we observe no ideologically tinged changes in judicial behavior. Further, we find that this effect is driven by accused infringers (defendants) being more successful in obtaining fee awards. The remainder of our paper proceeds as follows. In the “Discretion as a strategic institutional choice” section, we consider discretion as a strategic institutional choice by examining leading studies of legislative delegation to the bureaucracy employing a transaction cost approach. We then provide necessary background on fee shifting in patent litigation and explain the legal significance of the Supreme Court’s 2014 decisions in Octane Fitness and Highmark. Next, we consider changes to lower court discretion from the perspective of district court judges and examine how such changes may relate to the influence of ideology as a factor in judicial decision-making. After introducing our empirical strategy, along with our data and hypotheses, we present our results. Following a discussion of our findings, we offer concluding thoughts on the research’s broader significance for judicial politics and its potential lessons for future changes in the discretion afforded to judges on the lower federal courts.
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 3 Discretion as a strategic institutional choice According to Huber and Shipan (2002), allocations of discretionary authority by government decision-makers are not random; instead, they reflect strategic calculations about preference alignment, uncertainty, and institutional capacity. Although their framework is rooted in legislative delegations to the bureaucracy, its broader logic fits with the hierarchical nature of the federal judiciary. Just as legislatures must decide how much flexibility to grant bureaucratic agents (see Epstein and O’Halloran 1999), judicial principals (the Supreme Court and, as applicable, the circuit courts) similarly calibrate the boundaries of discretion within which district court judges operate. More specifically, this “transaction cost approach” to discretion holds that principals consider factors such as preference divergence, informational asymmetry, capacity constraints, and context (Epstein and O’Halloran 1999; Huber and Shipan 2002)—elements that map effectively onto judicial hierarchies. First, consider the divergence of preferences between principals and agents as well as the informational asymmetries that inherently exist between them. Huber and Shipan (2002) note that principals will be more likely to limit discretion when agents have different priorities. Just as Epstein and O’Halloran (1999) conceptualize legislative delegation as a series of choices between detailed policymaking by Congress (“internal production”) and delegation, appellate courts face similar situations when seeking to establish interpretive frameworks for lower courts. Lax’s (2012) work—especially the distinction it draws between bright line rules and standards—is an important illustration of this idea, and we discuss it more below. This framework helps explain discretionary dynamics in particular legal contexts, including intellectual property litigation, where federal courts have shaped feeshifting standards. Fee shifting and the “American Rule” The so-called “American Rule”—where each party bears responsibility for its own attorney’s fees regardless of the case outcome—is the default in US courts, but there are exceptions to it. These exceptions, known as fee shifting, allow courts to transfer the burden of legal fees from one party to another (see Singer 2010). They have been likened to a “tax” against non-prevailing parties, the main purposes of which are “to discourage frivolous litigation and to restore a prevailing litigant to their pre-dispute position as best as possible” (Downing 2021, 34). Federal law authorizes fee shifting in several places, with over 150 statutes discussing the practice in some form (e.g., Percival and Miller 1984). Courts have also created a handful of equitable exceptions to the American Rule since the turn of the twentieth century (see Root 2004, 585) but, in the main, the Rule has been consistently followed for more than 200 years (Singer 2010, 697). One important statutory exception to the American Rule appears in federal patent law under 35 U.S.C. §285, which allows courts to award reasonable attorney’s fees to prevailing parties in “exceptional cases.” Fee shifting provisions are also embedded in other areas of intellectual property, including copyright law (17 U.S.C. §505) and in a provision of the Lanham Act which governs trademark litigation (15 U.S.C. §1117(A)). The trademark provision uses the identical “exceptional cases” language to the patent statute, and some courts have imported the Supreme Court’s 2014 reformulation of the “exceptional cases” standard in the patent context to trademark issues (see Fair Wind Sailing, Inc. v. Dempster (3rd Circuit 2014)). In contrast, copyright’s fee-shifting provision utilizes different language and has not been subject to this potential spillover which, as we describe below,
4 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 makes it suitable for comparative analysis with patents. With this in mind, we provide an overview of fee shifting’s evolution as it relates to patent litigation and § 285. As Downing (2021, 40) observes of § 285, “Congress drafted the statute requiring that fees not be issued automatically but when a litigant’s behavior is somehow unique—more exactly, levied when said behavior is uniquely bad or incompetent.” But Congress did not precisely articulate the factors judges should analyze in making such determinations, leaving the interpretive work to the lower federal courts. Section 285 is symmetric—both prevailing patent holders and accused infringers may seek fees. However, when accused infringers prevail, the threat of fees has been deemed “the only deterrent to the … bringing of clearly unwarranted suits on invalid or unenforceable patents” (e.g., Mathis v. Spears 1988, 754).1 Prior to the Supreme Court’s revision of the standard in Octane Fitness and Highmark, what constituted an “exceptional case” was governed by the Federal Circuit’s decision in Brooks Furniture Manufacturing, Inc. v. Dutailier International Inc. (2005), which set an extraordinarily high threshold for litigants seeking attorney’s fees under § 285.2 Under this restrictive, two-pronged test, a case could only be considered exceptional for feeshifting purposes if (1) there was “material inappropriate conduct” during the litigation or (2) the litigation was brought in subjective bad faith and was objectively baseless (Jones 2015, 514). Under this interpretive gloss, awards of attorney’s fees under § 285 were relatively uncommon (Jiam 2015, 619). In 2014, the Supreme Court rejected that understanding. In Octane Fitness, Justice Sotomayor called the Brooks Furniture framework “unduly rigid” and noted that it “impermissibly encumbers the statutory grant of discretion to district courts.” Rather than exhibiting fidelity to the plain meaning of the statute, the Court criticized the Federal Circuit for “superimpose[ing] an inflexible framework onto statutory text that is inherently flexible” (Octane Fitness 2014, 553, 555). Both Octane Fitness and its companion Highmark, which articulated a more deferential standard of review appellate courts must utilize in reviewing fee determinations, contain numerous specific references to the importance of enhancing district court discretion: “The district court ‘is better positioned’ to decide whether a case is exceptional[] because it lives with the case over a prolonged period of time” (Highmark 2014, 564); “District courts may determine whether a case is ‘exceptional’ in the case-by-case exercise of their discretion, considering the totality of the circumstances” (Octane 2014, 554); and “[N]othing justifies such a high standard of proof. Section 285 demands a simple discretionary inquiry; it imposes no specific evidentiary burden, much less such a high one” (Octane 2014, 557). Reflecting this broader district court discretion, scattered evidence suggests district courts were granting fee awards more frequently after the Court’s decisions (e.g., Jiam 2015, 623; Jones 2015, 527; Downing 2021, 47). Octane Fitness and Highmark are ideal cases for examining the ideological consequences of expanded discretion. First, they were decided simultaneously in April 2014, providing a clear treatment date for causal identification. Second, the decisions represent a dramatic shift in legal standards, moving from the Federal Circuit’s “unduly rigid” two-pronged test 1 In the paper, we use “defendants” and “accused infringers” interchangeably to refer to parties alleged to have infringed a patent, and “plaintiffs” and “patent holders” to reference parties who assert patent rights. There are some circumstances (e.g., declaratory judgment actions) where the accused infringer is technically the procedural plaintiff. But, for our substantive purposes, all accused infringers are substantive defendants. 2 Since its creation in 1982, the Court of Appeals for the Federal Circuit (“Federal Circuit”) has possessed exclusive nationwide appellate jurisdiction of cases involving certain subject matter, most notably patents, with its decisions being subject to Supreme Court review (see Baum 1990, 2010).
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 5 to a more flexible “totality of the circumstances” approach. Third, copyright cases offer a natural comparison group—they have similar fee-shifting provisions but were unaffected by Octane-Highmark—allowing us to distinguish the reform’s impact from other, broader trends in intellectual property litigation. Fourth, their unanimous nature and procedural framing make Octane-Highmark a “least likely” case for observing ideological effects; finding such effects here would strengthen the expectation that they would occur more broadly. Finally, patent fee-shifting is a substantively important issue on its own merits, especially given concerns about non-practicing entities and frivolous litigation in the IP context. In combination, these characteristics make the cases a well-suited intervention for investigating the role of discretion in ideological decision-making. One interpretation of the Federal Circuit’s restrictive approach in Brooks Furniture is that it sought to limit district judges’ discretion out of concern those judges would otherwise pursue a range of conflicting approaches to awarding attorney’s fees. By this same logic, the Supreme Court’s 2014 rejection of the Brooks Furniture framework represented a different calculation about the costs and benefits of discretion—one that balanced policy objectives with informational advantages. Clearly, in acknowledging the fact that district courts are “better positioned” to determine whether a case is “exceptional,” the justices prioritized the informational advantage over concerns about doctrinal consistency. The Court’s “totality of the circumstances” approach (see Octane 2014, 554) represents just the sort of calculated trade-off Epstein and O’Halloran (1999) describe between decision costs and agency costs. It is worth noting that the Supreme Court’s decisions in Octane Fitness and Highmark were not ideologically motivated in the traditional sense. Both rulings were unanimous, authored by a liberal justice (Sotomayor), and framed entirely in terms of fidelity to the statutory text and the institutional competence of district courts. Nowhere in either decision did the Court telegraph an intention to prod lower courts to systematically favor conservative policy outcomes. The Court also made short work of the decisions—the Octane Fitness opinion was thirteen pages; Highmark spanned just five. Still, this does not mean the decisions lacked ideological consequences—expanding discretion can have asymmetric effects when, as here, prior rules had constrained some ideological proclivities more than others. Before Octane, the high bar for awarding fees may have disproportionately constrained conservative judges who might otherwise have wished to use fee-shifting as a tool for combatting frivolous litigation. More liberal judges, whose presumed preferences already aligned with the status quo, faced little constraint. By lowering the threshold necessary for district court judges to classify a case as “exceptional,” Octane freed more conservative judges to act on their previously constrained preferences while leaving the decision space of liberal judges largely unchanged. Simply stated, the Court need not have intended this outcome for it to occur—the asymmetry flows naturally from the interaction between the discretionary reform and pre-existing ideological divisions over litigation policy. District court judges, bounded discretion, and attorney fee awards To understand how the expansion of discretion through Octane Fitness and Highmark might manifest itself in judicial behavior, we examine Lax’s (2012) theoretical framework in greater depth, building on our earlier reference to his work. With its focus on how lower courts respond to the constraints and opportunities higher court doctrines create, Lax’s
6 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 perspective is a natural complement to the transaction cost approaches of Epstein and O’Halloran (1999) and Huber and Shipan (2002). While the latter authors concentrate on the strategic factors that shape allocations of discretion in the first place, Lax’s work shifts the focus to the ways in which judges utilize discretion and, as such, helps us understand how their behavior might be expected to change when those boundaries of discretion shift. The concept of “doctrinal intervals,” which Lax (2012) defines as the range of acceptable legal rules that lower courts can adopt within the constraints set for them by higher courts, helps explain the dynamics of judicial discretion. Mindful of the importance of hierarchical compliance, lower court judges operate within this range in order to avoid reversal even as they strategically pursue their preferred outcomes when possible. Higher courts construct doctrinal spaces with “strategic ambiguity,” to permit flexibility on the one hand while constraining extreme interpretations on the other. Such intervals can be narrowed with precise rules (e.g., Miranda v. Arizona 1966) or widened via more flexible standards such as a “totality of the circumstances” approach (Lax 2012, 765; e.g., Illinois v. Gates 1983; Strickland v. Washington 1984). Naturally, with bright line rules the bounds of district court discretion contract; conversely, with more open-ended standards such as Octane Fitness’s guidance in discerning whether a case is “exceptional,”3 the discretionary space expands. This, in turn, is likely to permit more space for ideological considerations to shape district court decisions because it allows judges on those courts more room to maneuver. In the context of attorney fee awards, greater discretion potentially affords district judges a greater ability to express their views about frivolous litigation and the appropriate consequences for it. This ideological asymmetry should manifest itself primarily when accused infringers seek attorney’s fees. In these cases—which constitute approximately two-thirds of fee requests—awarding fees serves to deter weak patent assertions, aligning with conservative policy preferences for limiting frivolous litigation. By contrast, when patent holders seek fees after prevailing, fee-shifting would incentivize patent enforcement rather than constrain it. Granting fees to prevailing patent holders does nothing to address concerns about vexatious litigation or patent trolls; indeed, it potentially encourages more patent enforcement activity. Thus, we expect conservative judges to be more responsive to expanded discretion specifically when accused infringers request fees. Finally, Lax’s theory of bounded discretion helps explain why we might observe different patterns across legal domains even if they involve similar statutory provisions. Although the Supreme Court expanded district court discretion in patent fee-shifting, interpretations of 17 U.S.C. §505 have been more consistent in providing that discretion when it comes to fee shifting in copyright cases. Section 505 contains no exceptional case requirement, and its doctrinal history is more consistent. In Fogerty v. Fantasy, Inc. (1994), the Supreme Court instructed district courts to grant copyright fee awards on an even-handed basis in a manner “faithful to the purposes of the Copyright Act” (Hyde and Sharrock 2004, 468). Although the Court clarified aspects of that decision in Kirtsaeng v. John Wiley & Sons (2016) two years after Octane Fitness, that decision differs in important respects from the patent intervention and, for reasons described below, is unlikely to confound our comparison. 3 “We hold, then, that an ‘exceptional’ case is simply one that stands out from others with respect to the substantive strength of a party’s litigating position (considering both the governing law and the facts of the case) or the unreasonable manner in which the case was litigated. District courts may determine whether a case is ‘exceptional’ in the case-by-case exercise of their discretion, considering the totality of the circumstances” (Octane Fitness 2014, 554).
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 7 Kirtsaeng essentially reaffirmed Fogerty’s longstanding framework rather than expanding discretion as Octane had done for patents. Notably, both parties in Kirtsaeng agreed that Fogerty had correctly established the appropriate framework for fee-shifting; they disagreed only on how district courts should apply that framework (Devinsky 2016). The Kirtsaeng Court emphasized that § 505 “clearly connotes discretion” and permits consideration of “several nonexclusive factors,” including “frivolousness, motivation, objective unreasonableness, and the need in particular circumstances to advance considerations of compensation and deterrence” (Kirtsaeng 2016, quoting Fogerty 510 U.S. at 534). The decision resolved a circuit split over how to weight factors within the existing Fogerty framework: one approach presumed prevailing parties would receive fees; another gave dispositive weight to objective reasonableness; and a third utilized a totality-of-the-circumstances approach that did not give pride of place to any single factor. The Court emphasized that § 505 gave the district courts “broad discretion” to award fees “even when the losing party advanced a reasonable claim or defense” based on other factors (Devinsky 2016). This represented a procedural correction to ensure the Fogerty standard was being properly applied, rather than a substantive transformation of the test itself.4 Although both Octane Fitness and Kirtsaeng were unanimous decisions, the nature of the Court’s agreement in the cases differed. Octane Fitness and Highmark were unanimous in that all justices agreed that the Federal Circuit’s restrictive framework was incorrect and needed to be jettisoned in favor of a more relaxing standard. Kirtsaeng was unanimous because the justices agreed the Fogerty framework was sound—the disagreement the Court resolved was, instead, about how lower courts should apply its multifactor balancing test (Kirtsaeng 2016; Kevinsky 2016). Put another way, in the patent cases the Court unanimously rejected an existing standard; in Kirtsaeng the justices reaffirmed Fogerty with clarifications about its proper application. Ideology and attorney fee awards The decision to award attorney’s fees is, of course, a determination made apart from the resolution of the substantive claims at issue and addresses litigation conduct rather than a party’s substantive intellectual property rights.5 Even so, these sorts of collateral issues are 4 Finally, an important institutional factor distinguishes the potential scope of impact between these interventions: patent appeals are exclusively heard by the Federal Circuit, whereas appeals in copyright cases are distributed across the twelve regional circuits. Even setting aside the doctrinal differences noted above, this structure ensured Octane would impact 100% of patent cases (where all district courts were bound by Brooks Furniture); Kirtsaeng affected only copyright fees in circuits with approaches requiring correction. Both domains involve comparable IP fee-shifting provisions adjudicated by federal district judges. The contrast between expanded discretion in patents post-Brooks Furniture and the Court’s reaffirmation of Fogerty offers a helpful comparative framework for our analyses. 5 Astute readers know the Supreme Court’s activity in patent law around this period was not limited to the fee-shifting cases we focus on. In particular, Myriad Genetics (June 2013) and Alice Corp. (June 2014) tightened patent eligibility standards related to biotechnology and software patents, respectively. These substantive changes to patent law concerned patent eligibility under 35 U.S.C. §101 rather than feeshifting under § 285. Beyond that, our placebo tests show no significant changes in fee-shifting behavior in 2013 when Myriad Genetics was decided. The timing of Alice is more proximate but, again, those two decisions about eligibility under § 101 were focused on specific IP categories; our controlling for areas of IP law in our analyses helps account for any differential effects those decisions might have had across patent types. Nor did the Federal Circuit issue any decisions after 2014 that substantively altered the Octane Fitness framework for fee-shifting; its post-2014 jurisprudence has been characterized by applying the Octane Fitness standard to specific factual scenarios (e.g., Lumen View v. FindtheBest.com 2016, which affirmed the district court’s judgment of exceptionality under § 285 but required recalculation of the fee
8 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 susceptible to the influence of ideologically charged decision-making. One recent study of district court decisions to stay patent litigation pending resolution of other relevant administrative proceedings found that “conservative district judges approach motions to stay consistent with conservative concerns regarding frivolous litigation” (Miller and Curry 2022, 167). That finding aligns with broader patterns in civil litigation, where concern about so-called “frivolous lawsuits” carry ideological valences, with conservative judges and politicians typically being more receptive to arguments about litigation abuse than their liberal counterparts. Common complaints relate to the burdens such litigation inflicts on the judicial system as well as the imposition of unwarranted costs on defendants (Reinert 2014). Patent litigation has become a particular flashpoint for these debates, with critics targeting “vexatious patent litigation” that allegedly “costs defendants and taxpayers tens of billions of dollars each year and delays justice for those who legitimately need a fair hearing of their claims” (Rader, Chien, and Hricik 2013). The non-practicing entity or “patent troll” has emerged as the archetypal example of such problematic litigation. These concerns over frivolousness appear to “animate the ideological dimension of making” in many patent cases (Miller and Curry 2022, 182). Interest group politics has also amplified ideological distinctions around issues of frivolous litigation, court access, and judicial efficiency. The push to limit frivolous lawsuits reflects clear partisan divides, with Republican politicians and conservative-leaning organizations like the US Chamber of Commerce and the American Tort Reform Association advocating strongly for civil litigation reform (Miller and Curry 2022, 174). Within the patent system in particular, major technology companies have often aligned themselves with these conservative interests in advocating for restrictions on patent litigation—especially those cases brought by non-practicing entities (Hosie 2008, 87). Fee shifting is a remedy that is directly responsive to these concerns. It allows courts to penalize parties who bring especially weak claims, while also sending a message to future potential litigants about the financial risks of pursuing questionable cases. In other words, fee-shifting creates a deterrent against frivolous lawsuits without restricting legitimate access to the courts. With the above framework in mind, we have three core expectations about the Supreme Court’s cases and their influence on district court decision-making in cases dealing with attorney’s fees. First, as a doctrinal and ideologically diagnostic perspective, we anticipate that fee awards in patent cases will become more likely after the Supreme Court’s decisions. Hypothesis 1 (Increased Discretion): District court judges will be more likely to award attorney’s fees in patent cases after Octane Fitness and Highmark than before the issuance of those decisions. Beyond this baseline expectation, we anticipate the effects of this expansion of judicial discretion to operate asymmetrically across judicial ideology. As previously discussed, conservatives tend to be more receptive to arguments about litigation abuse. Moreover, because the status quo ante already suited the presumed ideological proclivities of more liberal judges by making the awarding of fees difficult, there is further reason to doubt that liberal judges would significantly react to the Court’s shifting of the doctrinal space in the award; Intellectual Ventures v. Trend Micro 2019, which clarified that, although a single act can make a case exceptional, the district court must find the case as a whole stands out as exceptional under the “totality of the circumstances” test the Supreme Court announced in Octane Fitness).
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 9 two 2014 cases. At the same time, more conservative judges’ preferences would have been constrained by the rigid, pre-2014 framework, making it likely that they would be more responsive to the increased flexibility. Hypothesis 2 (Ideological Asymmetry): Conservative judges will be more likely than liberal judges to award attorney’s fees in patent cases after the 2014 decisions, with this effect concentrated among requests from accused infringers. Finally, we consider a third hypothesis to ensure that any findings we might observe in the patent realm are not simply part of broader trends in intellectual property litigation. For reasons we have described, copyright law provides an opportune comparative case for our purposes. If our theoretical framework is correct, any changes in the propensity of district court judges to award attorney’s fees after Octane Fitness and Highmark should be cabined within the patent domain: Hypothesis 3 (Domain-Specific): Any ideological effects observed in patent fee awards after Octane Fitness and Highmark will not extend to copyright fee awards. Data and variables Our data come from Lex Machina and include all patent cases in which a litigant requests the awarding of attorney’s fees between 2007 and 2024. This time period of collection is dictated by data availability from Lex Machina. We also collected a random selection of copyright cases (one-third of the total number of available cases) in which a litigant requested attorney’s fees in this same time period. Our dependent variable is a dichotomous measure of whether a judge voted to allow attorney’s fees in a case (awarded). This variable is coded one if the fees were granted and zero otherwise. For these cases, we collected data on the year in which the motion for attorney’s fees was decided and coded whether it occurred after the relevant Supreme Court decisions in April of 2014 (after SC decision). For cases occurring in 2014, we coded whether they occurred after the decisions in 2014 or before. For cases occurring after April of 2014, this variable is coded one and zero otherwise. In addition to this measure of our intervention, we coded whether or not a case was a patent case (one if yes, 0 otherwise). To account for judicial ideology, we use scores calculated by Boyd (2015), based on the technique first developed by Giles, Hettinger, and Peppers (2001). Higher values of the judicial conservatism measure represent more conservative judges. Our measure of whether a plaintiff is a high-volume plaintiff comes from Lex Machina. Following their direction, we coded as high-volume (equal to one if yes, zero otherwise) any plaintiff who brought more than ten cases in an issue area (either copyright or patent) in a single year. For every year after the initial year in which at least ten cases were filed a litigant is categorized as high-volume. To code for the complexity of these cases we counted the number of docket entries.6 For patent cases, we coded which party sought attorney’s fees (infringer seeks fees), distinguishing between the accused infringers (coded 1) and patent holders (coded 0). This allows us to examine whether any 6 We view docket entries as providing a rough proxy for case complexity, because it captures factors such as the extent of motion practice, discovery disputes, and the duration of litigation. We include the measure as a control variable but have no a priori expectations about its influence.
10 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00
Table 1 Descriptive statistics.
Variable
Mean
SD
Min.
Max.
Awarded
0.32
0.47
0
1
After SC decision
0.53
0.49
0
1
Patent case
0.36
0.48
0
1
Judicial conservatism
0.03
0.44
−0.64
0.69
High-volume plaintiff
0.31
0.46
0
1
Complexity
238
482
5
11,147
Infringer seeks fees
0.62
0.49
0
1
IP subject matter
n/a
n/a
1
6
ideological responses to expanded discretion operates symmetrically or is driven by defendant request as our theory predicts. Accused infringers sought fees in 62 percent of the patent cases in our sample. Finally, we include fixed effects for the subject matter of the intellectual property (IP subject matter) based on six mutually exclusive categories: technology and digital works; creative and entertainment; industrial and mechanical; life sciences and healthcare; consumer and commercial; and materials and chemical sciences. Appendix contains more extensive descriptions of these six categories and their distribution across the district courts. Table 1 displays descriptive statistics for these variables. Method We estimate a triple differences-in-differences design (DDD), that exploits variation across case types (patent vs. copyright), time (before vs. after treatment), and judicial ideology (liberal vs. conservative). This approach allows us to identify the differential effect of the treatment on judicial decision-making while controlling for time trends specific to each case type and ideology group. In a later section, we show that these results are robust using placebo tests that vary the time of treatment artificially as well as demonstrating that the parallel trends assumption holds. Our dependent variable throughout is whether, if asked, a district court judge awards attorney’s fees. Before turning to the regression results, we first demonstrate that our proposed treatment, the change in discretion created by the two 2014 Supreme Court decisions, altered otherwise parallel trends in our target group (patent cases) but not our control group (copyright cases). First, Table 2 shows a simple before/after difference in means across the groups of interest: case type, time, and judicial ideology. The difference in the rate of awarding between liberal and conservative judges7 in patent cases before the 2014 intervention is three percentage points. The difference in the rate of awarding after the intervention is 13 percentage points, a 400 percent increase in the difference between liberal and conservative judges. The before difference is not significant (and liberals are more likely to award), but the after difference is statistically significant, with conservatives more likely to make an award. In contrast, there is essentially no difference across liberal and conservative judges in copyright cases.
7 We use a continuous measure of judicial ideology in all regression models. Table 2 divides judges into “liberal” and “conservative” categories at the middle point of the ideology measure (0) for descriptive clarity in the difference-of-means tables.
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 11
Table 2 Difference in means. Patent Liberal Conservative Difference Copyright Liberal Conservative Difference
Before 2014 0.40 0.37 0.03 (0.31) Before 2014 0.21 0.22 0.01 (0.40)
After 2014 0.42 0.55 0.13 (0.02) After 2014 0.30 0.29 0.01 (0.40)
Figure 1 Parallel trends examination, liberal judges. Award rates in patent and copyright cases before and after the 2014 Supreme Court intervention, smoothed using LOWESS. Figures 1 and 2 display the differences in awards in patent and copyright cases. Figure 1 plots the trends in awards for liberal judges; figure 2 does the same for conservative judges. Results are smoothed using a local polynomial regression technique (LOWESS). The red (copyright) and blue (patent) lines in figure 1 are not perfectly parallel, as the rate of awards in copyright cases seems to increase before the 2014 Supreme Court intervention. In figure 2, we undertake the same exercise for conservative judges. Here the trends are considerably more parallel, with both increasing immediately before the 2014 intervention.8 Based on these visual inspections, we are confident of parallel trends in the award rates for conservative judges, but less confident in parallel trends for liberal judges. 8 Readers may view the upward trend in conservative awards in Figure 2 as undermining the argument that the 2014 Supreme Court decisions were a meaningful intervention. We use smoothing here because there are a relatively small number of cases in any given year and we wish to avoid idiosyncrasies caused by a small N. Another way to frame the trend in Figure 2 is to take the average number of awards in the 5 years before the treatment period and compare them to the 5-year period after the decisions. Note, this accounts for a substantial increase in awards in 2010. In the 5-year period before the 2014 decisions the average award rate is 36%; it rises to 54% in the 5-year period immediately following the decisions. This difference is highly statistically significant (t = 2.23; P = .01).
12 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 Figure 2 Parallel trends examination, conservative judges. Award rates in patent and copyright cases before and after the 2014 Supreme Court intervention, smoothed using LOWESS. Combined with the results from Table 2, we believe there is sufficient evidence for similar trends in the pre-intervention period to proceed to our triple DDD, especially given our special interest in the behavior of conservative judges. Among patent cases affected by Octane, we also examine whether the ideology effect differs depending on which party requests fees. Because Octane expanded discretion for all fee requests and did not specifically target defendant requests—the new standard was agnostic as to who was requesting fees—we view this analysis as testing where the ideology effect shows up in post-Octane cases rather than identifying a separate treatment dimension. That is, we are not claiming defendants were treated differently by Octane than plaintiffs; rather, the standard changed uniformly across all requestors. Instead, we test whether conservative judges used their expanded discretion differentially: more readily granting fees when accused infringers requested them than when patent holders did. We include a three-way interaction to test whether the ideological effects differ by which party requested fees. This supplementary analysis is reported in Appendix, with results visualized in the Results section. Results Table 3 displays results for the DDD regression. The regression includes each of the constituent terms as well as the three-way interaction necessary to test the triple differences-indifferences. Overall, the model fits the data reasonably well, with a highly significant Wald χ2 statistic. The model includes fixed effects for each of the broad substantive areas covered by patent and copyright law. Standard errors are robust and clustered by judge, of which there are 695 in our data.9 A number of terms involved in our interaction are significant, but interpreting them using just coefficients is fraught. To aid in interpretation, we graph the effects of the interaction for patent cases in figure 3. Key to our hypothesis is that the dashed line (representing 9 In the Appendix, we show nested regressions that progressively add in control variables and/or fixedeffects. Results do not vary from those presented here.
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 13
Table 3 Logit regression. DDD regression After SC decision Patent case Judicial conservatism High-volume plaintiff Complexity Interaction terms After SC × patent After SC × conservatism Patent × conservatism After SC × patent × conservatism Constant N N of judges N of districts Subject category fixed effects Wald χ2
Model 1 0.39 (.14)** 0.63 (.22)** 0.02 (.26) −0.26 (.14)* 0.00 (.00) 0.06 (.24) −0.07 (.33) −0.22 (.41) 0.90 (.53)* −1.26 (.22) 1,598 695 64 Yes 72.94 (0.00)
Figure 3 Patent cases. Predicted probability of attorney’s fee awards by judicial ideology before and after the 2014 Supreme Court decisions. results after the Supreme Court decisions) is statistically distinguishable from the solid line (representing results before the 2014 cases) for conservative judges. As judges get more conservative, they become significantly more likely to award attorney’s fees in patent cases after the 2014 Court decisions, but they are less likely to make such awards before those decisions. The results become statistically distinguishable from
14 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 Figure 4 Copyright cases. Predicted probability of attorney’s fee awards by judicial ideology before and after the 2014 Supreme Court decisions. one another at about the 60th percentile of conservatism (0.2 using our measure of conservatism). There is no significant difference for liberal judges in patent cases. Figure 4 repeats this exercise for copyright cases, where it is apparent that there is no statistically significant difference between the before and after periods defined by the Supreme Court decisions of 2014. Using a judge at the 75th percentile of conservatism to illustrate the differences in patent cases, results from Model 1 suggest that such a judge is 20 [10, 30] percentage points more likely to grant attorney’s fees after the Supreme Court intervention compared to before (95 percent confidence intervals are in brackets). Given that the median fee award is $470,000,10 this 20-percentage point increase represents a substantial shift in the potential costs of patent litigation. The same judge in copyright cases shows no statistically significant difference in the likelihood of awarding across the pre- and post-intervention time periods, with a difference of 7 [−1, 14] percentage points. Finally, note that among the control variables, only the high-volume plaintiff variable is significant and it is negative, indicating that attorney’s fees are less likely to be awarded when one of these high-volume plaintiffs files the case. Our assertion is that conservative judges will utilize the greater discretion created by the Octane and Highmark decisions based on a dislike of what they perceive as frivolous litigation. An observable implication of this hypothesis is that we should detect effects mainly for infringers (those being sued for infringement of a patent—i.e., defendants) as compared to those alleging infringement (i.e., plaintiffs). Figure 5 displays results in patent cases when we interact the position of the party requesting fees with ideology and whether the request came before or after the 2014 10 We were unable to collect complete information on award amounts—it is not consistently recoverable from the data to which we have access. We have data on the amount of the award in 94% of our patent cases, but no corresponding data for copyright cases.
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 15 Figure 5 Attorney fee awards by party position. Predicted probability of fee awards in patent cases by judicial ideology and Supreme Court intervention, separated by party requesting fees. (A) Infringers. (B) Non-infringers. Supreme Court decisions. The regression results used to generate these plots are provided in Appendix. A clear, and expected, pattern emerges that is consistent with Hypothesis 2. Namely, there is a statistically significant difference between infringers based on the Supreme Court intervention (panel A), but there is no statistically significant difference for non-infringers based on the Supreme Court intervention (panel B). This is evidence favoring our proposed mechanism—namely that conservative judges are reacting ideologically to perceptions of the frivolity of litigation. Robustness checks To further support the finding presented in Table 3, we undertook a series of placebo tests to see if our proposed date of intervention (2014) best fits the data. Table 4 displays the results of this exercise. Results suggest that assuming the intervention occurring in 2014 is most likely the best fit to the data as this date has the lowest Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) scores. The next best proposed intervention date is 2015, and this too provides evidence in favor of our proposed intervention since the Supreme Court decision was made one-third of the way into 2014 (in April), so some of the effect of the intervention is likely to have taken some of 2015 to diffuse through the lower federal courts. Given this finding, we have added confidence that we have correctly specified the timing of the proposed intervention. Another approach to placebo testing is to restrict the data to the period before the intervention and estimate models using false treatment dates. If the Supreme Court decisions in 2014 are the “true” treatment in the data, then these fake interventions should not be significant predictors. Taking this approach to the data from before 2014, we estimated two separate fake intervention models, one for 2012 and one for 2013. The Wald χ2 tests for these interactions were not significant (P = .12 for 2012 and P = .09 for 2013). This lends additional support to the assumption of parallel trends before intervention.
16 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00
Table 4 Testing intervention timing using placebos.
Year
AIC
2012 2013 2014 2015 2016
1954 1953 1950 1953 1955
BIC 2035 2034 2031 2033 2036
Values in italics indicate the proposed intervention date (2014), which produces the lowest AIC and BIC scores and therefore represents the best fit to the data.
Table 5 Parallel trends test. Relative year Judicial conservatism High volume plaintiff Complexity Interaction terms Relative year × judicial conservatism Constant N N of judges N of districts Subject category fixed effects Wald AUROC
Model 2 −0.14 (.11) 1.29 (.98) −0.59 (.71) −0.00 (.00) 0.29 (.25) −1.32 (.53) 153 116 43 Yes 11.17 (0.26) 0.66
Values in italics indicate the proposed intervention date (2014), which produces the lowest AIC and BIC scores and therefore represents the best fit to the data.
Additionally, to help confirm that there were no threatening pre-treatment trends in the data (and to help bolster our parallel trends assumption), we estimated a regression using relative lags from our treatment date. To formally assess the parallel trends assumption, we estimate a model using only pre-2014 patent cases, regressing fee-award decisions on a linear time trend, judicial ideology, and their interaction. That is, if a case was decided in 2013, its relative lag is −1, if in 2012 it is −2, and so on; an approach that follows others seeking to validate parallel trends assumptions (e.g., Autor 2003). We refer to this variable as relative year. The coefficient on the Relative Year x Judicial Conservatism interaction is small and statistically indistinguishable from zero, as displayed in Table 5. This indicates that conservative and liberal judges followed statistically similar trends in fee awards prior to Octane and Highmark. As a confirmation of the robustness of our substantive findings we used coarsened exact matching (CEM) (Blackwell et al. 2009) to pre-process our data and then re-estimate our triple-difference-in-difference regression (DDD). This approach allows us to address the possibility that our DDD results are driven by differences in the composition of cases assigned to judges. We implemented a two-stage matched DDD procedure. Using only pre2014 cases, we applied CEM to identify judges who heard comparable patent and copyright cases prior to the Supreme Court’s decisions, matching on case complexity, plaintiff type,
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 17
and IP subject-matter category. We then restricted the full dataset (pre- and posttreatment) to this set of judges who share common support and re-estimated the DDD model. The results are substantially unchanged: the three-way interaction remains positive and statistically significant, and the before/after difference for conservative judges in patent cases persists. We illustrate this in figure A1 in Appendix, which mirrors figure 3. This indicates that the observed ideological divergence does not result from differences in case mix or selection into patent versus copyright cases. Rather, the effect reflects a behavioral response among conservative judges to the expanded discretion created by Octane Fitness and Highmark. This matched-DDD check strengthens the causal interpretation of our main findings.
Conclusion In this manuscript, we demonstrate that increased discretion in awarding attorney’s fees can lead to asymmetrical ideological outcomes. Specifically, conservative judges are significantly more likely to award attorney’s fees when accused infringers request them in patent cases, consistent with conservative preferences for deterring weak patent assertions. Additionally, we compare trends in patent cases with those in copyright cases to reinforce our understanding that the observed ideological divergence stems from two Supreme Court decisions in 2014, which granted lower courts greater discretion to award fees in patent cases. Use of a triple DDD meant to rule out spurious time trends as the cause of the results we observe gives us heightened confidence. More broadly, our study has implications for the notion that increased discretion for lower court judges is likely to have non-neutral ideological consequences. This ideological imbalance is especially likely to be salient given the Supreme Court’s decision in Loper Bright (2024) to end Chevron (1984) deference to administrative agency rule interpretations. The ideological implications of the decision in Loper Bright are clear enough if one considers that all six conservative Justices voted in favor of increased discretion, while all three liberals voted against it. It shifts discretion from (perceived) liberal federal agencies to a more conservative set of decision makers (federal district court judges). We show that a seemingly less ideologically motivated move of this type in patent law has increased the ability of conservative federal judges to put their preferences into practice. Finally, our study has implications for patent law itself and for studies of the responsiveness of lower courts to rules established by the Supreme Court. With respect to patent law post-2014, conservative judges became substantially more likely to award attorney’s fees in patent cases, altering the litigation landscape. Further, and consistent with conservative preferences for deterring weak patent assertions, our analysis reveals this effect operates primarily when accused infringers defeat patent claims and request fees. This ideological diversion in fee-shifting may create strategic forum shopping incentives for patent litigants. Our work also supports the notion that lower federal court judges are responsive to the policy directions given to them by the Supreme Court (e.g., Masood, Kassow and Songer 2019), but that a district court judge’s own policy preferences may dictate the penetration of such proscriptions from the Court.
None declared.
Funding
18 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00
None declared.
Conflicts of interest
Appendix As described in the manuscript, we created six mutually exclusive categories to capture any possible impact of the general types of intellectual property subject matter at issue in each case. In constructing these categories—technology and digital works; creative and entertainment; industrial and mechanical; life sciences and healthcare; consumer and commercial; and materials and chemical sciences—we attempted to organize intellectual property into clear domains while accommodating the particular nature of the intellectual property at issue (i.e., patents and copyrights). The technology and digital works category included computer systems, software code, websites, databases, and networking technologies. For patents, this primarily covered computing-related inventions, telecommunications, and digital methods. For copyright, this included websites and digital media. Creative and entertainment represented the most common copyright category, covering artwork, songs, photographs, and books. The few patents in this category involved entertainment systems and devices. The industrial and mechanical category was most prominent in patents related to machinery, construction methods, and industrial equipment; for copyrights, it included architectural works, blueprints, and technical drawings. Life sciences and healthcare includes medical devices, pharmacological compositions, biological designs, and healthcare literature. Consumer and commercial includes fashion designs and apparel, commercial methods, retail products, furniture, and business systems. Finally, materials and chemical sciences includes chemical patents, scientific publications, and material innovations. Table A1 illustrates the distribution of patent and copyright cases across these six subject matter categories. Table A2 displays the distribution of cases across federal judicial districts. Our data encompasses cases from sixty-four districts, with representation across all major regions. The values in the table indicate the mean number of cases per district for patent and copyright cases, respectively. “—” indicates no cases in that category for the district.
Table A1 Distribution of cases by IP subject matter category.
IP category
Patent (%)
Technology and digital
37
works
Creative and
3
entertainment
Industrial and mechanical
18
Life sciences
16
and healthcare
Consumer
23
and commercial
Material and chemical
3
sciences
Copyright (%) 8 62 4 1 22 3
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 19
Table A2 Distribution of cases by judicial district.
District
Patent
Maine
5
Massachusetts
3.4
Puerto Rico
5
Connecticut
3
EDNY
4.3
NDNY
1
SDNY
3.5
WDNY
5
Delaware
2.2
New Jersey
4.2
EDPA
2.8
WDPA
4
Maryland
5
EDNC
2.3
WDNC
4
South Carolina
4.7
EDVA
2.3
NDWV
4
EDLA
2
WDLA
3
EDTX
1.9
NDTX
3.4
SDTX
3.4
WDTX
3.6
EDMI
2
WDMI
4.2
NDOH
2.3
EDTN
3
MDTN
4
WDTN
2
CDIL
6
NDIL
3.6
SDIL
4
SDIN
3.5
EDWI
4.6
WDWI
2.9
EDAR
3
NDIA
4
SDIA
4.5
Minnesota
3.9
EDMO
4
Nebraska
4
North Dakota
3
Arizona
3.2
Copyright 2 2.5 2 1.9 2.4 2.6 2.7 2 2.4 3 2.3 2 2.4 2.7 2.2 2 3.1 — 2.5 2 3 3.4 3.7 3.7 4 3 3.3 2 2.9 3 2.8 2.6 2 3.4 2.3 2 2.5 2.5 2.8 3 2.1 2.5 3 2.8 (continued)
20 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00
Table A2. (continued) District CDCA EDCA NDCA SDCA Hawaii Idaho Nevada Oregon WDWA Colorado Kansas NDOK WDOK Utah MDAL NDAL MDFL NDFL SDFL NDGA
Patent 3.2 3.8 1.9 2.8 3 1 2.8 3 2.9 3.8 1 2.5 4.3 3.6 5 1 3.3 4 3.6 4.5
Copyright 3.3 2.5 2.4 3 2.5 2.5 2.2 2.5 2.1 2 3.7 — 3.5 2.1 — 2 2.4 3.7 2.4 2.3
Table A3 Nested regressions. After SC decision Patent case Judicial conservatism High volume plaintiff Complexity Interaction terms After SC × patent After SC × conservatism Patent × conservatism After SC × patent × conservatism Constant N N of judges N of districts Subject category fixed effects Wald AUROC *indicates p < 0.05; **indicates p < 0.01
Model A1 0.40 (.14)** 0.61 (.21)** 0.11 (.13) −0.28 (.14)* 0.00 (.00) 0.14 (.24) — — — −1.28 (.22) 1,598 695 64 Yes 64.60 (0.00) 0.64
Model A2 0.43 (.14)** 0.81 (.18)** 0.02 (.25) — — −0.06 (.24) −0.09 (.33) −0.21 (.40) 0.93 (.52) −1.29 (.11) 1,598 695 64 No 63.97 (0.00) 0.63
Model A3 0.39 (.14)** 0.63 (.22)** 0.02 (.26) −0.26 (.14)* 0.00 (.00) 0.06 (.24) −0.07 (.33) −0.22 (.41) 0.90 (.53) −1.26 (.22) 1,598 695 64 Yes 72.94 (0.00) 0.65
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 21
Table A4 Matched sample regression. CEM After SC decision Patent case Judicial conservatism High volume plaintiff Complexity Interaction terms After SC × patent After SC × conservatism Patent × conservatism After SC × patent × conservatism Constant N N of judges N of districts Wald AUROC *indicates p < 0.05; **indicates p < 0.01
Model A4 0.37 (.18)* 0.50 (.22)* 0.03 (.25) −0.31 (.17) 0.00 (.00) 0.35 (.36) −0.26 (.40) −0.11 (.41) 0.68 (.72) −1.22 (.14) 1,057 426 48 51.32 (0.00) 0.67
Figure A1 Matched patent cases. Nested regressions To demonstrate the way that control variables affect our regression results we show in Table A3 nested regressions, which progressively add variables. Model A3 below is the same as Model 1 presented in the manuscript. Model A1 includes one interaction and IP subject matter fixed effects; Model A2 is stripped-down to include only the essential
22 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00
Table A5 Balance statistics for CEM. CEM diagnostics All Matched Unmatched Balance Complexity High-volume plaintiff IP code Overall
Copyright 507 505 2 Before 0.55 0.25 0.28 0.71
Patent 247 226 21 After 0.36 0.00 0.23 0.58
Table A6 Regression with infringer interactions. DV: awarded After SC decision (+) Judge ideology (+) Infringer Number of patents (−) High volume plaintiff (+) Complexity Interactions After SC × judge ideology After SC × infringer Infringer × judge ideology After SC × infringer × ideology Constant N N of districts Patent type FE Wald PRE *indicates p < 0.05; **indicates p < 0.01
Model A5 0.00 (.31) 0.23 (.47) −1.01 (.29)** 0.04 (.03) 0.76 (.28)** −0.00 (.00)** 0.78 (.67) 0.59 (.39) −0.96 (.64) 0.26 (.85) −0.07 (.30) 569 64 Yes 54.74 (0.00) 0.17
interactions and has no fixed-effects for subject matter area. As is clear from the coefficients on the constituent and interaction term, substantively there is little difference between Model A2 and Model A3. Robustness check To support the analysis presented in the manuscript, we estimated a model based on coarsened exact matching (CEM) in the data. Specifically, it may be that case composition for judges changes across our treatment (the relevant Supreme Court cases). Using only pre2014 cases (pre-Supreme Court treatment), we applied CEM to identify judges who heard comparable patent and copyright cases, matching on case complexity, plaintiff type (highvolume or not), and intellectual property subject category (as described above). We then restricted the full dataset (pre- and post-treatment) to this set of judges who share common
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 23 support and who heard cases on both sides of the proposed treatment. We then reestimated the triple-differences model displayed as Table 3 in the manuscript using this restricted set of observations with common support. Table A4 displays the results of the regression analysis after this matching preprocessing. Again, it is difficult to understand the substantive effects by looking only at the coefficients, so we provide a companion to figure 3 in the manuscript to illustrate the results: figure A1. Balance statistics are provided in Table A5, which shows significantly improved (if not perfect) balance across the three included variables. The results closely mirror those presented in the manuscript giving us added confidence that the treatment we propose is likely causal. Given the pre-processing of the data to match on case characteristics and the restriction of the data to only judges participating before and after the treatment we have eliminated the possibility that our results are driven by case composition differences. Put differently, the effect persists even after making patent and copyright cases as similar as possible on observable characteristics. Results to support figure 5 in the manuscript are presented in Table A6. Note, we have a more specific measure of complexity in patent cases (the number of patents at issue in the litigation, drawn from Lex Machina) here along with our measure of docket counts (complexity). References Autor, David H. 2003. “Outsourcing at Will: The Contribution of Unjust Dismissal Doctrine to the Growth of Employment Outsourcing.” Journal of Labor Economics 21: 1–42. Baum, Lawrence. 1990. “Specializing the Federal Courts: Neutral Reforms or Efforts to Shape Judicial Policy.” Judicature 74: 217–24. Baum, Lawrence. 2010. Specializing the Courts. Chicago: University of Chicago Press. Blackwell, Matthew, Stefano Iacus, Gary King, Giuseppe Porro. 2009. “CEM: Coarsened Exact Matching in Stata.” The Stata Journal 9: 524–46. Boyd, Christina L. 2015. “Federal District Court Judge Ideology Data.” http://cLboyd.net/. Devinsky, Paul. 2016. “Kirtsaeng II: Fees in Copyright Cases Depends on Reasonableness of Litigation Position.” https://www.mcdermottlaw.com/insights/kirtsaeng-ii-fees-copy right-cases-reasonableness/. Downing, Kristy J. 2021. “Rogue One: Section 285 Attorney’s Fees in Doctrinal Patent Ineligibility Cases.” Texas Intellectual Property Law Journal 30: 23–73. Epstein, David, and Sharyn O’Halloran. 1999. Delegating Powers: A Transaction Cost Politics Approach to Policy Making Under Separation of Powers. New York: Cambridge University Press. Giles, Micheal W., Virginia A. Hettinger, and Todd C. Peppers. 2001. “Picking Federal Judges: A Note on Policy and Partisan Selection Agendas.” Political Research Quarterly 54: 623–41. Hosie, Spencer. 2008. “Patent Trolls and the New Tort Reform: A Practitioner’s Perspective.” I/S: A Journal of Law and Policy for the Information Society 4: 75–87. Huber, John D., and Charles R. Shipan. 2002. Deliberate Discretion? The Institutional Foundations of Bureaucratic Autonomy. New York: Cambridge University Press. Hyde, Robert Aloysius, and Lisa M. Sharrock. 2004. “A Decade Down the Road but Still Running through the Jungle: A Critical Review of Post-Fogerty Fee Awards.” University of Kansas Law Review 52: 467–89.
24 • The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 Jiam, Hannah. 2015. “Fee-Shifting and Octane Fitness: An Empirical Approach Toward Understanding ‘Exceptional.’” Berkeley Technology Law Journal 30: 611–74. Johnson, Marilyn F., Karen K. Nelson, and Adam C. Pritchard. 2007. “Do the Merits Matter More? The Impact of the Private Securities Litigation Reform Act.” The Journal of Law, Economics, and Organization 23: 627–52. Jones, Darin. 2015. “A Shifting Landscape for Shifting Fees: Attorney-Fee Awards in Patent Suits After Octane and Highmark.” Washington Law Review 90: 505–43. Lax, Jeffrey R. 2012. “Political Constraints on Legal Doctrine.” The Journal of Politics 74: 765–81. Masood, Ali S., Benjamin J. Kassow, and Donald R. Songer. 2019. “The Aggregate Dynamics of Lower Court Responses to the U.S. Supreme Court.” Journal of Law and Courts 7: 159–86. Miller, Banks and Brett Curry. 2022. “To Stay or Not to Stay: Patent Litigation in the Federal District Courts.” Journal of Law and Courts 10: 167–88. Percival, Robert V. and Geoffrey P. Miller. 1984. “The Role of Attorney Fee Shifting in Public Interest Litigation.” Law and Contemporary Problems 47: 233–47. Rader, Randall, Colleen V. Chien, and David Hricik. 2013. “Make Patent Trolls Pay in Court.” New York Times, June 4, 2013. https://www.nytimes.com/2013/06/05/opin ion/make-patent-trolls-pay-in-court.html. Reinert, Alexander A. 2014. “Screening out Innovation: The Merits of Meritless Litigation.” Indiana Law Journal 89: 1191–236. Root, David A. 2004. “Attorney Fee-Shifting in America: Comparing, Contrasting, and Combining the ‘American Rule’ and ‘English Rule.’” Indiana International and Comparative Law Review 15: 583–617. Ryan, Sarah E. 2020. “Judicial Authority Under the First Step Act: What Congress Conferred Through Section 404.” Loyola University Chicago Law Journal 52: 67–125. Singer, Jacob. 2010. “Bad Faith Fee-Shifting in Federal Courts: What Conduct Qualifies?” St. John’s Law Review 84: 693–726. Stith, Kate and Jose A. Cabranes 1998. Fear of Judging: Sentencing Guidelines in the Federal Courts. Chicago: The University of Chicago Press. Cases cited Alice Corp. v. CLS Bank International, 573 U.S. 208 (2014). Ashcroft v. Iqbal. 556 U.S. 662 (2009). Association for Molecular Pathology v. Myriad Genetics, Inc., 569 U.S. 576 (2013). Bell Atlantic v. Twombly. 550 U.S. 544. (2007). Brooks Furniture Manufacturing, Inc. v. Dutailier International Inc. 393 F.3d 1378 (Fed. Cir. 2005). Chevron USA Inc. v. Natural Resource Defense Council, 467 US. 837 (1984) Fair Winds Sailing, Inc. v. Dempster, 764 F.3d 303 (2014). Fogerty v. Fantasy. 510 U.S. 517 (1994). Highmark Inc. v. Allcare Health Management System, Inc. 572 U.S. 559 (2014). Illinois v. Gates. 462 U.S. 213 (1983). Intellectual Ventures LLC v. Trend Micro Inc., 944 F. 3d 1380 (Fed. Cir. 2019).
The Journal of Law, Economics, and Organization, 2026, Vol, 00, Issue 00 • 25 Kirtsaeng v. John Wiley & Sons, Inc. 579 U.S. 197 (2016). Law v. Siegel. 571 U.S. 415 (2014). Loper Bright Enterprises v. Raimondo. 603 U.S. 369 (2024). Lumen View Technology LLC v FindTheBest.com. 811 F.3d 479 (Fed. Cir. 2016). Mathis v. Spears, 857 F.2d 749 (Fed. Cir. 1988). Miranda v. Arizona. 384 U.S. 436 (1966). Octane Fitness, LLC v. Icon Health & Fitness. 572 U.S. 545 (2014). Strickland v. Washington. 466 U.S. 668 (1984). United States v. Booker. 543 U.S. 220 (2005).
15 U.S.C. §1117(a) 17 U.S.C. §505 35 U.S.C. §101 35 U.S.C. §285
Statutes cited