EXPERTS, GENERALISTS, LAYPEOPLE, AND THE FEDERAL CIRCUIT. (2024)

TABLE OF CONTENTS I. INTRODUCTION 575 II. THE PATENT LANDSCAPE TODAY 579 A. Administrative Overhaul: from the AIA to Oil States 579 B. District Court Changes: Concentration, Pilots, and the 585 Pursuit of JuriesIII. METHODOLOGY 591 A. Data Source 591 B. Coding Procedure 592 IV. ANALYSIS 596 A. PTAB Data 597 B. District Court Data 605 C. Drawing Comparisons 610 D. Critiques and Concerns 617 V. MAKING SENSE OF THE DATA: ANTI-EXCEPTIONALISM AND 622 CONTEXT-SPECIFIC DEFERENCE VI. CONCLUSION 631

I. INTRODUCTION

I CANNOT STOP WITHOUT CALLING ATTENTION TO THE EXTRAORDINARY CONDITIONOF THE LAW WHICH MAKES IT POSSIBLE FOR A MAN WITHOUT ANY KNOWLEDGE OFEVEN THE RUDIMENTS OF CHEMISTRY TO PASS UPON SUCH QUESTIONS ASTHESE... FOR ONLY A TRAINED CHEMIST IS REALLY CAPABLE OF PASSING UPONSUCH FACTS... HOW LONG SHALL WE CONTINUE TO BLUNDER ALONG WITHOUT THEAID OF UNPARTISAN AND AUTHORITATIVE SCIENTIFIC ASSISTANCE IN THEADMINISTRATION OF JUSTICE, NO ONE KNOWS.--JUDGE LEARNED HAND (1)

Who should decide a patent case? There is no shortage ofanecdotal--and contradictory--answers, depending on who is asked. Jurorsare either woefully unqualified (2) or paragons of virtue (3). Judgesneed either more technical specialization (4)--or less. (5) Andadministrative tribunals are either "patent death squads" (6)or the only bulwark left against innovation-choking trolls (7).

The patent landscape has changed radically in the past decade,particularly in terms of decision-makers. The Leahy-Smith AmericaInvents Act ("AIA") (8) expanded the U.S. Patent and TrademarkOffice's authority as an administrative tribunal and greatlyincreased third-party participation in the patent grant (and post-grant)process. (9) The Patent Cases Pilot Program now funnels many patentcases to judges with enhanced patent training and experience, on top ofa litigant-driven concentration among a handful of districts and agrowing demand for juries. (10) Meanwhile, more than half of the seatson the Court of Appeals for the Federal Circuit--the sole appellatetribunal of right for these matters--have changed hands. (11)

Recent empirical studies of patent litigation have offeredconsiderable insight into, for example, the significance of differentareas of patented technology (12) or particular district courts (13) oncase outcomes. But few have shed light on the differences betweenclasses of adjudicators: administrative patent judges on the PatentTrial and Appeal Board ("PTAB"), district court judges ofvarious stripes, and juries. (14) And no studies offer a comprehensivepicture of the natural experiment currently unfolding. That is,comparable patent validity issues are decided separately by judges,juries, and administrators, but all of their decision-making is reviewedin turn by a singular, controlling entity: the Federal Circuit.Accordingly, examining and comparing the results of appeals to theFederal Circuit from each of these adjudicators offers a particularlyclear window into the relationships between these entities--and into thevarying effects of expertise and specialization in the patent worldoverall.

This Article capitalizes on the current adjudicatory structure ofpatent law, analyzing more than two thousand Federal Circuit cases andopinions--each hand-coded for validity findings and their disposition onappeal issue-by-issue. Though an incredibly time-intensive approach,(15) the result is a uniquely complete and clear dataset. Astraightforward empirical analysis of the data, moreover, challengesseveral longstanding and widespread anecdotal assumptions about patentadjudication.

In brief, the data suggests that the Federal Circuit affirmsfindings made by the PTAB reliably more often than findings made bydistrict court judges--particularly when the findings involve questionsof fact rather than questions of law. Whether the district court judge(or district itself) has more prior experience with patent cases appearsto be irrelevant to appellate results. whether the finding was made by ajury, on the other hand, is highly relevant, with those findingsaffirmed at the highest rate. Moreover, PTAB findings that invalidatepatent claims are affirmed more often than findings that uphold patentclaims. No similar pattern exists in district court appeals. And theunderlying technological subject matter of the patent at issue does notseem to perceptibly influence results on appeal for either.

The dataset thus tells a story about the Federal Circuit and patentlitigation that is more subtle than the dominant narratives have been todate. Despite a generally non-deferential relationship to the PTAB, theFederal Circuit appears to be placing greater faith in the scientificexpertise of its administrative patent judges. Likewise, given thehigher rate of affirmance for claim-invalidating findings from the PTAB,the "death squad" narrative seems entirely misplaced, unlessone accords the Federal Circuit the same appellation.

In contrast to scientific knowledge, pure legal specialization (asexemplified by the patent-heavy dockets among a handful of Texas andDelaware judges) appears to only have mixed results, at best, in termsof appellate outcome. Instead, it is findings made by jurors--theultimate laypersons and generalists--that survive most easily on appeal.

These trends suggest that the Federal Circuit has perhaps begun tointernalize the longstanding anti-exceptionalist messaging from scholarsand the Supreme Court. (16) That is, the Federal Circuit is treating thepatent space more and more like other areas of law. It is normalizingits relationship with the PTAB as one with more administrativedeference. It is declining to recognize an extra-legal stratificationamong district court judges. And it is treating jury-made findings withthe same level of extraordinary respect that they are afforded in lessscientific and technical areas of law. Given the nuances of the PTABdata, the Federal Circuit may also have internalized the policy-orientedmessaging behind Congress's most recent overhaul of patent law: badpatents need to go. (17)

The shape of patent litigation itself appears unlikely to changedrastically in the near future. subsequent scholarship would be wellserved to examine comparable data in some years' time to determinethe full durability of the results presented herein. They may turn outto merely reflect a transitory settling-in period after the fullimplementation of the AIA, but the Author suspects that they reflect anemergent status quo.

The remainder of the Article proceeds in four parts. Part IIbriefly outlines the recent changes to and overall trends in the patentlandscape, chief among them the drastically increased role of the Patentand Trademark Office and the long-term concentration of district courtlitigation. Part III describes the data collection process and codingmethod, including particular challenges and solutions. Part IV presentsvarious tabulations of the data, noting the most significant patternsand disparities of interest and addressing quantitative-level concernsor critiques. Finally, Part V concludes by contextualizing the empiricalresults among larger developments in patent law and the Federal Circuitas an institution.

II. THE PATENT LANDSCAPE TODAY

A. Administrative Overhaul: from the AIA to oil states

It is difficult to overstate the magnitude of changes to the Unitedstates patent system over the past decade--starting first and foremostwith the changes to its administrative apparatus: the U.S. Patent andTrademark Office ("USPTO").

In 2011, Congress passed the AIA, (18) "the most significantoverhaul to our patent system[] since the founding fathers." (19)Though historically considered a "weak agency" due to a lackof judicial deference or policymaking authority, (20) the AIA markedlyenhanced the USPTO's role in shaping patent law. Most importantly,the AIA established a battery of new adjudicative proceedings before thePTAB: post-grant review (21) and inter partes review, (22) wherein thirdparties may challenge the validity of a recently issued patent;derivation proceedings, wherein multiple parties contest ownership of asingle invention; (23) and supplemental examinations, wherein a patentowner may seek to correct errors made during prosecution. (24)

Crucially, these proceedings are formal and "trial-like"(25)--including everything from pretrial discovery and witness testimonyto cross-examination and oral argument--which bears on judicialdeference for regulations or rulings resulting therefrom. (26) Indeed,the USPTO's entitlement to Skidmore deference for itsinterpretation of the AIA has become increasingly accepted. (27) And, ata minimum, the AIA explicitly granted rulemaking power to the USPTO overthe "procedures" and "standards" applied in its newproceedings. (28)

These proceedings have, moreover, proven quite attractive tolitigants. More than 7,000 inter partes and post-grant review petitionshave been filed since the AIA's implementation in September of2012. (29) As recently as FY 2013, before these new proceedings began togenerate appeals, there were fewer than 150 appeals per year from theUSPTO to the Federal Circuit. (30) In FY 2016, there were nearly 650.(31) This drastic, steady increase has caused USPTO appeals, once amongthe smallest share of the Federal Circuit's docket, to supplantdistrict court appeals as the leading source of cases before theCircuit. (32)

At the same time, there has been vocal pushback against theUSPTO's expanded role. Former Chief Judge of the Federal Circuit,Randall R. Rader, once labeled the PTAB a patent "deathsquad," (33) referring to the PTAB's high invalidation rate:nearly 70% of all final written decisions in inter partes reviewinvalidate the patent-at-issue in full and a further 16% invalidate atleast some patent claims. (34) Patent owners and academics have broadlycriticized the new adjudicatory structure as generating costlylitigation, reducing incentives to innovate, and increasing marketplacerisk. (35) These critiques appear to have gained at least some tractionin Congress, with a handful of senators concluding that the AIA has had"unintended consequences" and proposing legislation that wouldmake it more difficult for the PTAB to invalidate challenged patents.(36)

On the other side, supporters of the USPTO's newpatent-invalidating procedures have argued that their seeminglyaggressive usage is a much-needed check on--and natural response to--apatent-granting system run amok. The rate at which the USPTO issuespatents has increased dramatically in recent years, spiking fromapproximately 90,000 utility patents granted in 1990 to more than300,000 by 2014. (37) In particular, the USPTO has issued an increasingnumber of software patents (now a majority of all patents granted (38)),which tend to be disproportionately litigation-prone. (39) Indeed,commentators observe that software patents are often especiallyimprecise and overbroad, whether due to oversight in the applicationprocess itself (40) or poorly defined underlying case law. (41)

Exacerbating matters, they argue, are patent trolls--entities thatacquire and enforce patents without actually practicing them--who havecapitalized on this broken system, using improvidently-issued patents tothreaten litigation against entire industries (42) and even user bases.(43) Supporters argue that the new USPTO proceedings are the best meansto disarm trolls, (44) and action groups have, in fact, seized theopportunity therein. (45) In brief, proponents suggest that the AIAoffers crucial balance that, in the long run, will greatly strengthenthe united States patent system overall.

The clash between these two viewpoints culminated last term inSupreme Court litigation challenging the constitutionality of theUSPTO's new AIA procedures outright: Oil States Energy Services,LLC v. Greene's Energy Group, LLC. (46) Petitioner Oil States--wholost its patent in an inter partes review proceeding before theUSPTO--argued that patents, as "private property rights," canonly be extinguished by an Article III court, not an Article Iadministrative tribunal such as the PTAB. (47) Respondent Greene'sEnergy--who had challenged Oil States's patent in the firstplace--argued that patents are "mere public rights," such thatPTAB review and revocation present no constitutional problem. (48)Despite the doctrinal nature of the litigants' core positions, thepolicy arguments outlined above clearly colored and informed theanalysis on both sides. (49)

Ultimately, the Court held that post-grant invalidation by theUSPTO is not unconstitutional:

Inter partes review falls squarely within the public-rights doctrine.This Court has recognized, and the parties do not dispute, that thedecision to grant a patent is a matter involving publicrights--specifically, the grant of a public franchise. Inter partesreview is simply a reconsideration of that grant, and Congress haspermissibly reserved the PTO's authority to conduct thatreconsideration. Thus, the PTO can do so without violating Article III.(50)

Practitioners and policymakers will doubtless continue to disputeand push on the precise boundaries of the USPTO's authority. Butthe Court's decision in Oil States means that proceedings beforethe USPTO will retain the same general shape for some time. The analysisof the USPTO's current relationship to the Federal Circuit anddistrict courts presented herein may therefore be fairly projected intothe foreseeable future.

B. District Court Changes: Concentration, Pilots, and the Pursuitof Juries

In parallel to the USPTO, federal district courts have undergonetheir own transformation. Specifically, three trends have changed theface of Article III patent adjudication over the past decade: theconcentration of patent litigation among a handful of districts, thePatent Pilot Judge Program, and the increasing reliance on juries.

First, the distribution of patent suit filings across the countryhas shifted dramatically in recent years. In 2001, the top three patentlitigation districts at that time--the Central and Northern Districts ofCalifornia as well as the Northern District of Illinois--received lessthan 22% of all patent cases in the country. (51) By 2016, the top threedistricts--now the Eastern District of Texas, the District of Delaware,and the Central District of California--accounted for more than 58%.(52) The Eastern District of Texas alone received 44%. (53)

Whereas the Central District of California's longstandingpreeminence is largely seen as a by-product of its tech sectorproximity, the sharp rise of the Eastern District of Texas and theDistrict of Delaware is commonly critiqued as the product of forumshopping for favorable procedural and administrative rules. (54) Inthese two districts, deadlines are accelerated, discovery is broad,defendant joinder is easy, and stays are exceptionally rare--even if theunderlying patent's validity is simultaneously under scrutiny atthe USPTO. (55) TO wit:

The preeminence of the Eastern District of Texas and the District ofDelaware as venues for patent litigation makes no sense according tomost economic indicators; it cannot be explained in terms offundamentals, such as economic activity, economic growth or the size ofthe local population... The Eastern District of Texas has gone to greatlengths to bend almost every procedural aspect of patent litigation infavor of plaintiffs. The District of Delaware has gone down the samepath, but not quite as far. (56)

Raw data suggests that this plaintiff-friendly explanation has atleast some truth to it; using 1997-2016 data, the Eastern District ofTexas had the highest plaintiff success rate of any district in thecountry, with the District of Delaware only a few spots behind. (57)

Pushback against this concentration has been considerable, rangingfrom criticism in popular media (58) to attempts at legislative reform(59) to Supreme Court litigation. In 2017, the Supreme Court heard TCHeartland LLC v. Kraft Foods Group Brands LLC, (60) a direct challengeto the governing interpretation of the patent venue statute (61) thathad enabled plaintiffs such considerable freedom in forum selection inthe first place. Under VE Holding Corp. v. Johnson Gas Appliance Co.,the Federal Circuit had held that venue was proper in patent suitsagainst corporate defendants in "any judicial district in which[the] defendant is subject to the court's personaljurisdiction." (62) In a unanimous opinion, the Supreme Courtrejected the Federal Circuit's broad interpretation, holding thatvenue is proper against a corporate defendant only where it: (1) isincorporated; or (2) "has a regular and established place ofbusiness." (63) When subsequently applying TC Heartland, theFederal Circuit held that a "regular and established place ofbusiness" specifically requires "physical" presence inthe district--not "merely... a virtual space or... electroniccommunications." (64)

This doctrinal arc represents a considerable narrowing of patentvenue choice--but not a significant reversal in concentration. Indeed,immediately after TC Heartland, early statistics indicated that patentcase filings decreased sharply in the Eastern District of Texas; (65)that shortfall, however, was largely offset by a commensurate increasein the District of Delaware. (66) More than a year later, that patternhas held steady. For the six-month period preceding TC Heartland, thetop five districts had a combined share of 61% of patent case filings;for the six-month period preceding this writing, that same combinedshare was 60%. (67) In other words, it appears that the forum-shoppingstopped by TC Heartland was occurring predominantly between districtsthat were already patent heavy; it was not pulling litigation away fromthe rest of the country. Regardless of which particular districts becomethe locus for patent litigation moving forward, concentration itselfthus appears to be here to stay.

Second, within certain districts, the Patent Pilot Program (68) hasfunneled patent cases to increasingly patent-experienced judges. Startedin 2011, the ten-year program has allowed the judges in thirteendistricts (69) to transfer any patent cases they are randomly assignedto the designated pilot judges in their district instead. (70) Theintent behind the program is "to encourage enhancement of expertisein patent cases among district court judges" by offering thedesignated judges greater exposure to and familiarity with patent cases.(71) Mid-pilot data from the Administrative Office of the U.S. Courtsconfirms that designated judges have indeed been receiving substantiallymore patent cases due to their participation; hundreds of patent caseshave been transferred in most of the selected districts, and more than3,000 have been transferred in the first five years of the programoverall. (72) This same preliminary data indicates that patent casesassigned to designated judges are terminated more quickly, cautiouslysuggesting gains in efficiency. (73)

Third, regardless of forum, the use of juries in patentinfringement suits has been steadily increasing. Situated in the largerhistorical context of patent law, this is a somewhat novelphenomenon--especially with regards to patent validity challenges--dueto the former separation of courts of law and courts of equity:

And because peripheral claiming came into its own with the 1870[Patent] Act, validity as we understand the term today became a realissue in patent litigation for the first time in these decades. Butbecause under the 1870 Act a patentee who wanted both an injunction anddamages had to proceed in a court of equity, virtually none of thepatent cases decided in this period were tried to a jury. Indeed, thedominance of equity in patent litigation was so complete that by1940... only 2.5% of patent suits were tried to a jury... (74)

But after the merger of law and equity in United States courts,(75) jury use in patent cases began to slowly tick upward. In the 1940sand 1950s, approximately 3.4% of patent cases were tried to a jury. (76)By the late 1970s, that figure tripled to 10%, (77) and by the late1990s, it tripled again to 32%. (78) In the past two decades, the trendhas only accelerated; juries today decide approximately 80% of patenttrials. (79)

Patent infringement is an exclusively civil offense, (80) unlikecopy-right (81) or trademark, (82) such that at least one of the partiesmust affirmatively request a jury if they wish to avoid a bench trial.(83) In practice, it is predominantly patentee-plaintiffs that areresponsible for the surge in jury use. (84) Juries are seen asdisproportionately plaintiff-friendly in patent cases, and justifiablyso: jury use is consistently associated with higher damage awards andincreased plaintiff win rates overall. (85) This has led to somepushback against jury use, whether grounded in the unique features ofpatent law (86) or arising from a more generalized suspicion of SeventhAmendment jury rights applying in complex cases. (87) Indeed, theFederal Circuit's landmark opinion over twenty years prior inMarkman v. Westview Instruments--holding that judges, not juries, have"the power and obligation to construe as a matter of law themeaning of language used in the patent claim[s]" (88)--wasinterpreted at the time as a partial embrace of such suspicion:

Today's decision also threatens to do indirectly what we have declinedto do directly, that is, create a 'complexity exception' to the SeventhAmendment for patent cases... But there is simply no reason to believethat judges are any more qualified than juries to resolve the complextechnical issues often present in patent cases... We have consistentlystressed that the same rules apply to patent cases as apply to allother civil disputes. (89)

Nevertheless, the vast majority of patent issues are stillpermitted to be submitted to juries, and, as indicated above, quiteoften are.

These three trends are, to some extent, intertwined andoverlapping. By design, many of the patent-heavy districts are alsoparticipants in the Patent Pilot Program. (90) This has led to a doubleconcentration: patent cases are preferentially filed in certaindistricts, and then preferentially (re)assigned to the dockets ofcertain judges. Chief Judge Rodney Gilstrap of the Eastern District ofTexas, a Pilot judge, is perhaps the apotheosis of this overlap; from2013 to 2018, Judge Gilstrap had more than 5,300 patent cases cross hisdocket, nearly outpacing the next four patent-heaviest judges combined.(91) Likewise, part of the reason why plaintiffs have congregated aroundcertain districts appears to be the comparative ease of reaching a jurytrial under their local rules. (92) Despite some critique and resistanceto these trends, the shape of district court patent litigation appearsunlikely to change particularly soon.

Hence, as with the USPTO, the analysis that follows will offer auseful picture of patent adjudication for the foreseeable future andshould ideally help inform efforts at reform or modification thereof.

III. METHODOLOGY

A. Data Source

This study covers all patent appeals docketed at the FederalCircuit between October 1, 2014, and September 30, 2016 (i.e., FY 2015and FY 2016), using Public Access to Court Electronic Records("PACER") as the document source. This time period offerscurrent, relevant data--particularly on post-AIA proceedings before thePTAB--while still ensuring that the overwhelming majority of cases havealready reached final appellate resolution. The direct use of PACER,moreover, allows for capture of cases that have no decision available onWestlaw, Lexis, or other secondary databases.

Of the 3,006 appeals filed at the Federal Circuit during thisperiod, 2,274 were appeals from district courts, the PTAB, or the ITC.(93) The Author read and annotated the materials for each of thosecases--direct review of the appellate docket, briefs, orders, andopinions as necessary--to most accurately determine the substantiveissues raised and final outcomes thereof. Those issues and outcomeswere, in turn, converted into data points, and compiled into a databaseon which statistical operations could be performed.

As other scholars have noted, this type of hand-coding is anextremely time-intensive task, (94) but offers significant advantagesover commercially available or machine-generated datasets. on thecommercial side, WestLaw's Lex Machina (95) platform is perhaps themost robust source available for empirics of patent cases--but does not,as of this writing, track appellate outcomes in any accessible ormanipulable way. on the machine-generated side, docket-level data, likethat found in PACER, may be mined directly, and to great effect incertain broad applications. (96) But using such surface-level data herewould eliminate the ability to distinguish appellate outcomesissue-by-issue, (97) and erroneously count irrelevant cases. (98) Evenotherwise comprehensive academic databases, such as the Federal CircuitData Project, (99) feature the same drawback; each case is treated as asingle unit, rather than a bundle of discrete claim sets and issues. Thehand-coding method presented below hence offers a uniquely direct andaccurate picture of Federal Circuit decision-making.

B. Coding Procedure

This study uses a claim-case combination as the unit of analysisfor coding. That is, for each case, the Author recorded the validityissues that were actually reviewed on appeal--as indexed by the set ofaffected patent claims. This approach attempts to best simulate thereality of decision-making: if the Federal Circuit and litigants treateda given set of claims as all rising or falling together on appeal, thenso does the dataset.

The following excerpt offers an illustrative example of briefingbefore the Federal Circuit at its most straightforward:

The district court ruled the "in response to" term recited in claims 1and 23 to be indefinite. However, the district court failed to performthe analysis required by Nautilus and this Court's opinions. The recordestablishes that the '445 patent's specification informs a person ofskill in the art as to the scope of the invention recited in claims 1and 23, including the "in response to" term. Nautilus thereforerequires reversal of the district court's ruling....The district court also ruled that the "hook inserted" term in claim 22is indefinite because it omitted a word. Here, the intrinsic evidencein the specification clarifies the "hook inserted" term's scope.Therefore, even without correcting the omission, the term meets thedefiniteness requirement. The record therefore establishes that the"hook inserted" term satisfies the Nautilus standard. (100)

Hence, this particular case provided two data points: one regardingthe definiteness of claims {1, 23} and one regarding the definiteness ofclaim {22}.

Where a set of claims was rejected or upheld on multiple grounds bythe trial forum, only the grounds that were actually reviewed by theFederal Circuit were included in the analysis. For example, in one case,the patent owner appealed the PTAB's determinations that the claimset {17, 18} was anticipated as well as obvious. (101) On appeal, theFederal Circuit affirmed on the basis of anticipation alone. (102)Therefore, that case provided only one data point--the anticipation of{17, 18}--with the explicitly unaddressed obviousness finding notincluded in the analysis.

For cases with written opinions, this method of coding was entirelystraightforward and required almost no subjective interpretation of thecase materials. More challenging, however, was the FederalCircuit's use of summary affirmances under Rule 36. (103) BecauseRule 36 affirmances do not explain the specific basis for affirming,some subjective coding decisions were required to avoid double-counting.

For example, in Enovsys LLC v. AT&T Mobility LLC, the patentowner appealed the district court's determinations that patentclaims {1, 11, 13} were anticipated as well as obvious. (104) For theultimate judgment below to stand, only one basis needed to be affirmed;claims {1, 11, 13} remain invalidated either way. To count this case astwo "affirm" data points would therefore tend to depart fromthe realities of appellate review; the Federal Circuit does not, as arule, engage in such redundant analyses. (105) In such ambiguouscirc*mstances, the Author selected the basis for affirming with the mostdeferential standard of review--in this example, anticipation--in aneffort to select the "narrowest" basis for affirmance. (106)Using written opinions as a guideline, this type of selection appears toreflect actual Federal Circuit practice. (107)

It should be emphasized, however, that this challenge was onlypresent in a minority of Rule 36 cases. More often, the set structure ofclaim bundles disputed on appeal made coding a matter ofstraightforward, if somewhat time consuming, set logic application. InIn re Clouding Corp., (108) for example, the patent owner appealed thefollowing separate determinations made by the PTAB: (1) claims {1-24} ofthe patent at issue were anticipated by International Pub. No. WO99/12098; and (2) claims {1-3, 6, 7, 13, 15, 16} were anticipated byU.S. Patent No. 5,778,389 (filed May 23, 1996). (109) The FederalCircuit affirmed under Rule 36, but the reasoning may nevertheless beparsed cleanly for quantitative purposes as follows. Because claim 4(among others) only appears in set (1) (i.e., it was not otherwiseindependently challenged on appeal), the finding for that set must havebeen reviewed and affirmed on appeal for the judgment below to stand.Moving one step further, because set (2) is a proper subset of set (1),one can assume as a matter of judicial economy that the Federal Circuitwould not superfluously review that issue as well--all the claims in set(2) are already handled by (1). So, altogether, the dataset includes Inre Clouding Corp. as exactly one data point--an anticipation affirmance.

This tracks the Federal Circuit's actual approach, asdemonstrated through the cases that do feature written opinions. Forexample:

[T]he Board's final decision found claims 1-12, 14, 15, and 17-29anticipated by U.S. Patent No. 6,496,867 to Beser ("Beser"), claims 1,2, 7-9, 12-17, 19-21, and 24-29 anticipated by U.S. Patent No.6,131,121 to Mattaway ("Mattaway"), and claims 1-15, 18-23, and 28-29anticipated by U.S. Patent No. 6,557,037 to Provino ("Provino")....After full review of the record and careful consideration, we find noerror in the Board's claims constructions or findings with respect tothe Mattaway and Provino references, which together cover all claims ofthe... patent. We do not, therefore, need to reach the merits of theBoard's decision with respect to Beser. (110)

This method--combined with the litigants' own tendency tonarrow the grounds of appeal to a handful of issues--rendered evencomplex Rule 36 cases easily codable.

After excluding cases in which there was no dispute on appeal as topatent validity (for example, cases where only the award of damages wasappealed) and cases where patent validity was disputed, but whollyunaddressed by the Federal Circuit (for example, cases where ajurisdictional defect was dispositive), there were 1,087 claim-case datapoints for empirical analysis. (111) In relevant part, each claim-casedata point included ten characteristics, ranging from case demographicsto substantive merits. (112)

Before turning to the results, a brief explanation is warrantedregarding patent technology class in particular. For each patent, theAuthor began with the USPTO's own internal class designation. (113)For example, Patent No. 7,647,460, entitled "Method and apparatusfor implementing a remote mirroring data facility without employing adedicated leased line to form the link between two remotely disposedstorage devices," is labeled by the USPTO as within class number711: "Electrical computers and digital data processing systems(memory)." (114) In turn, those class designations were condensedand simplified into six supercategories (115) for more sensiblequantitative use: (1) chemical; (2) computers and communications; (3)drugs and medical; (4) electrical and electronics; (5) mechanical; and(6) not otherwise classified. (116) The aforementioned 711 class, forexample, corresponds to the "computers and communications"supercategory, so the '460 patent was so designated.

The dataset thus offers a fairly robust picture of patent validitydecision-making from the perspective of the Federal Circuit, whileattempting to maintain mechanical objectivity in all aspects of codingto the maximum extent feasible. Part IV analyzes the results in detail,with the theoretical and doctrinal implications reserved until Part V.

IV. ANALYSIS

Sections IV.A and IV.B begin by briefly examining cross-tabulationsof the PTAB and district court data separately, noting disparities,distinctions, and outliers that arise and justifying refinements to thedata accordingly. Section IV.C then compares the two subsets side byside, including regressions on the refined data. Section IV.D addressespossible concerns and critiques with the techniques herein, inparticular the possibility of bias or error in coding andlitigant-driven selection effects.

A. PTAB Data

Starting with the highest level of generality, the raw results onappeal for patent validity findings made by the PTAB are presentedbelow:

Immediately, one sees the greater tendency for the Federal Circuitto vacate and remand instead of outright reversing the PTAB'sfindings. on further examination, it becomes clear that this is drivenin substantial part by the limitations of appellate review with respectto Article I tribunals (such as the PTAB) as contrasted with Article IIIcourts. That is, per SEC v. Chenery Corp. and its progeny, "anadministrative order cannot be upheld unless the grounds upon which theagency acted in exercising its powers were those upon which its actionscan be sustained." (117) Put differently, on appeal from a districtcourt, if the "decision below is correct, it must be affirmed,although the lower court relied upon a wrong ground or gave a wrongreason," (118) but if an agency finding "is not sustainable onthe administrative record made, then the... decision must be vacated andthe matter remanded... for further consideration." (119)

The Federal Circuit has carried this doctrine through to its reviewof PTAB decisions, highlighting in particular the need for adequate"explanation" from PTAB decision-makers. (120)

The following case, coded for the data set, provides an example ofthe doctrine in action:

Our precedent demands more than what the Board's opinion providedhere... [T]he Board must provide rationale for its findings tofacilitate our review... On remand, the Board must provide additionalfact findings and explanations for its findings relating to theanticipation and obviousness determinations as to claims 1-3, 5, 7-10,and 12-14 of the '960 patent. We take no position on whether the priorart, taken as a whole, anticipates or renders obvious the disputedclaims. (121)

Because PTAB findings that fail to satisfy these requirements arevacated and remanded, (122) whereas comparable district court findingswould not be, those twenty-six PTAB data points are hereafter pulledfrom the set to maximize comparability with district court outcomeslater on.

Even under a particularly strict interpretive standard--onlycounting opinions that explicitly invoke the "explanation" or"rationale" requirement by name and citation--this accountsfor a substantial portion of the PTAB's vacaturs and remands,leading to the following refinement:

Mathematically, this relies only on the assumption that thosevacaturs and remands would have otherwise been affirmances and reversalsat the same rate as the PTAB's other findings. Their removal thusfacilitates a better comparison to the district court results withoutany meaningful loss of clarity.

Going deeper, it's possible to distinguish between theaffirmance rates for different kinds of validity issues--anticipation,obviousness, and so on. Because the standards of review for these issuesare not identical, this offers an explanation for some of the variationin results on appeal. Specifically, subject-matter eligibility is aquestion of law reviewed de novo, (123) as is indefiniteness. (124)"[U]ltimate determinations of obviousness" are likewisequestions of law, reviewed "de novo"--but crucial"underlying factual findings" made by the PTAB are"re-view[ed] for substantial evidence." (125) Enablement has asimilar, mixed standard of review. (126) Anticipation, on the otherhand, is simply "a question of fact reviewed for substantialevidence." (127) Likewise, "[w]hether a claimed invention issupported by an adequate written description... is a question offact," reviewed for substantial evidence. (128) Doctrinally,"[s]ubstantial evidence review asks 'whether a reasonable factfinder could have arrived at the agency's decision'"through a holistic analysis of the entire record below, "takinginto account evidence that both justifies and detracts from anagency's decision." (129)

Differentiating between these categories yields the followingresults:

One quickly observes the overwhelming dominance of obviousness andanticipation findings. To avoid reading too much into the small samplesin certain categories--and to better facilitate comparisons--the tablebelow collapses the results into question type alone:

The result is surprising, at least at face value. The moredeferential the nominal standard of review, the lower the affirmancerate, particularly when comparing against questions of fact. Two detailscan, at least in part, explain these results. First and foremost: thePTAB's review of subject-matter eligibility is relativelycirc*mscribed. Section 101 challenges are not available in inter partesreexam, ex parte reexam, or inter partes review. Rather, the bulk ofsubject-matter eligibility PTAB determinations in this data set comefrom "covered business method" (130) (CBM) challenges. Theseare patents that specifically "claim[] a method or correspondingapparatus for performing data processing or other operations used in thepractice, administration, or management of a financial product orservice"--not "technological" inventions. (131) This maypreclude the PTAB from receiving more difficult and variedsubject-matter eligibility cases. That being said, the effect this hason the PTAB's overall affirmance rate as calculated at the outsetis truly negligible--less than one-tenth of a percentage point--due tothe very small number of such data points in the set.

Second: regardless of the particular validity issue, there areoften predicate claim constructions in dispute with their own standardsof review. Construction refers to the process of interpreting the scopeand meaning of a patent's claims, a distinct and explicit processin patent disputes that may proceed term-by-term. For example:

The parties dispute the construction of the term "pharmaceuticallyacceptable" as used in claims 1-5 of the '467 Patent, claims 1-4 of the'976 Patent, and claim 10 of the '427 Patent. "Pharmaceuticallyacceptable" is used to modify both the claimed dutasteride solvates aswell as the claimed carriers of the dutasteride... GSK argues that thespecification expressly defines "pharmaceutically acceptable" as "notdeleterious to the recipient thereof when administered as apharmaceutical." Defendants argue that "suitable for use in a finisheddrug product to be administered to a patient" is consistent with theordinary meaning of the term and that ordinary meaning applies here.(132)

Naturally, this construction process plays a "critical role innearly every patent case" and is "central to the evaluationof... validity." (133) Where there is "no dispute... aboutfindings or evidence of facts extrinsic to the patent"--in otherwords, when interpretation relies only on the text, prosecution history,and prior art of the patent itself--claim construction is solely amatter of law, and the Federal Circuit "conduct[s] a de novo reviewof the Board's... interpretation of the claim language." (134)Where extrinsic evidence is in dispute, any "subsidiary factfindings" by the PTAB are, again, "review[ed] for substantialevidence." (135)

When the Federal Circuit leaves claim constructions undisturbed, itis reasonable to primarily examine the actual validity issue appealed.But, if the Federal Circuit alters a claim construction on appeal, it isno longer reviewing the PTAB's validity findings, in truth. TheFederal Circuit either vacates and remands for new validity findingsunder the new construction or uses the existing record below to make itsown. (136)

Hence, removing data points where the claims were incorrectlyconstrued further hones the focus of the analysis on trial-leveldecision-making. only examining cases with accurate claim constructionwill further aid in maximizing comparability between the PTAB anddistrict court results. Until 2018, the PTAB and district courts useddifferent standards for claim construction, depending on the type ofproceeding, (137) and this adjustment minimizes the effect of thatdifference on the ultimate appellate results.

The following table thus reflects the PTAB data after removing datapoints where an underlying claim construction was modified by theFederal Circuit:

Here, one sees a more sensible result when comparing betweencategories--mixed or purely fact-based validity findings are affirmedmore often than those with de novo standards of review. The differencebetween mixed and purely factual findings is marginal.

Next, the table below examines the makeup of the PTAB data in termsof underlying patent technology class:

Unsurprisingly, cases involving computer and communicationstechnology compose the lion's share of PTAB validity findingsreviewed by the Federal Circuit. In terms of results on appeal, thestandard deviation between technological categories is quiteinsubstantial: only 1.4 percentage points (excluding the catchall"Other"). Put differently, there does not appear to be anyparticularly strong pattern or outlier that would skew the PTAB'soverall affirmance rate.

Finally, it's possible to examine the different types ofproceedings that take place before the PTAB. In addition to the newinter partes and CBM reviews authorized by the AIA, this datasetincludes appeals from the more longstanding inter partes and ex partereexaminations, (138) interference proceedings, and patent applications:

Note that appellate standards of review do not inherently differbetween these proceedings. Factfinding in an ex parte reexamination, forexample, is reviewed for "substantial evidence," (139) just asin inter partes (140) or CBM review. (141) With that equal footing inmind, there does not appear to be a particularly strong affirmancepattern between these categories; post-AIA proceedings are affirmed atrough parity with the more established or eliminated PTAB proceedings.(142) Accordingly, there is no distinction made between proceeding typesin the direct comparison to district court appeals in Section IV.C.First, however, a comparable examination of the district court dataitself is required.

B. District Court Data

As with the PTAB, the overall raw affirmance rate for validityfindings made by the district courts provides a starting point:

Although a full comparison with the PTAB results is reserved untilthe end, one can immediately see the much lower tendency for the FederalCircuit to vacate and remand district court findings involving validity,due to the lack of Chenery-type constraints.

The district courts' overall rate does, however, conceal onecritical distinction of its own: jury-made findings. It is worthemphasizing that, even in highly technical patent cases, juryfactfinding is a "black box" afforded an extraordinary levelof deference on appellate review. (143) Indeed, although juryfact-finding on questions of patent validity is quite rare--onlyforty-one instances in this data set (144)--those findings were affirmedat a rate notably higher than the remainder of the district court data(90.2%). Though the effect on overall rate is small, to obtain a moreaccurate picture of district court judge findings, those data points arehereafter removed from the analysis. Likewise, as with the PTAB, casesin which the Federal Circuit modified the court's predicate claimconstructions below are hereinafter removed as well.

As explained in earlier sections, due to a combination of thePatent Pilot Program and venue rules, patent litigation is highlyconcentrated among a handful of districts and judges. Conversely, thishas caused a relative dearth of patent cases elsewhere. Nevertheless,the data shows relative consistency across the country in terms ofappellate outcomes--whether viewed by judge or district. Starting withdistricts, over the past five years, the Eastern District of Texas, theDistrict of Delaware, the Central and Northern Districts of California,and the District of New Jersey collectively represented 72% of alldistrict court patent litigation in the country. (145) Separating outthe appeals from those top-five districts yields the following results:

Turning to the judges, only twenty-five district court judges inthe country saw more than one hundred patent cases on their docketduring that same five-year period. Separating these more experiencedjudges out from the rest of the set, one sees a different pattern:

That is, patent-heavy districts fare marginally better in terms ofappellate outcomes, whereas patent-heavy judges fare marginally worse.Neither pattern is particularly strong--only a few percentage points(146)--and indeed, using both metrics simultaneously results in a mereone percentage-point differential (147):

This lack of meaningful variation comports with the data availablethus far from the Federal Judicial Center's own review of thePatent Pilot Program. (148) Accordingly, the data from all districtcourt judges are hereinafter combined, with the understanding that thesystem may be fairly viewed as a whole for the time being.

Next, it is again possible to distinguish between the affirmancerates for different kinds of validity issues. In contrast to the PTAB,factual findings made by district court judges are reviewed for"clear error" rather than substantial evidence. (149)Doctrinally, this is a less forgiving standard of review thansubstantial evidence; (150) nevertheless, it still requires affirmanceunless, "despite some supporting evidence, [the court of appealsis] left with the definite and firm conviction that a mistake has beenmade." (151) Otherwise, review of validity issues is governed bythe same nominal law-versus-fact breakdown as the PTAB, including denovo review of questions of law: (152)

Findings involving at least some questions of law clearly dominatethe analysis, particularly subject-matter eligibility. As with the PTABdata, to avoid reading too much into the small samples in certaincategories, the results are collapsed into question type alone:

One sees that the counterintuitive pattern from earlier hasappeared again: the more deferential the nominal standard of review, thelower the affirmance rate. Here, however, the pattern persists evenafter the refinements to the data. That is, mixed and purely law-basedfindings from the district courts are affirmed reliably more often thanfact-based findings. That gap, moreover, is clearly and predominantlydriven by routine affirmance of subject-matter eligibility findings.

Finally, the results by underlying patent technology class arepresented below: (153)

Subdivided this finely, the number of observations in somecategories becomes rather small and easy to over-interpret.Nevertheless, two patterns may be fairly noted. First, cases involvingcomputers and communications technology again form the bulk of validityfindings that are reviewed by the Federal Circuit. Second, thedifference in affirmance rates is still quite small, suggesting a lackof meaningful variation between technology classes for purposes of thisanalysis. With all this in mind, a more direct comparison between thetwo trial-level fora follows.

C. Drawing Comparisons

Recombining the data, it's worth reiterating the refinementsfrom the preceding sections to maximize comparability: Chenery-typenon-affirmances are removed for the PTAB, jury findings are removed forthe district courts, and instances where claim constructions weremodified on appeal are removed for both. This yields the followingoverall results:

Three tendencies emerge from this picture: (1) the PTAB is affirmednotably more often than district courts on validity issues; (154) (2)district court decisions tend to be reversed outright, rather thanvacated or remanded; and (3) PTAB decisions, on the other hand,experience both types of non-affirmance in roughly equal measure.

Naturally, this broad view overlooks the aforementioned distinctionin standards of review. Tabulating the data in that fashion revealsfurther nuance:

One sees that the district courts' findings of law areactually affirmed at parity with the PTAB; the divergence in overallaffirmance rates is driven almost entirely by the more fact-basedcategories. (155) This dovetails somewhat with the district courts'backwards relationship between standards of review and actual affirmancerate, displayed again above.

As noted in the preceding sections, cases involving computers andcommunications technology dominate the dataset. Cases involving drugsand medical technologies are the second most common, though by a highermargin in district courts than at the PTAB. Across categories, one seesthat affirmance rates are relatively steady, with the PTAB holding aunilateral advantage:

The PTAB's affirmance rate is higher in every singletechnology category. Although the affirmance gap ranges from as littleas 3.7% (electrical and electronics) to as much as 10.7% (mechanical),it is difficult to read much into these granular distinctions given thesmall number of district court observations per category. Suffice tosay, no single area of technological subject matter is responsible forthe PTAB's over-all edge in terms of appellate outcomes.

Finally, one additional trend became clear when directly comparingfora: the disposition of claim-invalidating findings versusclaimupholding findings.

There is a distinct affirmance gap--approximately thirteenpercentage points--favoring PTAB findings that invalidate patent claimsover PTAB findings that uphold them. This gap does not exist betweencomparable district court appeals; to wit, there is a marginal gap inthe reverse direction for district court findings overall. Theadditional tabulations for each tribunal must be interpreted delicately,particularly with respect to the outright magnitude of affirmance rates,given the low number of observations per cell. Nevertheless, for thePTAB, one sees that the invalid-upheld gap is not driven by a particularcategory of findings; it persists across the board at a statisticallysignificant level, regardless of question type. For district courts, onthe other hand, the gap varies in magnitude and direction--and does notrise to the level of statistical significance within or acrosscategories.

Before turning to a more theoretical contextualization of thesetrends, a basic logistic regression model (156) is presented below toexamine the relative importance of different factors on appellateoutcome. Recall that the dataset used to produce the model is not asample, but rather the actual population of Federal Circuitdecision-making for FY 2015-2016. That is, the numbers below reflect theactual degree of variation and relative mathematical importance in theyears surveyed--not an "estimate" in the traditionalstatistical sense of attempting to extrapolate from a sample:

The unadjusted model evaluates each individual factor'sinfluence on the likelihood of being affirmed, without considering otherfactors. For example, if considered alone, PTAB findings are 1.85 timesmore likely than district court findings to be affirmed. The 95%confidence interval for this 1.85 figure is, in turn, (1.21, 2.83), witha Wald Chi-Square p-value of 0.0047.

The adjusted model, on the other hand, evaluates all factors'simultaneous influences on the likelihood of being affirmed: thetribunal below, the question type, the disposition of the underlyingclaim, the case year, and the patent-at-issue's technology class.For example, with all factors considered, PTAB findings are still 1.85times more likely to be affirmed than district court findings. But the95% confidence interval grows slightly to (1.01, 3.39), with acorresponding Wald Chi-Square p-value of 0.0462--thus remainingstatistically significant.

Overall then, observe the continued importance of the trial-leveltribunal for appellate outcomes, even after controlling for as manysalient case characteristics as feasible. This strongly suggests thatsomething about the PTAB itself, as compared to district courts, isincreasing its affirmance rate on appeal.

The remaining variables behave as one would expect based on thetabulations and cross-sections presented supra. Questions of law have apositive influence on affirmance rate; questions of fact have theopposite effect. Neither, however, appears statisticallysignificant--this is likely driven by the district courts' outsizedshare of law findings and strong standard-of-review pattern on the onehand, and the PTAB's outsize share of factual findings andcomparatively weak pattern on the other. The differences between thetechnology categories are again quite low, reinforcing the lack of ameaningful appellate affirmance pattern thereof. Finally, upholding theclaim below has a highly significant negative influence on appellateresult.

The strongest and most meaningful results of Part IV'sanalysis may be accordingly summarized as follows. First, the FederalCircuit affirms findings from the PTAB reliably more often than districtcourts (though less often than juries). This is particularly true whenthe findings are wholly or partially factual, as opposed to purequestions of law. Second, PTAB findings that invalidate patent claimsare affirmed substantially more often than findings that uphold patentclaims; no similar pattern exists for district court findings. Third,the underlying technological subject matter of the patent at issue doesnot appear to perceptibly influence results on appeal.

In Part V, these quantitative results are contextualized as part oflarger trends in patent jurisprudence. But first, Section IV.D addressesthe most likely concerns a reader may have with the quantitative methodpresented thus far: subjectivity and selection effects.

D. Critiques and Concerns

The act of grouping certain claims as rising and falling togethernecessarily introduced a small degree of subjective decision-making intothe coding process. Accordingly, robust intercoder reliability tests(157) were performed to ensure that there was no systemic bias or errorin the Author's own procedure. Three other researchers--all formerFederal Circuit law clerks--were collectively assigned a random 15%sample of the cases in this data set to independently re-code. Agreementwith the master dataset overall, measured as a raw percentage ofmatching data cells, exceeded 99%. Agreement remained above 99% evenwhen restricted to the more substantive data cells: the holding below,the result on appeal, and whether a claim construction was modified onappeal. As expected, divergence was at its highest with respect toclaim-grouping decisions themselves, but the other researchers'decisions still matched the master dataset more than 95% of the time.Altogether, these results indicate that no pervasive errors have skewedthe results, such that the data presented above is an accuratequantitative representation of the cases coded.

Separate from accuracy, a reasonable critique of the data presentedthus far would be the possibility of selection effects as explainingsome, or all, of the patterns above. This critique has essentially twodimensions to it: (1) the decision litigants make in choosing betweentrial fora; and (2) the decision litigants make in whether or not toappeal. To the extent the PTAB receives a meaningfully differentproportion of patent validity challenges than the district courts--or ameaningfully different subset of the PTAB's cases actually progressto an appeal at the Federal Circuit--it could impact this data.

Observe at the outset, however, that many forum-selection orappeal-decision differences would not themselves affect the overallaffirmance rate for a given tribunal, which is the chief concern of thisanalysis. For example, imagine that the district courts receive adisproportionate share of litigation where the patent-at-issue isgenuinely valid, whereas the PTAB receives a disproportionately invalidshare. If the goal were to directly compare the patent invalidation ratebetween tribunals, this phenomenon would clearly dominate the results.But it would not contribute to a PTAB affirmance advantage on appeal,unless one assumes that valid-patent cases are intrinsically morecomplex or difficult to adjudicate than invalid-patent cases. Likewise,imagine that litigants are more likely to file an appeal in casesinvolving computer and communications technology than cases involvingelectrical and electronic inventions. Even if the district courts seemto receive an outsize portion of the former (and the PTAB, the latter),(158) affirmance rates will only be affected if it is more difficult, onaverage, to reach the correct result (or, strictly speaking, the resultthat the Federal Circuit considers correct) in one over the other.

In other words, the principal concern for purposes of this analysisis whether the PTAB receives a disproportionately easy share of cases,roughly speaking. The remainder of this section presents severalquantitative indices and overall heuristics that suggest such adifference is unlikely to exist at a level high enough to influence theresults of this analysis.

First, consider how the PTAB's share of patent validitychallenges has changed over time. The years surveyed in this study, FY2015 and FY 2016, represent a paradigm shift at the Federal Circuit:(159)

As shown in the chart above, appeals from the PTAB to the FederalCircuit have surged upward, surpassing appeals from district courts forthe first time in FY 2016. With this dataset, it is possible to comparethe results on appeal for each tribunal, before and after that shift:

Both tribunals saw a marked increase in affirmance in FY 2016compared to FY 2015, and a slight convergence in their overallaffirmance gap. This renders the "easier cases" hypothesissomewhat suspect; as the PTAB siphons away more and more cases fromdistrict courts, one would expect the affirmance gap between the two tobe far less stable. Put differently: there are only so many "easycases" out there, yet the PTAB continues to outperform as its sharegrows. For this same reason, if selection effects were at play, onemight expect the year variable to meaningfully impact affirmance rate inthe more complete regression analysis. Instead, it was among the weakestoverall, after controlling for other factors--and the trial-leveltribunal remains quite significant even after its inclusion.

Second, consider the Federal Circuit's varying use ofdissents. If nothing else, a dissenting opinion on appeal suggests thatreasonable minds may have differed on an issue's appropriateresolution. Or, in other words, the issue was not so clear-cut andsimple as to necessarily command unanimity. The rate of dissents onappeal between the PTAB and district courts' validity decisions arecompared below: (160)

One indeed observes a difference, but quite marginal--less than twopercentage points. (161) To put this in terms of raw numbers, just fouror five fewer dissents in district court appeals--over two full fiscalyears--would equalize the data. The lack of a stronger pattern hereagain undercuts the "easier cases" suggestion.

Third, consider what happens if the PTAB data is arranged not bypre- or post-AIA procedures (162) but instead by which proceduresactually do put the PTAB in direct competition with district courts.Inter partes review, both types of reexamination, and CBM review allserve to invalidate imprudently-granted patents, a function that is alsofulfilled by district courts' declaratory judgment power or as anaffirmative defense in infringement litigation. Patent applications andinterference proceedings, on the other hand, have no such analog orsubstitute.

One sees that the affirmance rate is slightly lower for proceedingswith no other choice of forum, but just barely. (163) And indeed, whencomparing the most meaningful and representative subcategoriesdirectly--inter partes reviews and patent applications--the differenceis even more marginal. Altogether, it seems considerably unlikely thatthe PTAB's overall affirmance advantage is solely the result of itreceiving easier cases compared to district courts.

Setting quantitative indices aside, it's worth emphasizing theunlikeliness of selection effects as a matter of heuristics. First,recall that district courts adjudicate infringement as well as validitydisputes--typically as part of the same case--whereas the PTAB can onlyadjudicate validity. In cases where the district court did addressinfringement and validity, the losing party has a much strongerincentive to appeal. If the plaintiff-patentee lost, a victory on appealmeans not only the revitalization of its patent, but also potentiallyinfringement remedies. If the defendant-infringer lost, a victory onappeal means not only avoiding past damages, but also carte blanche toopenly infringe the (now-dead) patent moving forward. Compoundingmatters, fee-shifting may also be on the line in district courtproceedings, whereas it is generally unavailable in PTAB disputes. (164)Altogether then, one would expect parties to be more likely to appealmarginal district court cases rather than marginal PTAB cases--thepotential reward (or averted loss) is higher. These cases, in turn,would artificially inflate the district court affirmance rate on appeal;the Federal Circuit would be receiving disproportionatelynon-meritorious appeals from district courts.

Moreover, observe that PTAB litigation is, by design, far cheaperand faster than district court litigation. (165) On average then, onemight expect the parties that reach final judgment in district court tohave deeper pockets and more long-term strategies, and thus be more ableto bear the time and costs of challenging the judgment on subsequentappeal. Again, all else being equal, this should drive the districtcourts' affirmance rate up, if anything; these litigants can affordto appeal more broadly. Put differently, poorer litigants--or thosefacing significant time pressures--are occasionally forced to leave evenmeritorious appeals on the table.

Accordingly, the results outlined in Section IV.C likely reflectmeaningful patterns in validity appeals and are not merely an artifactof coding bias or underlying selection effects. Part V proceeds to fitthese patterns into a larger picture of contemporary patent law.

V. MAKING SENSE OF THE DATA: ANTI-EXCEPTIONALISM ANDCONTEXT-SPECIFIC DEFERENCE

Because the dataset at the heart of this analysis is based onappeals, there is a fundamental question that needs to be examined: arethe differences in appellate outcome the result of qualitativelydifferent adjudication below or the consequence of unevenly applieddeference on review? By teasing the two apart, it's possible to seea glimpse of the Federal Circuit's evolving relationships--andperhaps a more nuanced role for specialization and expertise in thepatent sphere overall.

As a matter of first principles, recall that factual findings madeby the PTAB are subject to the "substantial evidence" standardof review, whereas district court judge findings are subject to"clear error" review. (166) The latter standard is, nominally,a less forgiving one. (167) Hence, this difference likely plays at leastsome role in explaining the affirmance gap between the PTAB and districtcourts, particularly since the gap tends to grow as the findings atissue become more based in fact. Compelling scholarship, however,suggests that the intrinsic difference between the two standards ofreview is marginal elsewhere in the law as a matter of practice. (168)To wit, when addressing the Federal Circuit's standards of reviewvis-a-vis the USPTO, the Supreme Court itself opined:

The [substantial evidence] standard, as we have said, is somewhat lessstrict than the [clear error] standard. But the difference is a subtleone--so fine that (apart from the present case) we have failed touncover a single instance in which a reviewing court conceded that useof one standard rather than the other would in fact have produced adifferent outcome... The difficulty of finding such a case may in partreflect the basic similarity of the reviewing task, which requiresjudges to apply logic and experience to an evidentiary record, whetherthat record was made in a court or by an agency. It may in part reflectthe difficulty of attempting to capture in a form of words intangiblefactors such as the judicial confidence in the fairness of thefactfinding process. (169)

It seems highly unlikely that the words "substantialevidence" versus "clear error" are, therefore,sufficiently compelling as to create a difference in appellate outcomes.Put differently, even if the stated deference regime is influencingthese results, the strength of that regime is ultimately predicated onthe relationship of judicial confidence that motivates its applicationby the Federal Circuit in the first place. An examination of thatrelationship is therefore appropriate.

Reading the case law more broadly suggests that judicial confidencein the PTAB has, to say the least, not been particularly high. On thecontrary, the Federal Circuit's relationship with the PTAB to dateappears almost universally exacting. To start, major Federal Circuitdecisions have generally served to cabin the USPTO's authority, notexpand it. For example, the Federal Circuit has emphatically rejectedthe application of Chevron deference to the USPTO's interpretationof substantive patent law (170) and, occasionally, even the rulesgoverning its own administrative proceedings. (171) The otheradministrative apparatuses in the Federal Circuit's purviewreceive, by stark contrast, full-throated deference (172)--including, inat least one instance, the Trademark Trial and Appeal Board within theUSPTO. (173) Indeed, at least a handful of Federal Circuit judges haveexpressed support for equalizing the PTAB and district court standardsof review outright:

Merck now urges this court to sit en banc to decide whether applicationof a more searching standard of review--clear error--is required forappeals from inter partes review proceedings... under the [AIA]... Iagree that application of the substantial evidence standard of reviewis seemingly inconsistent with the purpose and content of the AIA. Thiscourt is bound by binding Supreme Court precedent... and this court'sown... to apply the substantial evidence standard of review to factualfindings by the Board, however. Because Congress failed to expresslychange the standard of review employed by this court in reviewing Boarddecisions when it created... the AIA, we are not free to do so now.(174)

The Federal Circuit as a whole has likewise tended to maximizereviewability of PTAB proceedings, (175) while narrowing theirterritory, (176) preclusive effect, (177) and even cost-recoveringcapabilities. (178)

In more routine cases, as noted in Part III, the Federal Circuitoften rejects--sharply--the PTAB's final written decisions asinsufficiently rigorous or unclear:

The Board's procedural obligations are not satisfied merely because aparticular fact might be found somewhere amidst the evidence submittedby the parties... [However, w]e will not decide whether the Boardviolated [appellant]'s procedural rights. To make that decision, wewould need to be able to determine what evidence the Board relied on tosupport its implicit factual findings, how the Board interpreted thatevidence, and what inferences the Board drew from it. The Board'sopinion does not sufficiently permit such determinations. (179)

This tension between the Federal Circuit and the PTAB ishighlighted even further by the timbre of Circuit judges writing morefreely in dissent:

The current practice of assigning the same PTAB panel to both instituteand conduct an inter partes review is not only contrary to the statute,but has the taint of prejudgment... It is our judicial obligation toensure agency compliance with statutory text and purpose. The departureby the PTo is not only contrary to the statute, but has devastatingconsequences for the public confidence in post-grant proceedings andthe patent system as a whole. The nation's economic health depends onpublic confidence in an unbiased and balanced patent system. (180)

Or, of course, in public statements off the bench. (181)

In summation, the Federal Circuit does not appear to be affordingthe PTAB especially lenient consideration. And the nominal deferenceregime does not otherwise appear sufficiently strong to force adifferential in the face of resistance or, at best, equivocation.Accordingly, it seems highly likely that the PTAB's affirmanceadvantage over the district courts relates--at least in some part--tosubstantively different decision-making. The PTAB's affirmanceadvantage does grow considerably as the findings at issue become morefact-based, but this tracks the nature of the PTAB itself.Administrative patent judges are required by statute to possess"competent... scientific ability," (182) and in practice"[a]ll of [them] have specialized technical degrees... andexperience" (183) that are brought to bear on their specific caseassignments. (184) Federal district court judges, despite many virtues,generally do not have any background in the sciences at all, (185) letalone one relevant to the particular patent cases on their dockets.Despite the fairly widespread criticism of PTAB decision-making andpost-grant procedures, the most straightforward conclusion is that itsstructure of expertise has, on the margins, aided decision-making on thethorny scientific questions endemic to patent disputes.

But the data suggests that expertise (and deference informed bythat expertise) is not the only dynamic in play. Recall the stark gap inthe PTAB's claim-invalidating affirmance rate compared to itsclaim-upholding affirmance rate. (186) Do the PTAB's administrativepatent judges only leverage their scientific expertise when strikingdown junk patents, and then set it aside when reviewing solid ones?Likewise, consider the relative insignificance of technology categories:the PTAB showed generally steady affirmance rates between categories(187) and maintained its edge over district courts across the board.(188) All else being equal, one might expect the PTAB's expertiseto be more salient in adjudicating, say, the merits of chemical orelectrical patents instead of purely mechanically-oriented ones. (189)On the flip side, why aren't the de facto specialized districtcourts and judges reaping similar benefits? And why, of alladjudicators, do layperson juries perform the best on appeal?

These additional patterns suggest that the Federal Circuit may beinternalizing message streams that it has been receiving for some time.First and foremost, a message from the Supreme Court: patentexceptionalism needs to be reined in. Scholars generally fit the SupremeCourt's frequent reversals (190) of Federal Circuit law into apattern of anti-exceptionalism. (191) That is, the Supreme Court isinstructing the Federal Circuit to treat patent law, broadly speaking,more like any other species of law where possible. (192) The FederalCircuit's historically non-deferential relationship to thePTAB--"act[ing] like the head of an agency reining in waywardadministrative law judges" (193)--has been, in its own way, a formof exceptionalism. (194) The data presented herein suggest that perhapswe are experiencing the first cautious steps towards anti-exceptionalisminstead: a degree of cautious deference, applied narrowly to theexpertise-driven factual findings of the PTAB.

Layperson jurors, of course, not only lack relevant technologicalsubject matter expertise--they lack experience with the entire legalsystem, let alone patent law. And indeed, in previous years, the FederalCircuit was criticized for seemingly having too little faith injurors' abilities. (195) Nevertheless, the data presented hereinsuggest that their findings now hold a distinct edge on appeal. (196)This too is arguably a form of emergent anti-exceptionalism; despite theunique intricacies and complexities of patent cases, the Federal Circuitis treating jury-made findings with the same level of extra care anddeference that they are afforded in other areas of law. (197) Whateverthe merits of a patent exception to the Seventh Amendment jury guaranteebased on perceived case complexity, (198) the Federal Circuit does notappear so persuaded.

Turning to the district courts generally, recall the considerablesimilarity in appellate results between patent-heavy and more typicaljudges and districts. (199) On the one hand, these adjudicators bydefinition have greater expertise and familiarity with patent law thantheir colleagues. On the other hand, unlike in the administrative lawspace, there is no legal framework for actually recognizing thatcultivated expertise. (200) A judge is a judge is a judge. The lack ofspecial deference afforded to these heavy-hitter judges is thuspotentially anti-exceptionalism at work again. That is, the FederalCircuit is resisting the tendency to create a patent-specificstratification of district courts and judges. Instead, it isinternalizing more general legal principles that tend to reject notionsof inter-judge variation.

A second message stream comes from Congress instead: bad patentsneed to be culled. Despite ex post grumblings and halting attempts atmodification, (201) the AIA remains the most recent comprehensivelegislative intervention in the patent space. And the driving mission ofthat intervention--at least with respect to post-grant proceedings--wasperhaps as unambiguous as congressional intent can be:

Congress has not enacted comprehensive patent law reform in nearly 60years. The object of the patent law today must remain true to theconstitutional command, but its form needs to change, both to correctflaws in the system that have become unbearable, and to accommodatechanges in the economy and the litigation practices in the patentrealm... The decisions reflect a growing sense that questionablepatents are too easily obtained and are too difficult to challenge...[T]he Committee's attention [is] on the value of... improving patentquality and providing a more efficient system for challenging patentsthat should not have issued... (202)

This idea--that low-quality patents are choking Americaninnovation--pervades the AIA's legislative history and context.(203) Given the Federal Circuit's near-exclusive dominion overpatent law, in addition to its more agency-like stance, it may havenaturally internalized this idea as well. (204) In the Circuit'sown words: "Congress... saw powerful reasons to utilize theexperience of the PTO for an important public purpose--to correct theagency's own errors in issuing patents in the first place."(205) Accordingly, there may be a form of context-mediated deference atplay: when the PTAB determines that a patent needs to go, it ischanneling those legislative intentions, and the Federal Circuit ishence less inclined on the margins to reverse course. District courts,on the other hand, were left untouched by the AIA--creating nocomparable pull, and no comparable differential.

Patent practitioners and policymakers may have their own, lessabstract, takeaways from the overall empirical analysis presentedherein. The fears of litigation concentration appear overblown in somerespects, given marginal inter-district and inter-judge variation.Instead, the increased reliance on juries--enabled in part by certaindistricts and judges more than others--may be far more salient. From aprocedural standpoint, if jury findings are being affirmed solely due toblack-box deference, there may come a point where effective appellatereview is essentially circumvented. From an outcome-focused standpoint,one may rightly question whether deference to juries--even now--ismerely papering over weaker decision-making. Meanwhile, on theadministrative side, the PTAB does not appear to be misbehaving at allon substantive issues. Rather, it is operating well within the boundsarticulated to it by the Federal Circuit. Again, however, sharpercritics may question whether the Federal Circuit is articulating thecorrect bounds in the first place, particularly given the constraintsand messaging--real or perceived--that it faces from the Supreme Courtand Congress.

VI. CONCLUSION

After coding and analyzing two full years of Federal Circuitdecision-making, the results paint a surprising picture of itsrelationship to the PTAB, to district courts, and to juries. ThePTAB's expertise--whether putative or real--is cautiously rewardedby a high affirmance rate advantage on fact-specific issues. Districtcourt specialization, on the other hand, seems to carry little weight inany category. Instead, it is layperson juries that appear to garner themost appellate respect. What unifies these trends is, perhaps, anembrace by the Federal Circuit of the anti-exceptionalist messaging fromacademia and the Supreme Court. Simultaneously, the Federal Circuit maybe internalizing more policy-oriented messaging from Congress,expressing a particular willingness to affirm the PTAB when it thins thepatent herd. A reexamination of comparable data in the future would shedlight on the durability of these patterns. Absent changes to theunderlying shape of the patent landscape, however, one might expect themto continue for some time.

Matthew G. Sipe (*)

(*) Frank H. Marks Visiting Associate Professor, George WashingtonUniversity Law School; J.D., Yale Law School; B.A., University ofVirginia. With sincere thanks to Professors Michael Abramowicz, JonasAnderson, Andrew Bradt, Daniel Brean, Jeanne Fromer, Dmitry Karshtedt,Mark Lemley, sarah Rajec, Ted sichelman, and Liza Vertinsky for theircritique, inspiration, and support. A project of this scope and scalewould not have been possible without the generous assistance andresources provided by the supreme court Fellows Program, including inparticular the guidance and feedback of the Federal Judicial Center andthe Administrative Office of the U.S. Courts. Additional thanks to themembers of the Judicial Conference and Supreme Court Fellows Commissionwho provided invaluable feedback during draft and proposal stages.Finally, I am indebted to Shayon Ghosh, Patrick Holvey, Taylor King,Guohui Pan, and Ryan Watzel for their additional assistance andceaseless encouragement.

(1.) Parke-Davis & Co. v. H.K. Mulford Co., 189 F. 95, 115(S.D.N.Y. 1911).

(2.) See, e.g., Judicial Panel Discussions on Science and the Law,25 CONN. L. REV. 1127, 1145 (1993) (statement of Judge Covello, U.S.District Judge for the District of Connecticut) ("Honest to God, Idon't see how you could try a patent matter to a jury.... It'sfactually so complicated."); Case for Special Juries in ComplexCivil Litigation, 89 YALE L.J. 1155, 1155, 1158 n.18 (1980) (quotingWarren E. Burger, The Use of Lay Jurors in Complicated Cases, Remarks tothe Conference of State Chief Justices 3-5 (Aug. 7, 1979)).

(3.) See, e.g., Federal Judges, Academics, and PractitionersConsider Patent Jury Trials at Engelberg Center and Civil Jury ProjectConference, NYU L. NEWS (Oct. 18, 2016),http://www.law.nyu.edu/news/trial-by-jury-of-patent-cases-engelberg-center-civil-jury-project-judges-practitioners [https://perma.cc/7VF4-HQHL]("[J]uries can sort out even the most complex issues when given theproper tools, and... almost always arrive at conclusions which arerational, fair, and... justified....").

(4.) Parke-Davis & Co., 189 F. at 115 (providing the openingquotation above); see generally Scott Brewer, Scientific ExpertTestimony and Intellectual Due Process, 107 YALE L.J. 1535 (1998).

(5.) See, e.g., Richard Revesz, Specialized Courts and theAdministrative Lawmaking System, 138 U. PA. L. REV. 1111, 1146 (1990)("[S]pecialized courts tend to [control administrative action] lesseffectively than generalist courts because they are more likely toexhibit systemic biases....").

(6.) Peter Pitts, 'Patent Death Squads' vs. Innovation,WALL ST. J. (June 10, 2015, 7:23 PM),http://www.wsj.com/articles/patent-death-squads-vs-innovation-1433978591(last visited May 10, 2019) (quoting remarks from then-Chief Judge ofthe Federal Circuit Randall R. Rader).

(7.) See, e.g., Timothy Seppala, One of the Most Profitable PatentTrolls Has Been Defanged, ENGADGET (Mar. 26, 2016),https://www.engadget.com/2016/03/26/uniloc-patent-troll-defeated[https://perma.cc/X7SE-H463]; Kelly Knaub, 'Anti-Troll'Targets Notorious NPE's Patent for PTAB Review, LAW360 (July 27,2016, 8:53 PM EDT), https://www.law360.com/articles/822011 (last visitedMay 10, 2019).

(8.) See Leahy-Smith America Invents Act [section] 18, Pub. L. No.112-29, 125 Stat. 284 (codified in scattered sections of 35 U.S.C.).

(9.) See infra Section II.A.

(10.) See infra Section II.B.

(11.) See U.S. Court of Appeals for the Federal Circuit: Judges,FED. JUD. CTR., https://www.fjc.gov/history/courts/u.s.-court-appeals-federal-circuit-judges [https://perma.cc/74FH-7AE5].

(12.) See, e.g., John Allison & Lisa Ouellette, How CourtsAdjudicate Patent Definiteness and Disclosure, 65 DUKE L.J. 609, 621-25(2016); see generally Mark Lemley et al., Our Divided Patent System, 82U. CHI. L. REV. 1073 (2015).

(13.) See, e.g., Jack DaSilva, Forum Shopping Under the PatentCases Pilot Program, 97 J. PAT. & TRADEMARK OFF. SOC'Y 630,639-50 (2015).

(14.) Then-professor Kimberly Moore's work comparing districtcourt judge and jury-made patent findings is an excellent example in theliterature, but it relies on case data now more than two decades old andexcludes PTAB cases entirely. See Kimberly Moore, Judges, Juries, andPatent Cases: An Empirical Peek Inside the Black Box, 99 MICH. L. REV.365, 367 (2000) (examining all patent trials from 1983 through 1999 andcomparing outcomes between bench and jury trials). Likewise, ProfessorsSaurabh Vishnubhakat, Arti Rai, and Jay Kesan offer a comprehensivelitigant-centric comparison of district courts versus the PTAB, but donot address appellate results. See generally Saurabh Vishnubhakat etal., Strategic Decision Making in Dual PTAB and District CourtProceedings, 31 BERKELEY TECH. L.J. 45 (2016). There is, finally, asubstantial body of empirical work examining the PTAB in isolation--buteven that scholarship tends to ignore appellate-level inquiry in favorof trial-level statistics. See, e.g., Yasser El-Gamal et al., The NewBattlefield: One Year of Inter Partes Review Under the America InventsAct, 42 AIPLA Q.J. 39 (2014).

(15.) It is also usually cost-prohibitive. PACER generally chargesper-page fees for access. See Electronic Public Access Fee Schedule,PACER (Dec. 1, 2013), https://www.pacer.gov/documents/epa_feesched.pdf[https://perma.cc/M5CL-FF5U]. To examine the orders, briefs, motions,and opinions for a dozen cases, the fees could easily add up to hundredsof dollars; for the thousands of cases analyzed herein, the fees wouldbe astronomical. Accordingly, I reiterate my deep gratitude to theAdministrative Office of the U.S. Courts, in coordination with theSupreme Court Fellows program, for granting me direct, unmetered accessto electronic case files--thereby enabling this unique research.

(16.) See infra notes 186-190 and accompanying text.

(17.) See infra notes 198-201 and accompanying text.

(18.) Pub. L. No. 112-29, 125 Stat. 284 (2011).

(19.) See, e.g., David Kappos, USPTO Director, Re-Inventing the USPatent System, USPTO (Sept. 16, 2011),http://www.uspto.gov/blog/director/entry/re_inventing_the_us_patent[https://perma.cc/AFA9-RE9R].

(20.) See, e.g., Arti K. Rai, Patent Validity Across the ExecutiveBranch: Ex Ante Foundations for Policy Development, 61 DUKE L.J. 1237,1238-39 (2012) (recognizing that prior to the passage of the AIA, thepatent debate focused on Federal Circuit decisions, not the USPTO, whichlacked any substantive rulemaking power); Samiyyah R. Ali, The GreatBalancing Act: The Effect of the America Invents Act on the Division ofPower Between the Patent and Trademark Office and the Federal Circuit,69 VAND. L. REV. 217, 222-23 (2016) (finding that the lack of both acongressional delegation and Federal Circuit deference to the USPTOcreated the perception of a "weak administrative agency"). Forexamples of typical pre-AIA case law, see Koninklijke Philips Elecs.N.V. v. Cardiac Sci. Operating Co., 590 F.3d 1326, 1336 (Fed. Cir. 2010)("The PTO lacks substantive rulemaking authority."); Merck& Co. v. Kessler, 80 F.3d 1543, 1549-50 (Fed. Cir. 1996)("[T]he broadest of the USPTO's rulemaking powers... does notgrant the Commissioner the authority to issue substantive rules.")(emphasis in original); and Animal Legal Def. Fund v. Quigg, 932 F.2d920, 930 (Fed. Cir. 1991) ("The authority granted [to the USPTO] isdirected to the 'conduct of proceedings' before the Office. Asubstantive declaration with regard to the Commissioner'sinterpretation of the patent statutes... does not fall within the usualinterpretation of such statutory language.").

(21.) See 35 U.S.C. [section] 321(b) (2012).

(22.) See id. [section] 311.

(23.) See id. [section][section] 135, 146, 291.

(24.) See id. [section] 257(a).

(25.) Melissa F. Wasserman, The Changing Guard of Patent Law:Chevron Deference for the PTO, 54 WM. & MARY L. REV. 1959, 1983(2013) (observing that the AIA's legislative history revealsCongress's intent to establish a formal adjudication process); seealso Sarah Tran, Patent Powers, 25 HARV. J.L. & TECH. 609, 631(2012) (noting that the AIA "give[s] the USPTO broad control overits new trial-like proceedings").

(26.) United States v. Mead Corp., 533 U.S. 218, 229 (2001)("We have recognized a very good indicator of delegation meritingChevron treatment in express congressional authorizations to engage inthe process of rulemaking or adjudication that produces regulations orrulings...."); see also Rai, supra note 20, at 1280 ("In fact,the executive branch could also use the postgrant-review [sic] authorityconferred upon the USPTO by the AIA to go one step further. As adoctrinal matter, under current Supreme Court precedent... thegovernment could ask for Chevron deference toward decisions made inpostgrant review proceedings.").

(27.) See, e.g., Ass'n for Molecular Pathology v. U.S. Patent& Trademark Office, 689 F.3d 1303, 1357 (Fed. Cir. 2012) ("[W]eowe deference [to the USPTO] only commensurate with 'thethoroughness of its consideration and the validity of itsreasoning.'" (quoting Merck & Co. v. Kessler, 80 F.3d1543, 1550 (Fed. Cir. 1996))), aff'd in part and rev'd in partsub nom. Ass'n for Molecular Pathology v. Myriad Genetics, Inc.,569 U.S. 576 (2013); Univ. of Mass. v. Kappos, 903 F. Supp. 2d 77, 84(D.D.C. 2012) ("[T]he PTO's determination is not entitled toChevron deference.... Instead, the PTO is only entitled to deferenceunder Skidmore "); see generally Skidmore v. Swift & Co., 323U.S. 134, 140 (1944); Jim Rossi, Respecting Deference: ConceptualizingSkidmore Within the Architecture of Chevron, 42 WM. & MARY L. REV.1105, 1125-27 (2001) ("Christensen clarifies that Skidmoredeference applies to most agency interpretive and policy statementsoutside of adjudication and notice-and-comment rulemaking.").

(28.) See, e.g., 35 U.S.C. [section] 135(b) (2012) (granting theUSPTO authority to set forth "standards" of "sufficientevidence to prove and rebut a claim of derivation"); id. [section]316(a)(2) (granting the USPTO authority to set forth "the standardsfor the showing of sufficient grounds to institute" inter partesreview); id. [section] 362(a)(2) (granting the USPTO authority to setforth "the standards for the showing of sufficient grounds toinstitute" post-grant review); id. [section] 316(a)(5) (grantingthe PTO authority to set forth the "standards" for"discovery"); id. [section] 316(a)(9) (granting the USPTOauthority to set forth "standards" for when amending a patentis proper). Courts generally recognize "standard-setting" as aform of substantive rulemaking. See, e.g., Whitman v. Am. TruckingAss'ns, 531 U.S. 457, 472-77 (2001); JEM Broad. Co. v. FCC, 22 F.3d320, 327 (D.C. Cir. 1994) ("[O]ur task is to identify whichsubstantive effects are sufficiently grave so that notice and commentare needed to safeguard the policies underlying the APA.")(internal quotation marks omitted).

(29.) USPTO, TRIAL STATISTICS at 3 (Nov. 2017),https://www.uspto.gov/sites/default/files/documents/trial_statistics_nov2017.pdf [https://perma.cc/SCM3-DUZ3].

(30.) U.S. COURT OF APPEALS FOR THE FED. CIR., APPEALS FILED INMAJOR ORIGINS 2008-2017 (Sept. 2017),http://www.cafc.uscourts.gov/sites/default/files/the-court/statistics/Hist_Caseld_by_Major_Origin_10-year.pdf [https://perma.cc/TC3F-7YJ4].

(31.) See id.

(32.) In FY 2016, for example, there were approximately 550district court appeals, which declined further to 500 in FY 2017. See id

(33.) See Pitt, supra note 6. It's worth adding that even thePTAB's own Chief Judge at the time, James Smith, was inclined toagree with the assessment. Ryan Davis, The PTAB's 'DeathSquad' Label, LAW360 (Aug. 14, 2014),https://www.law360.com/articles/567550/ptab-s-death-squad-label-not-totally-off-base-chief-says (last visited May 11, 2019). ("If weweren't, in part, doing some 'death squadding,' we wouldnot be doing what the statute calls on us to do.").

(34.) USPTO, PATENT TRIAL AND APPEAL BD. STATISTICS 10 (Sept.2016), https://www.uspto.gov/sites/default/files/documents/aia_statistics_september2016A.pdf [https://perma.cc/SZQ8-4K6T].

(35.) See, e.g., Richard Baker, America Invents Act Cost the USEconomy Over $1 Trillion, PATENTLYO (June 8, 2015),https://patentlyo.com/patent/2015/06/america-invents-trillion.html[https://perma.cc/HTV3-KRFX]; United Inventors Ass'n of America,Support the Patent System Fix, UIA (June 28, 2017),http://www.uiausa.org/single-post/2017/06/28/Support-the-Patent-System-Fix [https://perma.cc/22G6-FHP3] (advocating an end to inter partes andpost-grant proceedings before the USPTO); Gregory Dolin, Dubious PatentReform, 56 B.C. L. REV. 881, 881 (2015) ("[T]he newly createdsystem is open to abuse... such abuse occurs, and... the costs... aresubstantial.").

(36.) STRONGER Patents Act of 2017, S. 1390, 115th Cong. (2017)(among other things, narrowing the claim construction standards used inPTAB proceedings and creating an explicit presumption of validity).

(37.) USPTO, U.S. PATENT STATISTICS CHART CALENDAR YEARS 1963-2014,http://www.uspto.gov/web/offices/ac/ido/oeip/taf/us_stat.htm; see alsoGene Quinn, The Rise of Patent Litigation in America: 1980-2012,IPWATCHDOG (Apr. 9, 2013),http://www.ipwatchdog.com/2013/04/09/the-rise-of-patent-litigation-inamerica-1980-2012/id=38910/[https://perma.cc/VPJ4-CB3S].

(38.) U.S. GOV'T ACCOUNTABILITY OFFICE, GAO-13-465,INTELLECTUAL PROPERTY: ASSESSING FACTORS THAT AFFECT PATENT INFRINGEMENTLITIGATION COULD HELP IMPROVE PATENT QUALITY 11-13 (2013),http://www.gao.gov/assets/660/657103.pdf [https://perma.cc/E342-K7SL].

(39.) Id. at 21-24.

(40.) Id. at 28-30 ("Language describing emergingtechnologies, such as software, may be inherently imprecise becausethese technologies are constantly evolving... claims in software patentssometimes define the scope of the invention by encompassing an entirefunction--like sending an e-mail--rather than the specific means ofperforming that function.... [S]ome patents, particularlysoftware-related patents, should never have been issued because theywere obvious, not novel, or lacked definiteness.") (emphasisadded).

(41.) See, e.g., Brian Fung, The Supreme Court's Decision onSoftware Patents Still Doesn't Settle the Bigger Question, WASH.POST: THE SWITCH (June 20, 2014),http://www.washingtonpost.com/blogs/the-switch/wp/2014/06/20/the-supreme-courts-decision-onsoftware-patents-still-doesnt-settle-the-bigger-question (last visited Feb. 15, 2019) ("[The Court] didn't do muchto say what kinds of software should be patentable. In other words, thecourt decided the most basic conflict in the case, but more or lessdeclined to offer guidance for other, future cases.") (emphasis inoriginal); Robert Merges, Go Ask Alice--What Can You Patent After Alicev. CLS Bank?, SCOTUSBLOG (June 20, 2014, 12:04 PM),http://www.scotusblog.com/2014/06/symposium-go-ask-alice-what-can-you-patent-after-alice-v-cls-bank [https://perma.cc/B998-UP4S] ("To saywe did not get an answer is to miss the depth of the non-answer we didget.").

(42.) Perhaps most infamously, Personal Audio, LLC, used a patentcovering a "System for Disseminating Media Content RepresentingEpisodes in a Serialized Sequence" to claim infringement againstany major podcasting media groups, including CBS, NBC, and Fox. U.S.Patent No. 8,112,504 B2 (filed Feb. 7. 2012). That patent wassubsequently invalidated by the USPTO on inter partes review. SeePodcast 'Patent Troll' Faces Blow After US Ruling, BBC (Apr.13, 2015), http://www.bbc.com/news/technology-32286340[https://perma.cc/CLP3-BBVS]; This American Life: When Patents Attack...Part Two!, CHI. PUB. RADIO (May 31, 2013),https://www.thisamericanlife.org/496/when-patents-attack-part-two[https://perma.cc/9MLL-BR34]. See generally Personal Audio, LLC v.Electronic Frontier Foundation, 867 F.3d 1246 (Fed. Cir. 2017)(affirming the USPTO's invalidation decision).

(43.) See, e.g., Daniel Nazer, Infamous Wi-Fi Patent Troll Settlesfor Peanuts, ELECTRONIC FRONTIER FOUND. (Feb. 7, 2014),https://www.eff.org/deeplinks/2014/02/infamous-wi-fipatent-troll-settles-peanuts [https://perma.cc/QK95-6VBx] ("Armed with some patentspurchased from Broadcom, Innovatio sent thousands of letters targetinghotels and cafes that provide wi-Fi for customers... demand[ing] as muchas $2,500 per location.").

(44.) See, e.g., Josh Landau, IPR Successes: Cleaning up Messes,PATENT PROGRESS (Sept. 26, 2017),https://www.patentprogress.org/2017/09/26/ipr-successes-cleaning-messes/[https://perma.cc/A3BV-7ZX2] ("Using [inter partes review],APTWater succeeded in achieving a quick... settlement from the patenttroll after the PTAB determined that the patents were likely invalid....That money is money APTWater can spend developing new technologies forwater treatment.").

(45.) Unified Challenges the Three Most Prolific Patent Trolls of2016, UNIFIED PATENTS LLC (July 27, 2016),https://www.unifiedpatents.com/news/2016/7/27/unified-challenges-the-three-most-prolific-patent-trolls-of-2016 [https://perma.cc/FJ73-ZVBL]("Unified... refuses to pay off [trolls], instead disrupting anddeterring them by challenging poor-quality patents. As part of itsactivities, Unified has analyzed the patentability of more than ahundred patents and filed almost 50 inter partes reviews... since2013."); see also supra note 7.

(46.) 138 S. Ct. 1365, 1375 (2018) (contrasting a view of patentsas "private property" with the "public rightsdoctrine").

(47.) Petition for a Writ of Certiorari at 2-4, Oil States, 138 S.Ct. 1365 (No. 16-712).

(48.) Brief in Opposition of Certiorari at 1-3, Oil States, 138 S.Ct. 1365 (No. 16-712).

(49.) For examples from the petitioner's side, see Brief forPetitioner at 48, Oil States, 138 S. Ct. 1365 (No. 16-712) (referencingthe PTAB's "death squadding" tendency and potential biastowards high invalidation rates); Brief of Thirty-Nine Affected PatentOwners as Amici Curiae Supporting Petitioner at 9, Oil States, 138 S.Ct. 1365 (No. 16-712) (citing the increased "risk" and reduced"incentive [to] innovate[e]" as a result of the PTAB'sprocedures); Brief of the PhRMA Owners as Amici Curiae SupportingPetitioner at 29-31, Oil States, 138 S. Ct. 1365 (No. 16-712) (arguing"the special need for stability" in patent rights, both to"encourage innovation" and enable downstream marketcoordination); Brief of University of New Mexico as Amicus Curiae inSupport of Petitioner at 18, Oil States, 138 S. Ct. 1365 (No. 16-712)("[U]niversity-industry projects advance scientific knowledge andinnovation while providing a system for basic research.... [Inter partesreview] has hurt this process by devaluing patents and undermining thepatent system."). For examples from the respondent's side, seeBrief for Respondent at 20, Oil States, 138 S. Ct. 1365 (No. 16-712)("Congress considered it critical that the PTO have the ability toreexamine issued patents... [to] keep strong patents in the system whileremoving illegitimate ones, thereby helping to restore confidence in theeffectiveness of our patent system.") (citation omitted); Brief ofUSGMC as Amicus Curiae Supporting Respondents at 3-4, 21, Oil States,138 S. Ct. 1365 (No. 16-712) (referencing the need to stop "patenttroll" litigation); Brief of Amicus Curiae Dell, Inc. et al. asAmici Curiae Supporting Respondents at 17-21, Oil States, 138 S. Ct.1365 (No. 16-712) (referencing patent trolls, as well as "wastefullitigation costs" generally); Brief of Askeladden LLC as AmiciCuriae Supporting Respondents at 9-12, Oil States, 138 S. Ct. 1365 (No.16-712) (referencing patent trolls, as well as the "lowquality" of many granted patents).

(50.) Oil States, 138 S. Ct. at 1373 (emphasis in original).

(51.) Kimberly A. Moore, Forum Shopping in Patent Cases: DoesGeographic Choice Affect Innovation, 79 N.C. L. REV. 889, 903 (2001).

(52.) LEX MACHINA, http://law.lexmachina.com[https://perma.cc/N9PU-YNRM] (searching for case type "patent"and cases filed between Jan. 1, 2015 to Dec. 31, 2015).

(53.) Id.

(54.) See, e.g., J. Jonas Anderson, Court Competition for PatentCases, 163 U. PA. L. REV. 631 (2015).

(55.) See id. at 666-77; Daniel Klerman & Greg Reilly, ForumSelling, 89 S. CAL. L. REV. 263-65 (2016).

(56.) Matthew Sag, IP Litigation in U.S. District Courts: 1994 to2014, 101 IOWA L. REV. 1065, 1095-96 (2016).

(57.) PRICEWATERHOUSE COOPERS, 2017 PATENT LITIGATION STUDY: CHANGEON THE HORIZON? 22 (2017),http://www.ipwatchdog.com/wp-content/uploads/2017/05/2017-Patent-Litigation-Study_PwC.pdf [https://perma.cc/6MEV-HP4E].

(58.) See, e.g., This American Life: When Patents Attack!, CHI.PUB. RADIO (July 22, 2011),https://www.thisamericanlife.org/441/when-patents-attack[https://perma.cc/22R7-UK7K] ("Why would a company rent an officein a tiny town in East Texas... and leave it completely empty for ayear? The answer involves... a war waging right now, all across thesoftware and tech industries."); Joe Nocera, The Town That TrollsBuilt, BLOOMBERG: OPINION (May 25, 2017, 2:41 PM),https://www.bloomberg.com/view/articles/2017-05-25/the-texas-town-that-patent-trolls-built-j34rlmjc (last visited May 11, 2019); Adam Liptak,Supreme Court Considers Why Patent Trolls Love Texas, N.Y. TIMES (Mar.27, 2017), https://www.ny-times.com/2017/03/27/business/supreme-court-patent-trolls-tc-heartland-kraft.html [https://perma.cc/935E-JD87].

(59.) See Venue Equity and Non-Uniformity Elimination Act of 2016,S. 2733, 114th Cong. (2016).

(60.) 137 S. Ct. 1514 (2017).

(61.) 28 U.S.C. [section] 1400 (b) (2012).

(62.) 917 F.2d 1574, 1584 (Fed. Cir. 1990) (discussing history ofCongress's 1988 amendments to the general venue statute, 28 U.S.C.[section] 1391 (1988)).

(63.) TC Heartland, 137 S. Ct. at 1519.

(64.) In re Cray Inc., 871 F.3d 1355, 1362 (Fed. Cir. 2017).

(65.) See, e.g., Owen Byrd, Patent Litigation Trends in the ThreeMonths After TC Heartland, LEX MACHINA (Oct. 18, 2017),https://lexmachina.com/patent-litigation-trends-in-the-three-months-after-t-c-heartland/[https://perma.cc/W5UH-8LBU].

(66.) See id. (finding that the sum share of patent filings in theEastern District of Texas and the District of Delaware decreased by onlyseven percentage points in the first three months after TC Heartland).This is somewhat unsurprising, given the dominance of Delaware as aplace of incorporation. Why Businesses Choose Delaware, DELAWARE.GOV,https://corplaw.delaware.gov/why-businesses-choose-delaware/[https://perma.cc/WXJ6-S82A] ("Indeed, more than 60 percent of theFortune 500 companies are incorporated in Delaware.").

(67.) LEX MACHINA, http://law.lexmachina.com[https://perma.cc/N9PU-YNRM] (searching for case type"patent," then comparing between cases filed between Dec. 22,2016 and May 22, 2017, and cases filed between Oct. 1, 2018 and Apr. 1,2019).

(68.) Patent Cases Pilot Program, Pub. L. No. 111-349, 124 Stat.3674 (2011).

(69.) Participating districts include the Central District ofCalifornia, the Northern District of California, the Southern Districtof California, the Northern District of Illinois, the District ofMaryland, the District of New Jersey, the District of Nevada, theEastern District of New York, the Southern District of New York, theWestern District of Pennsylvania, the Western District of Tennessee, theEastern District of Texas, and the Northern District of Texas. MARGARETS. WILLIAMS, REBECCA EYRE, & JOE CECIL, FED. JUDICIAL CTR., PATENTPILOT PROGRAM: FIVE-YEAR REPORT 1-3 (April 2016),https://www.fjc.gov/sites/default/files/2016/Patent%20Pilot%20Program%20Five-Year%20Report%20(2016).pdf [https://perma.cc/B3 SZ-TATG].

(70.) See id. at 2.

(71.) Patent Cases Pilot Program, Pub. L. No. 111-349, 124 Stat.3674 (2011) (codified at 28 U.S.C.[section] 137 (2012)).

(72.) WILLIAMS ET AL., supra note 69, at 10-12.

(73.) Id. at 22-23 ("On average, cases before designatedjudges take less time than those before nondesignated judges, and thedifferences are statistically significant.").

(74.) Mark A. Lemley, Why Do Juries Decide if Patents Are Valid?,99 VA. L. REV. 1673, 1703-04 (2013) (citing Gary M. Ropski,Constitutional and Procedural Aspects of the Use of Juries in PatentLitigation, 58 J. PAT. OFF. SOC'Y 609, 673 (1976)).

(75.) See FED. R. CIV. P. 2 (effective Sept. 16, 1938).

(76.) Lemley, supra note 74, at 1706.

(77.) See HERBERT F. SCHWARTZ, FED. JUDICIAL CTR., PATENT LAW ANDPRACTICE 130 (2d ed. 1995); see also Mark A. Lemley et al., Rush toJudgment? Trial Length and Outcomes in Patent Cases, 41 AIPLA Q.J. 169,174 n.2 (2013).

(78.) PRICEWATERHOUSECOOPERS, supra note 57, at 6 (excludinginfringement suits relating to Abbreviated New Drug Applications foroutlier reasons).

(79.) See id.; see also Lemley, supra note 74, at 1706.

(80.) See, e.g., OFFICE OF LEGAL EDUC., EXEC. OFFICE FOR U.S.ATTORNEYS, PROSECUTING INTELLECTUAL PROPERTY CRIMES 8 (4th ed. 2013),https://www.justice.gov/sites/default/files/criminal-ccips/legacy/2015/03/26/prosecuting_ip_crimes_manual_2013.pdf [https://perma.cc/L8CE-3L4D]("[T]here are no criminal--only civil--penalties for committingpatent infringement."); Jacob S. Sherkow, Patent Infringement asCriminal Conduct, 19 MICH. TELECOMM. & TECH. L. REV. 1, 2 n.6("[N]othing in the patent statute provides criminal penalties forany violation of its provisions.") (citing 35 U.S.C.[section][section] 281-297 (2012)).

(81.) See 17 U.S.C. [section] 506(a) (2012); 18 U.S.C. [section]2319 (2012).

(82.) See 18 U.S.C. [section] 2320 (2012 & Supp. IV 2016).

(83.) FED. R. CIV. P. 38(b).

(84.) See, e.g., Kimberly A. Moore, Jury Demands: Who'sAsking?, 17 BERKELEY TECH. L.J. 847, 855 (2002) ("[P]laintiffsdemanded a jury significantly more often than did defendants; plaintiffsdemanded a jury in 71% of all cases."); Richard A. Posner, WhyThere Are Too Many Patents in America, ATLANTIC (July 12, 2012),https://www.theatlantic.com/business/archive/2012/07/why-there-are-too-many-patents-in-america/259725/[https://perma.cc/62GP-RLGT]("[P]atent plaintiffs tend to request trial by jury because theybelieve that jurors tend to favor patentees.... ").

(85.) See PRICEWATERHOUSE COOPERS, supra note 78, at 6; Kimberly A.Moore, Populism and Patents, 82 N.Y.U. L. REV. 69, 107 (2007); Lemley,supra note 74, at 174-85; Andrei Iancu & Jay Chung, Real Reasons theEastern District of Texas Draws Patent Cases--Beyond Lore and Anecdote,14 SMU SCI. & TECH. L. REV. 299, 305 (2011) (finding a nationwidepatentee win rate in jury trials of 68%).

(86.) See Judicial Panel Discussions, supra note 2; Lemley, supranote 74, at 1675-76 (critiquing jury use in patent cases as, among otherthings, ahistorical); Arti K. Rai, Specialized Trial Courts:Concentrating Expertise on Fact, 17 BERKELEY TECH. L.J. 877, 889 (2002)("Indeed, a significant analytical and empirical literature...suggests that lay persons, confronted with competing expert accounts ofa scientific or technological dispute, are unlikely to make consideredjudgments."). Moreover, arguments along these lines have been madeby patent litigants themselves for at least a half-century. See, e.g.,Tights, Inc. v. Stanley, 441 F.2d 336, 340 (4th Cir. 1971) ("Thebriefs filed in opposition... argue strongly that patent cases are toocomplicated and difficult for a jury to deal with.").

(87.) See, e.g., James Oldman, On the Question of a ComplexityException to the Seventh Amendment Guarantee of Trial by Jury, 71 OHIOST. L.J. 1031, 1051-53 (2010); Frank M. Loo, A Rationale for anException to the Seventh Amendment Right to a Jury Trial: In re JapaneseElectronic Products Antitrust Litigation, 30 CLEV. ST. L. REV. 647,654-55 (1981); see generally Patrick Devlin, Jury Trial of ComplexCases: English Practice at the Time of the Seventh Amendment, 80 COLUM.L. REV. 43 (1980). There are indications of at least some skepticismfrom the Supreme Court itself. See, e.g., Ross v. Bernard, 396 U.S. 531,538 n.10 (1970) (stating that the jury triability of an issue "isdetermined by considering, first, the pre-merger [of equity and law]custom with reference to such questions; second, the remedy sought; and,third, the practical abilities and limitations of juries")(emphasis added).

(88.) 52 F.3d 967, 979 (Fed. Cir. 1995), aff'd, 515 U.S. 1192(1995).

(89.) Id. at 993 (Mayer, J., concurring in the judgment).

(90.) Pub. L. No. 111-349, [section] 1(b)(2)(A), 124 Stat. 3674,3675 (2011) (stating that the pilot districts shall be selected fromamong those with either "the largest number of patent and plantvariety protection cases" or specialized "local rules forpatent and plant variety protection cases").

(91.) Those judges are, in order, Judge Robert Schroeder (EasternDistrict of Texas, 1,921 patent cases), Chief Judge Leonard Stark(District of Delaware, 1,216 patent cases), Judge Richard Andrews(District of Delaware, 1,131 patent cases, and Judge Gregory Sleet(District of Delaware, 1,088 patent cases). LEX MACHINA,http://law.lexmachina.com [https://perma.cc/N9PU-YNRM] (searching forcase type "patent" and cases filed between Jan. 1, 2013 toJan. 1, 2018).

(92.) See, e.g., Brian J. Love & James Yoon, PredictablyExpensive: A Critical Look at Patent Litigation in the Eastern Districtof Texas, 20 STAN. TECH. L. REV. 1, 18 (2017) ("[B]ecause EastTexas patent cases are both unlikely to be transferred out of thedistrict and unlikely to be completely resolved by summary judgment,they are... disproportionately likely to go to trial."); Sag, supranote 56, at 1100 (noting the reduced availability of summary judgment inthe District of Delaware and Eastern District of Texas, andcomparatively greater reliance on "jury verdict[s]");Anderson, supra note 54, at 674-75.

(93.) See U.S. COURT OF APPEALS FOR THE FED. CIRCUIT, YEAR-TO-DATEACTIVITY (Sept. 30, 2016),http://www.cafc.uscourts.gov/sites/default/files/the-court/statistics/revYTD_Activity_9.30.16.pdf [https://perma.cc/4VHE-CVGX]; U.S. COURT OFAPPEALS FOR THE FED. CIRCUIT, YEAR-TO-DATE ACTIVITY (Sept. 30, 2015),http://www.cafc.uscourts.gov/sites/default/files/the-court/statistics/ytd_activity_9_30_15.pdf [https://perma.cc/HA87-YKGX].

(94.) Lemley et al., supra note 12, at 1081 ("Coding ofoutcomes, especially in patent cases, is notoriously difficult and timeconsuming, requiring deep knowledge of patent law and litigation as wellas the motivation to devote long hours to the task.").

(95.) LEX MACHINA, http://law.lexmachina.com[https://perma.cc/N9PU-YNRM].

(96.) See, e.g., Theodore Eisenberg & Martin T. Wells, TrialOutcomes and Demographics: Is There a Bronx Effect?, 80 TEX. L. REV.1839, 1843-45 (2002); Jason Scott Johnston & Joel Waldfogel, DoesRepeat Play Elicit Cooperation? Evidence from Federal Civil Litigation,31 J. LEGAL STUD. 39, 46-47 (2002). But see, e.g., Christina L. Boyd& David A. Hoffman, The Use and Unreliability of Federal Nature ofSuit Codes, 2017 MICH. ST. L. REV. 997, 1024-25 (2017) (finding highvariance among PACER data labeling the underlying cause of action).

(97.) PACER docket data for Federal Circuit cases typicallyindicates the final outcome, e.g., "affirmed in part and reversedin part," but offers no way to tell which outcome applies to whichfindings made below (or whether certain findings were addressed at allon their merits, rather than mooted).

(98.) Consider, for example, Orenshteyn v. Citrix Systems, Inc.,609 F. App'x 654 (Fed. Cir. 2015) (summary affirmance). The appealis labeled as "Patent Infringement" on PACER due to its causeof action below, but the only issue actually challenged by appellantswas the district court's award of sanctions pursuant to 28 U.S.C.[section] 1927 (2012). See Brief of Appellant at 17-20, Orenshteyn v.Citrix Systems, Inc., 609 F. App'x 654 (Fed. Cir. 2015) (No.15-1056).

(99.) The Federal Circuit Data Project, UNIV. OF IOWA,https://empirical.law.mowa.edu/compendium-federal-circuit-decisions[https://perma.cc/Q783-NCP4].

(100.) Opening Brief of Plaintiff-Appellant at 17-18, TrustedKnight Corp. v. Int'l Bus. Machs. Corp., 681 F. App'x 898(Fed. Cir. 2016) (No. 16-1510).

(101.) See Brief of Appellant at 22-23, Schoeller-BleckmannOilfield Equip. AG v. Churchill Drilling Tools US, Inc., 664 F.App'x 949 (Fed. Cir. 2016) (No. 16-1494).

(102.) See Schoeller-Bleckmann Oilfield Equip. AG, 664 F.App'x at 953 ("We need not and do not address the Board'sconclusions that... Bourgoyne rendered obvious claims 17 and 18.").

(103.) FED. CIR. R. 36 (stating that "[t]he court may enter ajudgment of affirmance without opinion, citing this rule, when... anopinion would have no precedential value" and the decision belowpresents no error requiring reversal, vacatur, or remand).

(104.) See Corrected Brief for Appellant Enovsys LLC at 6, 48-54,Enovsys LLC v. AT&T Mobility LLC, 678 F. App'x 1025 (Fed. Cir.2016) (No. 16-1691, -1767).

(105.) See, e.g., Duke University v. Biomarin Pharm. Inc., 685 F.App'x 967, 976 (Fed. Cir. 2017) ("Because addressingDuke's arguments relating to whether van Bree anticipates claims 1and 20 resolves this appeal,... we need not address Duke'sarguments relating to the Board's conclusion that claims 1 and 20were [also] unpatentable as obvious.").

(106.) It is worth noting that this issue only applied in theinvalidity context. If the PTAB, for example, found a given set ofclaims both not anticipated and not obvious, then a Rule 36 judgmentwould necessarily be affirming both findings. Otherwise, the endresult--patent validity--would not stand.

(107.) See, e.g., MPHJ Tech. Invs., LLC v. Ricoh Americas Corp.,847 F.3d 1363, 1370-71 (Fed. Cir. 2017); TVIIM, LLC v. McAfee, Inc., 851F.3d 1356, 1364 (Fed. Cir. 2017); Schoeller-Bleckmann Oilfield Equip.AG, 664 F. App'x at 953; Prolitec, Inc. v. Scentair Techs., Inc.,807 F.3d 1353, 1362 (Fed. Cir. 2015). There are instances of the FederalCircuit taking the opposite approach, but they are considerably rarer.See, e.g., In re Anderson, 662 F. App'x 958, 965 (Fed. Cir. 2016).

(108.) 628 F. App'x 766 (Fed. Cir. 2016).

(109.) See Appellant's Brief at 2, In re Clouding Corp., 628F. App'x at 766 (No. 15-1615).

(110.) VirnetX Inc. v. Apple, Inc., 671 F. App'x 789, 789-90(Fed. Cir. 2016) (emphasis added).

(111.) As of this writing, only three lead cases from the relevanttime period remain unresolved, and hence are not included in theanalysis: Docket Nos. 16-1919, 16-2222, and 16-2231.

(112.) In full: (1) case number, date, and Federal Circuit panel;(2) precedential status of the opinion; (3) patent number(s) andclaim(s) in dispute; (4) patent technology class; (5) specific tribunalbelow (e.g., Eastern District of Texas, PTAB, etc.); (6) for districtcourts, the particular judge; (7) validity finding(s) by the tribunalbelow (e.g., non-obvious, ineligible subject matter, etc.); (8) contextof finding below (e.g., summary judgment, inter partes review, etc.);(9) whether any predicate claim constructions were modified on appeal;and (10) disposition of the finding on appeal (e.g., affirmed, vacated,etc.).

(113.) See USPTO, US Classes by Number with Title,https://www.uspto.gov/web/patents/classification/selectnumwithtitle.htm[https://perma.cc/HZL7-GGUP].

(114.) U.S. Patent No. 7,647,460 (filed Sept. 23, 1997).

(115.) This study uses the same supercategory conversion system asProfessors Bronwyn Hall, Adam Jaffe, and Manuel Trajtenberg in theirscholarship for the National Bureau of Economic Research. See Bronwyn H.Hall et al., The NBER Patent Citations Data File: Lessons, Insights, andMethodological Tools 13 (Nat'l Bureau of Econ. Research, WorkingPaper No. 8498, 2001). Although their supercategory system is notnecessarily unique, it is in particularly common use. See, e.g.,Vishnubhakat, supra note 14, at 66; Gregory Nemet & Evan Johnson, DoImportant Inventions Benefit from Knowledge Originating in OtherTechnological Domains?, 41 RES. POL'Y 190, 192 (2012); AlbertoGalasso & Mark Schankerman, Patent and Cumulative Innovation: CausalEvidence from the Courts, 130 Q.J. ECON. 317, 332 (2014); Shawn P.Miller, Where's the Innovation: An Analysis of the Quantity andQuality of Anticipated and Obvious Patents, 18 VA. J.L. & TECH. 1,23 n.59 (2013) (using the HJT system, among others).

(116.) Hall et al., supra note 115, at 13.

(117.) SEC v. Chenery Corp., 318 U.S. 80, 95 (1943). See generallyKevin M. Stack, The Constitutional Foundations of Chenery, 116 YALE L.J.952 (2007) (sketching the doctrine's background, evolution, andsignificance).

(118.) Helvering v. Gowran, 302 U.S. 238, 245 (1937). See alsoJerry L. Mashaw, Small Things like Reasons Are Put in a Jar: Reason andLegitimacy in the Administrative State, 70 FORDHAM L. REV. 17, 19-26(2001) (contrasting judicial review of agency action and judicialdecision-making).

(119.) Camp v. Pitts, 411 U.S. 138, 143 (1973).

(120.) In re Nuvasive, Inc., 842 F.3d 1376, 1382 (Fed. Cir. 2016)(citing Chenery, 318 U.S. at 94; Motor Vehicle Mfrs. Ass'n v. StateFarm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983)).

(121.) Google Inc. v. Intellectual Ventures II LLC, 701 F.App'x 946, 955 (Fed. Cir. 2017).

(122.) In re Hodges, 882 F.3d 1107, 1117 (Fed. Cir. 2018)("The Board has failed to explain its reasoning.... When faced withsimilarly deficient factual findings, 'we have consistently vacatedand remanded for further proceedings.'") (quoting In re VanOs, 844 F.3d 1359, 1362 (Fed. Cir. 2017)).

(123.) In re Ferguson, 558 F.3d 1359, 1363 (Fed. Cir. 2009)(quoting In re Bilski, 545 F.3d 943, 951 (Fed. Cir. 2008) (en banc),aff'd sub nom. Bilski v. Kappos, 561 U.S. 592 (2010)).

(124.) Biosig Instruments, Inc. v. Nautilus, Inc., 783 F.3d 1374,1377 (Fed. Cir. 2015) (citing Interval Licensing LLC v. AOL, Inc., 766F.3d 1364, 1370 (Fed. Cir. 2014)).

(125.) Ariosa Diagnostics v. Verinata Health, Inc., 805 F.3d 1359,1364 (Fed. Cir. 2015) ("[T]he scope and content of the prior art,the differences between the prior art and the claimed invention, thelevel of ordinary skill in the art, the presence or absence of amotivation to combine or modify with a reasonable expectation ofsuccess, and objective indicia of non-obviousness.").

(126.) In re Morsa, 713 F.3d 104, 109 (Fed. Cir. 2013)("Enablement is a question of law based on underlying factualfindings.").

(127.) Id. (citing In re Gleave, 560 F.3d 1331, 1334-35 (Fed. Cir.2009)); see also Kennametal, Inc. v. Ingersoll Cutting Tool Co., 780F.3d 1376, 1381 (Fed. Cir. 2015) ("Anticipation... is a question of

fact... [which] [w]e review... for substantial evidence....").

(128.) In re Owens, 710 F.3d 1362, 1366 (Fed. Cir. 2013) (citingAriad Pharm., Inc. v. Eli Lilly & Co., 598 F.3d 1336, 1351 (Fed.Cir. 2010) (en banc)).

(129.) Intelligent Bio-Systems, Inc. v. Illumina Cambridge Ltd.,821 F.3d 1359, 1366 (Fed. Cir. 2016) (quoting In re Gartside, 203 F.3d1305, 1312 (Fed. Cir. 2000)).

(130.) See Leahy-Smith America Invents Act [section] 18, Pub. L.No. 112-29, 125 Stat. 284.

(131.) Id.; see generally Versata Development Grp., Inc. v. SAPAm., Inc., 793 F.3d 1306, 1310-11 (Fed. Cir. 2015) (explaining thehistory and structure of CBM patents and CBM review).

(132.) GlaxoSmithKline LLC v. Anchen Pharmaceuticals, Inc., 2012 WL5594540, at *4 (D. Del. Nov. 15, 2012); see generally Markman v.Westview Instruments Inc., 517 U.S. 370 (1996) (establishing the modernclaim construction process).

(133.) Peter S. Menell et al., Patent Claim Construction: A ModernSynthesis and Structured Framework, 25 BERKELEY TECH. L.J. 711, 714(2010).

(134.) Straight Path IP Grp., Inc. v. Sipnet EU S.R.O., 806 F.3d1356, 1360 (Fed. Cir. 2015) (emphasis added); see also Phillips v. AWHCorp., 415 F.3d 1303, 1317 (Fed. Cir. 2005) (outlining the distinctionbetween extrinsic and intrinsic evidence). Extrinsic evidence istypically inventor testimony, expert declarations, and documentaryevidence such as dictionaries, technical manuals, and scientifictreatises. See Menell et al., supra note 133, at 721.

(135.) PPC Broadband, Inc. v. Corning Optical Commc'ns RF,LLC, 815 F.3d 747, 751 (Fed. Cir. 2016).

(136.) In this data set, the Federal Circuit modified trial-levelclaim constructions in 54 of the 1,087 data points. Just over half(55.6%) of such instances led to some form of remand. The remainder wereeither affirmed or reversed outright. The latter practice has beencriticized. See, e.g., Homeland Housewares, LLC v. Whirlpool Corp., 865F.3d 1372, 1379 (Fed. Cir. 2017) (Newman, J., dissenting) ("Mycolleagues' de novo findings are contrary to the record, overstepour appellate role, and are incorrect in fact and law."); see Rai,Specialized Trial Courts, supra note 86, at 883-85 (suggesting that anaversion to remand after modifying claim construction demonstrates alack of deference from the Federal Circuit to the tribunal below).

(137.) Before 2018, post-grant procedures before the PTAB relied onthe "broadest reasonable interpretation" standard for claimconstruction, the same standard used in patent prosecution. 37 C.F.R.[section] 42.100(b) (2017); see also Cuozzo Speed Techs., LLC v. Lee,136 S. Ct. 2131, 2136 (2016) (affirming the PTAB's use of broadestreasonable claim construction in post-grant procedures); cf. In re Am.Acad. of Sci. Tech Ctr., 367 F.3d 1359, 1364 (Fed. Cir. 2004)("During examination, 'claims... are to be given theirbroadest reasonable interpretation consistent with thespecification....'") (quoting In re Bond, 910 F.2d 831, 833(Fed. Cir. 1990)). District courts, on the other hand, apply thePhillips standard of claim construction: unless otherwise compelled bythe patent's specification, its prosecution history, orparticularly compelling extrinsic evidence, claim language is given itsordinary and customary meaning (from the perspective of a person ofordinary skill in the art). See, e.g., Ericsson, Inc. v. D-Link Sys.,Inc., 773 F.3d 1201, 1217-18 (Fed. Cir. 2014) (citing Phillips v. AWHCorp., 415 F.3d 1301, 1312-13, 1315-17 (Fed. Cir. 2005) (en banc)).After a rule change on October 3, 2018, the PTAB now uses the same claimconstruction standard as district courts in its post-grant procedures.Changes to the Claim Construction Standard for Interpreting Claims inTrial Proceedings Before the Patent Trial and Appeal Board, 83 Fed. Reg.51,358 (Oct. 11, 2018) (to be codified at 37 C.F.R. [section] 42.100).

(138.) Ex parte reexamination, created in 1980, was the firstmechanism for challenging the validity of issued patents before the PTOitself--by way of petition. See 35 U.S.C. [section][section] 301-307(2018); see generally Raymond A. Mercado, Ensuring the Integrity ofAdministrative Challenges to Patents: Lessons from Reexamination, 14COLUM. SCI. & TECH. L. REV. 558, 561-68 (2013); Mark D. Janis,Rethinking Reexamination: Toward a Viable Administrative RevocationSystem for U.S. Patent Law, 11 HARV. J.L. & TECH. 1 (1997). In 1999,Congress added an inter partes option, allowing the challenger toparticipate in the reexamination directly, including briefing andappeal. See 35 U.S.C. [section][section] 311-318 (2012). The AIAreplaced inter partes reexamination with post-grant and inter partesreview, which are more formal, trial-like, and robust. See supra notes22-28 and accompanying text.

(139.) See, e.g., In re Man Machine Interface Techs. LLC, 822 F.3d1282, 1285 (Fed. Cir. 2016).

(140.) See, e.g., Dynamic Drinkware, LLC v. Nat'l Graphics,Inc., 800 F.3d 1375, 1378 (Fed. Cir. 2015).

(141.) See, e.g., Blue Calypso, LLC v. Groupon, Inc., 815 F.3d1331, 1337, 1341 (Fed. Cir. 2016).

(142.) Mathematically, comparing the post-AIA and pre-AIAaffirmance rates directly yields a chi-square statistic of 0.89, and ap-value of 0.35--further supporting the lack of a meaningful pattern.

(143.) See, e.g., Moore, Judges, Juries, and Patent Cases, supranote 14, at 367-69 (examining all patent trials from 1983 through 1999and comparing outcomes between bench and jury trials).

(144.) These instances were defined strictly, only including juryverdicts actually dispositive on the given issue (40), and one advisoryverdict to which the district court judge explicitly deferred.

(145.) LEX MACHINA, http://law.lexmachina.com[https://perma.cc/N9PU-YNRM] (searching for case type "patent"and cases filed between Jan. 1, 2014 to Jan. 1, 2019).

(146.) Neither the district- nor judge-level analyses yieldsignificant chi-square statistics. Comparing the affirmance rates forpatent-heavy districts to standard ones results in a chi-square value of0.40, corresponding to a p-value of 0.53. Comparing the affirmance ratesfor patent-heavy judges to standard ones results in a chi-square valueof 0.90, corresponding to a p-value of 0.34.

(147.) Again, the chi-square statistic from comparing thesecategories--less than 0.01--strongly suggests the lack of a meaningfulpattern, corresponding to a p-value of 0.98.

(148.) WILLIAMS ET AL., supra note 69, at 32, 36 (finding"that appeals of pilot cases are no more or less common thanappeals of nonpilot cases," and that "pilot cases and nonpilotcases are 'correct'"--as measured by results onappeal--"at approximately the same rate").

(149.) See, e.g., Teva Pharm. USA, Inc. v. Sandoz, Inc., 135 S. Ct.831, 838 (2015) ("This factual determination, like all otherfactual determinations, must be reviewed for clear error.") (citingPullman-Standard v. Swint, 456 U.S. 273, 287 (1982)).

(150.) See infra text accompanying notes 167-169.

(151.) Impax Labs. Inc. v. Lannett Holdings Inc., 893 F.3d 1372,1378 (Fed. Cir. 2018) (citing United States v. U.S. Gypsum Co., 333 U.S.364, 395 (1948)).

(152.) There was, additionally, a single case that involved afinding of prior conception--a "question of law... based uponunderlying factual determinations." Price v. Symsek, 988 F.2d 1187,1190 (Fed. Cir. 1993). That case was affirmed. There were no data pointsinvolving enablement.

(153.) There were two design patent cases involved in the data set,which could not properly be categorized using the HJT (utility patent)typology. Each yielded a single data point. The first, Yao-Hung Huang v.Marklyn Group, Inc., 636 F. App'x 795 (Fed. Cir. 2016), was a Rule36 affirmance of the district court's finding of anticipation. Thesecond, 3form, Inc. v. Lumicor, Inc., 678 F. App'x 1002 (Fed. Cir.2017), affirmed the district court's finding of obviousness in afull opinion.

(154.) Comparing the affirm vs. non-affirm rates between the PTABand district courts directly yields a chi-square statistic of 8.16,corresponding to a highly significant p-value of 0.0043.

(155.) As indicated in the table, Fisher's exact testdemonstrates that the difference between the affirmance rate of the PTABand district courts is only statistically significant within thecategory of findings of fact. Fisher's exact test, rather thanchi-squared, is used here due to the relatively small sample sizes percategory.

(156.) Because the dependent variable in this analysis is binary("affirmed" or "not"), logistic regression offers amore accurate picture of the underlying data than typical linearregression models would. Cf. FRED C. PAMPEL, LOGISTIC REGRESSION: APRIMER 17 (2000); SCOTT MENARD, LOGISTIC REGRESSION: FROM INTRODUCTORYTO ADVANCED CONCEPTS AND APPLICATIONS 1-2 (2010).

(157.) Intercoder reliability refers to the extent to whichindependent coders identically evaluate the same materials. Seegenerally Matthew Lombard et al., Practical Resources for Assessing andReporting Intercoder Reliability in Content Analysis Research Projects(June 1, 2010), http://matthewlombard.com/reliability/index_print.html[https://perma.cc/97QP-UYL6].

(158.) See supra Table 17.

(159.) U.S. COURT OF APPEALS FOR THE FED. CIR., CASELOAD BY MAJORORIGIN 2008-2017, http://www.cafc.uscourts.gov/sites/default/files/the-court/statistics/Hist_Caseld_by_Major_Origin_10-year.pdf[https://perma.cc/5PC7-9JR3].

(160.) Dissents were coded strictly and narrowly. That is, if aFederal Circuit judge dissented, but on an issue other than the relevantvalidity finding itself (for example, arguing that the tribunal belowlacked jurisdiction), then it was not counted as a "dissent"for these purposes.

(161.) In particular, comparing the dissent vs. non-dissent ratebetween the two tribunals directly yields a chi-square statistic of1.25, corresponding to a p-value of 0.2629, suggesting a lack ofstatistical significance.

(162.) See supra Table 7.

(163.) Comparing the affirm vs. non-affirm rate between theforum-choice and no-choice categories directly yields a chi-squarestatistic of 0.61, corresponding to a p-value of 0.4341, suggesting alack of statistical significance.

(164.) Recovery of fees generated in PTAB disputes is effectivelylimited to circ*mstances where there is parallel district courtlitigation, and the district court applies its own fee-shifting powerbroadly under 35 U.S.C. [section] 285 (2012). See generally, Megan M. LaBelle, Fee Shifting for PTAB Proceedings, 34 TEX. INTELL. PROP. L.J. 367(2016) (examining the use of [section] 285 to shift PTAB fees, andnoting the lack of direct fee-shifting power for the PTAB itself).

(165.) See, e.g., AM. INTELL. PROP. L. ASS'N, 2015 REPORT OFTHE ECONOMIC SURVEY at 37-38 (2015) (finding the median cost of all IPRsto be $275,000, compared to $600,000 for district court patentlitigation with less than $1 million at risk, and up to $5 million incosts for cases with more than $25 million at risk); IPRs: Reality Amidthe Pyrotechnics, RATIONALPATENT (July 2, 2015),https://www.rpxcorp.com/2015/07/02/iprs-reality-amid-the-pyrotechnics/(last visited May 11, 2019).

(166.) See supra notes 149-150 and accompanying text.

(167.) See, e.g., In re Gartside, 203 F.3d 1305, 1312 (Fed. Cir.2000); Universal Camera Corp. v. NLRB, 340 U.S. 474, 477 (1951).

(168.) See, e.g., Amanda Peters, The Meaning, Measure, and Misuseof Standards of Review, 13 LEWIS & CLARK L. REV. 233, 245-46 (2009)("The problem with this [substantial evidence] standard... is thatthere is effectively no difference between it and the clearly erroneousstandard of review."); Paul R. Verkuil, An Outcomes Analysis ofScope of Review Standards, 44 WM. & MARY L. REV. 679, 689 n.36(2002) ("'Clearly erroneous' is distinguished from'substantial evidence' in theory although the two standardsare often equated in practice.").

(169.) Dickinson v. Zurko, 527 U.S. 150, 162-63 (1999).

(170.) See, e.g., Belkin Int'l, Inc. v. Kappos, 696 F.3d 1379,1381 (Fed. Cir. 2012) (stating that the Board's interpretation ofunderlying patent statutes is "review[ed] de novo"); Merck& Co., Inc. v. Kessler, 80 F.3d 1543, 1549-50 (Fed. Cir. 1996);Glaxo Operations UK Ltd. v. Quigg, 894 F.2d 392, 398 (Fed. Cir. 1990).See Tran, supra note 25, at 616 ("The Federal Circuit... hasassumed exclusive responsibility for making substantive interpretationsof the Patent Act..."); Sapna Kumar, Expert Court, Expert Agency,44 U.C. DAVIS L. REV. 1547, 1550 (2011).

(171.) See, e.g., Aqua Prods., Inc. v. Matal, 872 F.3d 1290, 1296(Fed. Cir. 2017) (en banc) (rejecting Chevron deference for theUSPTO's interpretation of 35 U.S.C. [section] 316(e) (2012),governing the assignment of burdens of proof in inter partes reviewproceedings); Biogen MA, Inc. v. Japanese Found. for Cancer Res., 785F.3d 648, 657 (Fed. Cir. 2015) (rejecting Chevron deference for theUSPTO's interpretation of AIA [section] 3(n)(1), governing theeffective date of the AIA's new procedures).

(172.) See, e.g., Keener v. United States, 551 F.3d 1358, 1363(Fed. Cir. 2009) (granting Chevron deference to the Treasury'sinterpretation of the Internal Revenue Code); Patterson v. Dep't ofthe Interior, 424 F.3d 1151, 1159 (Fed. Cir. 2005) (granting Chevrondeference to the Office of Personnel Management); Tunik v. Merit Sys.Prot. Bd., 407 F.3d 1326, 1336 (Fed. Cir. 2005) (granting Chevrondeference to the Merit Systems Protection Board).

(173.) See Eastman Kodak Co. v. Bell & Howell Document Mgmt.Prods. Co., 994 F.2d 1569, 1571 (Fed. Cir. 1993) (applying Chevrondeference to the TTAB's interpretation of substantive provisions ofthe Lanham Act). But see Melissa F. Wasserman, What Administrative LawCan Teach the Trademark System, 93 WASH. U. L. REV. 1511, 1543 (2016)("[T]he Eastman Kodak opinion has had a limited impact on theFederal Circuit's jurisprudence.").

(174.) Merck & Cie v. Gnosis S.P.A., 820 F.3d 432, 433 (Fed.Cir. 2016) (O'Malley, J., concurring in denial of rehearing enbanc) (joined by Judges Wallach and Stoll); see also id. at 438 (Newman,J., dissenting in denial of rehearing en banc) ("With thesesubstantive consequences, it is not reasonable to infer the legislativeintent to apply highly deferential review to issues traditionallysubjected to appellate review for... clear error.").

(175.) See, e.g., Wi-Fi One, LLC v. Broadcom Corp., 878 F.3d 1364,1367 (Fed. Cir. 2018) (en banc) (holding that initial time-bardeterminations in inter partes reviews are appealable, despite generalbar on appealing the PTAB's decision to institute review); VersataDev. Grp., Inc. v. SAP Am., Inc., 793 F.3d 1306, 1320 (Fed. Cir. 2015)(holding that the initial eligibility determination for CBM review isappealable, despite same bar); see generally 35 U.S.C.[section][section] 314(d), 315(b), 324(e) (2012).

(176.) See, e.g., Secure Axcess, LLC v. PNC Bank Nat'lAss'n, 848 F.3d 1370, 1373 (Fed. Cir. 2017) (holding that the PTABhad adopted an impermissibly broad definition of which patents areeligible for CBM review), cert. granted, 138 S. Ct. 1982 (2018); UnwiredPlanet, LLC v. Google Inc., 841 F.3d 1376, 1379 (Fed. Cir. 2016).

(177.) See, e.g., Shaw Indus. Grp., Inc. v. Automated Creel Sys.,Inc., 817 F.3d 1293, 1299-300 (Fed. Cir. 2016) (holding that estoppeldoes not apply to invalidity challenges rejected by the PTAB at theinstitution stage).

(178.) See, e.g., Nantkwest, Inc. v. Iancu, 898 F.3d 1177, 1180(Fed. Cir. 2018) (en banc) (holding that 35 U.S.C. [section] 145 (2012),which entitles the USPTO to collect "[a]ll the expenses" ofdefending against administrative challenges in district court, does notinclude attorneys fees).

(179.) Rovalma, S.A. v. Bohler-Edelstahl GmbH & Co. KG, 856F.3d 1019, 1029-30 (Fed. Cir. 2017).

(180.) Ethicon Endo-Surgery, Inc. v. Covidien LP, 826 F.3d 1366,1368-69 (Fed. Cir. 2016) (Newman, J., dissenting); see also VirnetX Inc.v. Apple Inc., 665 F. App'x 880, 890 (Fed. Cir. 2016)(O'Malley, J., dissenting) ("[The PTAB] appears to have simply'substituted its own expertise for record evidence.'...Allowing the Board to continue this practice... would only exacerbatethe trend towards a 'haze of so-called expertise' that thiscourt and the Supreme Court have admonished against.").

(181.) See, e.g., Pitts, supra note 6 (quoting then-Chief Judge ofthe Federal Circuit Randall Rader, referring to PTAB panels as"death squads killing property rights").

(182.) 35 U.S.C. [section] 6 (2012).

(183.) Gene Quinn, PTAB Chief Judge Defends APJs, IPWATCHDOG (Mar.8, 2018), http://www.ipwatchdog.com/2018/03/08/ptab-chief-judge-defends-apjs/id=94528/ [https://perma.cc/FCP6-TYZV] (statement by PTAB ChiefJudge David Ruschke).

(184.) See, e.g., Brief Amicus Curiae of the Patent Trial andAppeal Board Bar Ass'n in Support of Neither Party at 6, Oil StatesEnergy Servs., LLC v. Greene's Energy Grp., LLC, 138 S. Ct. 1365(2018) (No. 16-712) ("These judges have special technical and legalexpertise, and at least one of them typically has a technical backgroundand work experience related to the subject matter of the patent inquestion."); USPTO, ORGANIZATIONAL STRUCTURE AND ADMINISTRATION OFTHE PATENT TRIAL AND APPEAL BOARD 2,https://www.uspto.gov/sites/default/files/documents/Organizational%20Structure%20of%20the%20Board% 20May%2012%202015.pdf[https://perma.cc/9BRK-8987] ("The Vice Chief Judges each manage adivision consisting of judges and patent attorneys. Currently, there aresix sections in each division.... Each section covers a broad technicalfocus....").

(185.) Though a rough measure, the Federal Judicial Centerindicates that, of 1305 sitting federal judges, only 16% (209) possess aminimum of an undergraduate science degree (B.S., B.S.E., B.S.E.E.,B.Sc., A.S., A.A.S.). Biographical Directory of Article III FederalJudges, 1789-Present, FED. JUDICIAL CTR.,https://www.fjc.gov/history/judges [https://perma.cc/M7HV-KWJP]. Seegenerally Charlie Stiernberg, Science, Patent Law, and EpistemicLegitimacy, 27 HARV. J.L. & TECH. 279, 299 (2013) (noting "thelack of technically trained judges" as "measurabl[y]impact[ing] the patent system"); Peter Lee, Patent Law and the TwoCultures, 120 YALE L.J. 2, 6-7 (2010).

(186.) See supra Table 18.

(187.) See supra Table 6.

(188.) See supra Table 17.

(189.) To wit, the PTAB's advantage over district courts isactually higher in the mechanical category as compared to chemical orelectrical patent cases. See id.

(190.) See, e.g., Roy E. Hofer & Joshua H. James, Supreme CourtReversal Rates for Federal Circuit Cases, 6 LANDSLIDE 40 (2014) (findingan 83.3% reversal rate, the highest of all thirteen circuits, for the1999-2008 period).

(191.) See, e.g., Tejas N. Narechania, Certiorari, Universality,and a Patent Puzzle, 116 MICH. L. REV. 1345, 1349 (2018) (finding"[in]consistency across substantive fields of law" to be asubstantial factor in granting cert on Federal Circuit cases); PeterLee, The Supreme Assimilation of Patent Law, 114 MICH. L. REV. 1413,1453-56 (2016); Robin Feldman, Ending Patent Exceptionalism andStructuring the Rule of Reason: The Supreme Court Opens the Door forBoth, 15 MINN. J.L. SCI. & TECH. 61 (2014). This trend in SupremeCourt jurisprudence is, in many ways, not necessarily unique to patentlaw either. See, e.g., Cynthia Estlund, Are Unions a ConstitutionalAnomaly?, 114 MICH. L. REV. 169 (2015) (regarding anti-exceptionalism inlabor law); Abigail R. Moncrieff, Understanding the Failure ofHealth-Care Exceptionalism in the Supreme Court's ObamacareDecision, 142 CHEST 559 (2012) (regarding anti-exceptionalism in healthcare law); Kristin E. Hickman, Agency-Specific Precedents: RationalIgnorance or Deliberate Strategy?, 89 TEX. L. REV. 89, 108-10 (2010)(regarding anti-exceptionalism in tax law).

(192.) See, e.g., Octane Fitness, LLC v. Icon Health & Fitness,134 S. Ct. 1749, 1755-58 (2014) (rejecting the Federal Circuit'sspecialized test for determining "exceptional" cases underfee-award provisions in favor of "ordinary meaning" and the"general 'American rule'"); Medlmmune, Inc. v.Genentech, Inc., 549 U.S. 118, 132-37 (2007) (rejecting the FederalCircuit's patent-specific test for "actual controversy"under the Declaratory Judgment Act in favor of generalist precedent);eBay Inc. v. MercExchange, LLC, 547 U.S. 388, 393-94 (2006) (rejectingthe Federal Circuit's "general rule... [of] issu[ing]permanent injunctions against patent infringement," mandating useof the more typical four-factor test instead); Independent Ink, Inc. v.Illinois Tool Works Inc., 547 U.S. 28, 40-42 (2006) (rejecting theFederal Circuit's rule presuming "market power" inantitrust tying cases involving patented products).

(193.) Sapna Kumar, The Accidental Agency?, 65 FLA. L. REV. 229,232 (2013).

(194.) See, e.g., Christopher J. Walker, Chevron Deference andPatent Exceptionalism, 65 DUKE L.J. ONLINE 149, 157 (2016); Stuart MinorBenjamin & Arti K. Rai, Administrative Power in the Era of PatentStare Decisis, 65 DUKE L.J. 1563, 1563 (2016) ("PTAB decisionmakingcould be structured in a manner that should, under conventionaladministrative law principles, merit Chevron deference. In alllikelihood, the chief roadblock to Chevron is not formal administrativelaw, but specific challenges within the patent regime.").

(195.) For previous scholarly works criticizing the Federal Circuitfor appearing to lack appropriate deference to juries, see William C.Rooklidge & Matthew F. Weil, Judicial Hyperactivity: The FederalCircuit's Discomfort with its Appellate Role, 15 BERKELEY TECH.L.J. 725, 739-40 (2000); Ted D. Lee & Michelle Evans, The Charade:Trying a Patent Case to All "Three" Juries, 8 TEX. INTELL.PROP. L.J. 1, 14 (1999); Gregory D. Leibold, In Juries We Do Not Trust:Appellate Review of Patent-Infringement Litigation, 67 U. COLO. L. REV.623, 625-26 (1996); Arti K. Rai, Engaging Facts and Policy: AMulti-Institutional Approach to Patent System Reform, 103 COLUM. L. REV.1035, 1056 (2003). Indeed, the jury affirmance rate presented in thisarticle--92.3%--is a notable increase compared to previous empiricalstudies. See, e.g., Moore, Judges, Juries, and Patent Cases, supra note14, at 397 (finding a jury affirmance rate of 78%, based on 1983-1999data).

(196.) See supra note 143 and accompanying text.

(197.) See, e.g., Kevin Casey et al., Standards of Appellate Reviewin the Federal Circuit: Substance and Semantics, 11 FED. CIR. B.J. 279,307-08 (2002) ("The foundation for jury trials in civil litigationis, of course, the Seventh Amendment to the Constitution. Appellatechallenges to jury findings rarely succeed, because the SeventhAmendment proscribes review of such findings even more than [FederalRules of Civil Procedure] 52 restricts review of trial court findings offact."); Edward H. Cooper, Civil Rule 50(A): Rationing andRationalizing the Resources of Appellate Review, 63 NOTRE DAME L. REV.645, 650 (1988) ("Judge findings are accorded somewhat lessdeference than jury findings, at least in common comparisons of theclear error standard with directed verdict standards."). But seeKevin M. Clermont & Theodore Eisenberg, 3 AM. L. & ECON. REV.125 (2001) (finding that jury deference is not equally leveraged acrossall fields of law).

(198.) See supra notes 84-89 and accompanying text.

(199.) See supra Tables 9, 10, 11.

(200.) See generally Michael Goodman, What's So Special AboutPatent Law?, 26 FORDHAM INTELL. PROP. MEDIA & ENT. L.J. 797, 800(2016) ("Some proposals focus on the development of legal--ratherthan technical--expertise, thereby running up against the Court'srecent declarations about the lack of need for specialization of thattype.").

(201.) See supra notes 33-36 and accompanying text.

(202.) H.R. REP. NO. 112-98, Part 1, at 38-40 (2011) (emphasesadded).

(203.) See, e.g., 153 CONG. REC. H23,941 (2007) ("The rapidpace of innovation and increasingly complex patent filings have strainedthe Patent and Trademark office and patent claims of questionablevalidity have been granted."); Press Release, Office of the PressSec'y, Executive Actions: Answering the President's Call toStrengthen Our Patent System and Foster Innovation (Feb. 20, 2014)https://obamawhitehouse.archives.gov/the-press-of-fice/2014/02/20/fact-sheet-executive-actions-answering-president-s-call-strengthen-our-p[https://perma.cc/46JT-AN4K]) ("The United States Patent andTrademark Office (USPTO) has now successfully implemented [the AIA],enabling many reforms that are leading to higher patent quality,including post-grant patent review proceedings at the USPTo....").Joe Matal's two-part guide to the legislative history of the AIA isan exquisitely comprehensive resource from an insider'sperspective, and links the creation of specific provisions directly tothe need to combat patent trolls. See Joe Matal, A Guide to theLegislative History of the America Invents Act: Part II of II, 21 FED.CIR. B.J. 539, 592 (2012) ("Section 299 should put an end to apractice that had become a favorite tactic of patent trolls: suing alarge number of unrelated patent defendants in a single action.").At the time of authorship, Joe Matal was the Judiciary Committee Counselto Senator Jon Kyl, one of the principal architects of the AIA. UntilJanuary 2019, Joe Matal was the Acting Deputy General Counsel andSolicitor for the USPTO. About Us, USPTO,https://www.uspto.gov/about-us/executive-biographies/joseph-matal[https://perma.cc/5CLF-WHD9].

(204.) To wit, half of the currently active judges on the FederalCircuit were either appointed contemporaneously with the AIA'spassage and implementation, or shortly thereafter: Judges Jimmie Reyna,Evan Wallach, Richard Taranto, Raymond Chen, Todd Hughes, and KaraStoll. See Judges, U.S. COURT OF APPEALS FOR THE FED. CIR.,http://www.cafc.uscourts.gov/judges [https://perma.cc/N3C7-HB6W].

(205.) MCM Portfolio LLC v. Hewlett-Packard Co., 812 F.3d 1284,1290 (Fed. Cir. 2015).

Table 1: Disposition of PTAB Validity Findings on Appeal (Unadjusted)Affirm Rate Reverse Rate Vacate and/or Observations Remand Rate83.5% 6.6% 9.9% 771Table 2: Disposition of PTAB Validity Findings (Partially Adjusted)Affirm Rate Reverse Rate Vacate and/or Observations Remand Rate86.4% 6.9% 6.7% 745Table 3: Disposition of PTAB Validity Findings by Issue (PartiallyAdjusted)Validity Question Affirm Reverse Vacate ObservationsIssue Rate Rate and/or Remand RateSubject-Matter Law 90.0% 6.7% 3.3% 30EligibilityDefiniteness Law 66.7% 0% 33.3% 3Obviousness Mixed 87.5% 5.5% 7.0% 456 Law/FactEnablement Mixed 100% 0% 0% 5 Law/FactAnticipation Fact 84.0% 9.5% 6.5% 231Written Fact 85% 10% 5% 20DescriptionTable 4: Disposition of PTAB Validity Findings by Question Type(Partially Adjusted)Question Affirm Reverse Vacate ObservationsType Rate Rate and/or Remand RateLaw 87.9% 6.1% 6.1% 33Mixed 87.6% 5.4% 7.0% 461Law/FactFact 84.1% 9.6% 6.4% 251Table 5: Disposition of PTAB Validity Findings by Question TypeQuestion Affirm Reverse Vacate ObservationsType Rate Rate and/or Remand RateLaw 87.9% 6.1% 6.1% 33Mixed 91.2% 4.8% 4.1% 441Law/FactFact 91.7% 4.8% 3.5% 229Table 6: Disposition of PTAB Validity Findings by Technology ClassTechnology Affirm Percentage of ObservationsClass Rate Cases(1) Chemical 90.6% 7.5% 53(2) Computers & Communications 91.2% 48.4% 340(3) Drugs & Medical 87.8% 12.8% 90(4) Electrical & Electronics 91.9% 10.5% 74(5) Mechanical 89.7% 11.1% 78(6) Other 97.1% 9.7% 68Table 7: Disposition of PTAB Validity Findings by Proceeding TypeProceeding Type Affirm Rate Percentage of Cases ObservationsApplication 90.1% 10.1% 71Inter Partes Reexam 95.0% 19.8% 139Ex Parte Reexam 91.7% 3.4% 24Interference 85.0% 2.8% 20Pre-AIA Combined 92.5% 36.1% 254Inter Partes Review 90.8% 59.0% 415CBM Review 85.3% 4.8% 34Post-AIA Combined 90.4% 63.9% 449Table 8: Disposition of District Court Validity Findings on Appeal(Unadjusted)Affirm Rate Reverse Rate Vacate and/or Remand Rate Observations83.5% 12.3% 4.1% 316Table 9: Disposition of District Court Validity Findings by DistricttypeDistrict Type Affirm Rate Percentage of Cases ObservationsPatent-Heavy Districts 86.3% 49.6% 131Standard Districts 83.5% 50.4% 133Table 10: Disposition of District Court Validity Findings by JudgetypeJudge Type Affirm Percentage of Observations Rate CasesPatent-Heavy District Court Judges 82.2% 38.3% 101Standard District Court Judges 86.5% 61.7% 163Table 11: Disposition of District Court Validity Findings by Judge andDistrict TypeJudge/District Type Affirm Rate Data Share ObservationsPatent-Heavy Judges or Districts 84.9% 52.7% 139Standard Judges and Districts 84.8% 47.4% 125Table 12: Disposition of District Court Validity Findings by IssueValidity Question Affirm Reverse Vacate ObservationsIssue Type Rate Rate and/or Remand RateSubject-Matter Law 89.9% 8.6% 1.4% 139EligibilityDefiniteness Law 81.0% 19.0% 0% 42Obviousness Mixed 87.0% 10.9% 2.2% 46 Law/FactAnticipation Fact 60.0% 20.0% 20.0% 25Written Fact 81.8% 18.2% 0% 11DescriptionTable 13: Disposition of District Court Validity Findings by QuestionTypeQuestion Type Affirm Reverse Vacate and/or Observations Rate Rate Remand RateLaw 87.9% 11.1% 1.1% 181Mixed 87.2% 10.6% 2.1% 47Law/FactFact 66.7% 19.4% 13.9% 36Table 14: Disposition of District Court Validity Findings by TechnologyClassTechnology Class Affirm Percentage of Observations Rate Cases(1) Chemical 85.7% 5.3% 14(2) Computers & Communications 85.7% 53.2% 140(3) Drugs & Medical 81.8% 20.9% 55(4) Electrical & Electronics 88.2% 6.5% 17(5) Mechanical 79.0% 7.2% 19(6) Other 88.9% 6.8% 18Table 15: Disposition of Validity Findings by Trial-Level ForumTrial-Level Affirm Reverse Vacate and/or Remand ObservationsForum Rate Rate RatePTAB 91.2% 4.8% 4.0% 703District Court 84.9% 12.1% 3.0% 264Table 16: Disposition of Validity Findings by Question Type andTrial-Level ForumQuestion Trial-Level Affirm Non-Affirm ObservationsType Forum Rate RateLaw PTAB 87.9% (29) 12.1% (4) 214 Fisher's Exact Test: p=1.00 District 87.9% (159) 12.2% (22) CourtsMixed PTAB 91.2% (402) 8.8% (39) 488Law/Fact Fisher's Exact Test: p=0.42 District 87.2% (41) 12.8% (6) CourtsFact PTAB 91.7% (210) 8.3% (19) 265 Fisher's Exact Test: p=0.0002 District 66.7% (24) 33.3% (12) CourtsTable 17: Disposition of Validity Findings by Technology Class andTrial-Level ForumTechnology Trial-Level Affirm Non-Affirm ObservationsClass Forum Rate Rate(1) Chemical PTAB 90.6% (48) 9.4% (5) 67 Fisher's Exact Test: p=0.6304 District 85.7% (12) 14.3% (2) Courts(2) Computers & PTAB 91.2% (310) 8.8% (30) 480Communications Chi-Squared Test: p=0.075 District 85.7% (120) 14.3% (20) Courts(3) Drugs & PTAB 87.8% (79) 12.2% (11) 145Medical Chi-squared Test: p=0.3225 District 81.8% (45) 18.2% (10) Courts(4) Electrical & PTAB 91.9% (68) 8.1% (6) 91Electronics Fisher's Exact Test: p=0.64 District 88.2% (15) 11.8% (2) Courts(5) Mechanical PTAB 89.7% (70) 10.3% (8) 97 Fisher's Exact Test: p=0.2429 District 79.0% (15) 21.1% (4) Courts(6) Other PTAB 97.1% (66) 2.9% (2) 86 Fisher's Exact Test: p=0.1917 District 88.9% (16) 11.1% (2) CourtsTable 18: Disposition of Validity Findings by Question Type,Trial-Level Forum, and Claim DispositionQuestion Tribunal Affirm Rate Affirm RateType Below (Claim-Invalidating) (Claim-Upholding)Over-all PTAB 93.2% 80.4% (551) (90) District 84.6% 85.7% Courts (170) (54)Law PTAB 93.6% 0.0% (29) (0) District 88.7% 81.8% Courts (141) (18)Mixed PTAB 92.6% 82.8% (349) (53) District 773% 96.0% Courts (17) (24)Fact PTAB 94.5% 80.4% (173) (37) District 60.0% 75.0% Courts (12) (12)Question Invalid-Upheld ObservationsType Affirmance GapOver-all +12.8% 703 Chi-Squared Test: p<0.0001 -1.1% 264 Chi-Squared Test: p=0.8261Law +93.6% 33 Fisher's Exact Test: p=0.0114 +6.9% 181 Fisher's Exact Test: p=0.3155Mixed +9.8% 441 Chi-Squared Test: p=0.011 -18.7% 47 Fisher's Exact Test: p=0.0848Fact +14.1% 229 Fisher's Exact Test: p=0.0048 -15.0% 36 Fisher's Exact Test: p=0.4815Table 19: Logistic Regression Models of Appellate AffirmanceVariables Unadjusted Model Aadjusted Model Odds-Ratio 95% P-Trend Odds-Ratio CITribunal BelowDistrict 1.00 Referent 1.00CourtPTAB 1.85 1.21-2.83 0.0047 1.85Question of LawNo 1.00 Referent 1.00Yes 0.81 0.51-1.30 0.3880 1.12Question of FactNo 1.00 Referent 1.00Yes 0.85 0.54-1.33 0.4748 0.84Disposition of ClaimInvalidated 1.00 Referent 1.00Upheld 0.46 0.29-0.72 0.0008 0.52Case YearFY2016 1.00 Referent 1.00FY2015 0.55 0.37-0.84 0.0057 0.57HJT-ChemicalNo 1.00 Referent 1.00Yes 1.01 0.45-2.28 0.9780 0.46HJT-Computer &CommunicalionsNo 1.00 Referent 1.00Yes 1.03 0.68-1.55 0.8949 0.40HJT-Drug & MedicalNo 1.00 Referent 1.00Yes 0.65 0.39-1.08 0.0965 0.37HJT-Elections &ElectronicNo 1.00 Referent 1.00Yes 1.25 0.59-2.66 0.5674 0.46HJT-MechanicalNo 1.00 Referent 1.00Yes 0.82 0.43-1.55 0.5382 0.31Variables Aadjusted Model 95% P-Trend CITribunal BelowDistrict ReferentCourtPTAB 1.01-3.39 0.0462Question of LawNo ReferentYes 0.54-2.32 0.7613Question of FactNo ReferentYes 0.51-1.38 0.4826Disposition of ClaimInvalidated ReferentUpheld 0.32-0.85 0.0091Case YearFY2016 ReferentFY2015 0.37-0.87 0.0093HJT-ChemicalNo ReferentYes 0.13-1.69 0.2435HJT-Computer &CommunicalionsNo ReferentYes 0.14-1.16 0.0922HJT-Drug & MedicalNo ReferentYes 0.12-1.15 0.0854HJT-Elections &ElectronicNo ReferentYes 0.13-1.61 0.2228HJT-MechanicalNo ReferentYes 0.09-1.00 0.0504Table 20: Disposition of Validity Findings by Trial-Level Forum and YearTrial-Level FY 2015 FY 2016 ObservationsForum Affirm Rate Affirm RatePTAB 88.8% 93.0% 703 (276) (365) Chi-Squared Test: p=0.0426District Courts 81.6% 88.3% 264 (115) (109) Chi-Squared Test: p=0.1106Table 21: Dissent Rate by Trial-Level ForumTrial-Level Dissent Rate ObservationsForumPTAB 5.0% 703 (35)District Courts 6.9% 264 (18)Table 22: Disposition of PTAB Validity Findings by Proceeding typeProceeding Type Affirm Rate Percentage Observations of CasesInter Partes Review 90.8% 59.0% 415 (377)Inter Partes Reexam 95.0% 19.8% 139 (132)Ex Partes Reexam 91.7% 3.4% 24 (22)CBM Review 85.3% 4.8% 34 (29)Forum-Choice 91.5% 87.0% 612Combined (560)Application 90.1% 10.1% 71 (64)Interference 85.0% 2.8% 20 (17)No Choice Combined 89.0% 12.9% 91 (81)