Diffusion of Intellectual Property Precedent

Amy Semet, Center for the Study of Democratic Politics, Princeton University

Introduction

In examining judicial influence, scholars have traditionally focused on the case outcome to determine court ideology and influence.[1] However, in recent years, scholars have increasingly employed new statistical techniques such as network analysis and textual analysis in order to gain a greater appreciation of actual opinion content examining how judges reach their decisions and what precedent they rely on to support their analysis.[2]Much of this work has been done by judicial politics scholars who often fail to consider the intricacies of how law operates and how cases may differ depending by legal issue. Moreover, most of the work in this regard focuses on hot button civil rights and procedures issues. Application of these techniques to the more intricate and difficult to understand study of intellectual property law is left wanting. This multi-article or book project represents an attempt to merge the legalistic aspects of case analysis with the analytical tools of judicial politics to gain a better understanding of how courts develop intellectual property law precedent.

In this project, I aspire to use these more sophisticated statistical tools to gain a greater appreciation of how intellectual property law doctrine diffuses in the judicial system. I seek to answer the following questions: 1) How do courts cite precedent? In patent law, for instance, do district courts favor citation of Supreme Court over Federal Circuit precedent? How in turn does the Federal Circuit cite precedent, that is, does it favor certain district courts over others?; 2) How do higher appellate court induce compliance among the lower courts in their citation of precedent? Scholars have often debated whether the Federal Circuit optimally transmits its precedent to the lower court.[3] Using textual analysis, what can we say about how well both the Supreme Court and appellate courts transmit precedent?; 3) Finally, while most of my analysis focuses on patent law, we can gain much from using these same techniques to gain a better appreciation of how precedent diffusion for patent law (which has a centralized appellate court) differs from that of copyright or trademark law (where appellate courts are heard regionally). Such analysis lends insight into whether specialized appellate courts work better than diffuse, regional courts in inducing compliance among the lower judiciary. Answers to these questions also offer normative guidance into how intellectual property law is formed and whether institutional structure influences how precedent is diffused.

How Do Courts Cite and Use Precedent in Patent Cases?

Questions: 1) As compared to the Supreme Court, do district courts find Federal Circuit precedent more authoritative in reaching their conclusions in patent law cases? In turn, how does the Federal Circuit treat the citation of lower court district court decisions? Does it favor some district courts over others?Do district court find other district courts more persuasive than others in patent cases?

Patent law is unique in the sense that a single appellate court – the Federal Circuit – exists to encourage uniformity in patent law.[4]Since the Supreme Court hears so few patent cases, the Federal Circuit is often the decisive authority for district courts to rely on for patent law doctrine, often being nicknamed the “Supreme Court of Patents.”[5] Nonetheless, since Supreme Court precedent technically controls and since the Supreme Court has been increasingly active in recent years in overturning Federal Circuit decisions, it is a worthwhile question to explore whether district courts rely more on the Federal Circuit or the Supreme Court in both citing precedent and in devising new precedent to apply to future cases. David Pekarek-Krohn and Emerson Tiller explored these issues, finding that in their study of about 250 opinions from 2007 to 2009 that federal district courts treat the Federal Circuit as more authoritative than the Supreme Court on patent law than they treat regional courts on copyright law.[6]

In addition to studying how the district courts apply precedent, I hope to explore inward citations of how the district courts in turn cite precedent among each other and how the Federal Circuit cites district court precedent. In 2011, Congress passed a patent pilot program whereby cases would be funneled to judges in the 14 districts hearing the most patent cases on a voluntary basis.[7]The pilot program assumes that judges with more experience will make better decisions and will adjudicate cases faster because the learning curve will be lower.[8]The chosen districts – including plaintiff-friendly Eastern District of Texas and tech-friendly Northern District of California – represent districts containing judges with the most experience in patent law.[9] Any district court within the top fifteen district courts in the year 2010 had the option to participate in the program, with district courts that opt in having at least three judges designated as “patent judges.”[10] As a theoretical matter, one might anticipate that precedent coming from those districts might differ from other districts, that is, both other district courts as well as the Federal Circuit might think that precedent coming from those district courts was more authoritative than precedent coming from district courts like the District of Vermont, a district that rarely hears patent cases.

To conduct the analysis, I am in the process of coding cases over a ten year period from 2006 to 2016. There have been some important changes in patent law over the time period, as the Supreme Court has made several substantive decisions affecting key issues in patent law. Moreover, the Supreme Court has increasingly both taken and overturned more Federal Circuit cases than in past years. Looking at cases over a long time frame – even if one randomly samples a given selection – allows one to leverage changes both before and after these procedural and substantive developments. Moreover, decisions over this time frame allow one to ascertain the before and after effects of the patent law pilot project which began in 2011.

Fortunately, Westlaw/Shepherds presents users with both citations as well as a Table of Authorities to easily glean both the depth by which the case was cited as well as the type of citation. Most legal analysis of case citations treats all citations the same, as scholars simply make up a Python script to gather citations in an automatic grab. However, only by reading the case can one really ascertain how the citation is used. For instance, the court may simply cite a case to make a procedural point, or they make use a case in a string cite as reference for clearly known patent law principles. By coding the cases for both how the case is cited (i.e, string cite, in depth discussion, used in support of argument, used to distinguish, used as a negative reference) as well as depth of discussion, one gets a better qualitative idea of how courts use precedent which in turns informs the quantitative analysis conducted later on.

In addition to coding for how cases are cited, one would also code for common attributes such as legal issue, procedural posture of the case, year, district, judge, ideology of the judge or panel, and other factors pertinent as control variables. In addition, one could also code for whether the opinion favors the patent owner or not.

There are a few different ways I could do the statistical analysis. One can employ the tools of network analysis in order to ascertain citation counts, with each judicial citation representing a latent judgement from the court about what legal principles are important in deciding the given case. After devising a graph of all the in and out citations, one could get a better sense of exactly what cases are the most influential in terms of both citations in and citations out. Specifically, I will create a measure of case centrality based on citation patterns using the techniques used by Fowler et al.in their analysis of United States Supreme Court opinions.[11] Network analysis maps out relationships between cases, using two key concepts: “nodes” (i.e., the people or information within a setting) and “links” (i.e., the relationship between nodes). Court cases are considered “nodes” within the network, linked together to other case nodes through citations to precedent.[12] “Links” take one of two forms: outward, referring to all precedents a case cites, and inward, referencing all subsequent opinions that cite that case.[13] A citation analysis analyzing the combination of links and nodes provides a way to measure case importance, referring to the central role that the particular case plays in influencing precedent. To create the system of nodes and links, we cite the citing case, the cited case, and any information we gleaned from the Shepherd’s analysis. As noted above, there are different ways you can do the analysis, depending on how you treat string cites or negative citations. Some scholars, like Fowler et al. argue that in ascertaining the legal relevance of the case, it is simply important to capture whether a case was cited rather than how it is cited (such as in a string cite).[14] I would employ both methods so as to analyze two conceptually different concepts, that is, case relevance (whether a case is cited, which would include all case cites) and alternatively, legal vitality or authoritativeness of the case (based on how the case is cited).[15]

We next need to get at a measure of centrality to measure case importance, using the methods used by Fowler et al. After mapping out all the inward and outward citations, we get a better picture of which cases are important and which are not. The total number of “links” between each node is the “degree.”[16] As Fowler describes, network theory posits a few ways in which to determine which cases are more or less central to the network. Degree centrality measures the number of citations from other cases (inward citations), while eigenvector centrality is an alternative measure that estimates simultaneously the importance of all cases in the network.[17] More recently, scholars have borrowed from computer science to use another measure of centrality that allows one to draw on both inward and outward citations to a greater degree.[18] Developed by Kleinberg, this procedure relies on both outwardly citations (which are cases that cite to many other relevant decisions, also known as a “hub”) and inwardly relevant citations (which are casesthat are widely cited by other widely cited cases, also known as “authority”).[19] Put simply, cases that are outwardly relevant cite many inwardly relevant cases and inwardly relevant cases cite many outwardly relevant cases.[20]By devising both outwardly and inwardly relevant scores one can identify precedents that are both influential (inwardly relevant) as well as those who has a strong basis in other precedent (outwardly relevant).[21] Moreover, we can also assess whether certain types of cases are more or less important as precedent. In their analysis of the U.S. Supreme Court, James Fowler and SangickJeon found that reversed cases tend to be more important as precedent, and that the cases that reversed those cases became some of the Court’s most important cases.[22] They also found that the Court is careful to ground overturned cases in past precedent.[23]Similarly, I would hypothesize that the findings may be similar here, with both the Supreme Court and the Federal Circuit taking care to envelope overturned cases in robust precedent. Moreover, we may find that overtime the relevance of precedent changes. Although the time span I use here may be too limited to measure such changes in time, it may be possible to capture some sharp breaks in precedent given that the Supreme Court has been so active in its review.

Alternatively, I could use a method similar to that employed in the working paper by Pekarek-Krohn and Epstein.[24] In their analysis, Pekarek-Krohn and Epstein compared Federal Circuit citations in patent cases to citations of copyright cases in the regional appellate court by measuring how the proportions different. Specifically, they measured two proportions, the case proportion (# Circuit Court Cases/# Supreme Court Case + #Circuit Court Cases) or the cite proportion (# Circuit Court Cites/# Supreme Court Cites + # Circuit Court Cites).[25] They then measured the weights for both the Supreme Court (Supreme Court Cites/Supreme Court Cases) and the Circuit Weight (Circuit Court Cites/Circuit Court Cases).[26] The case proportion measures what precedent the district court relied on, while the cite proportion measures how a case is used, in an attempt to capture how much the district court links the facts and reasoning of the case to that of the higher court.[27] The weights in turn measure how much emphasis a district court puts on the opinions of each institution.[28] I would expand beyond Pekarek-Krohn and Epstein by developing a number of different case proportions based on how cases are cited. For instance, I would have some case proportions or case citations employing only cases that were discussed in depth, thus eliminating string cite cases or cases in which the court is only citing the case for a well-known legal point. Similar to Pekerek and Epstein, I could conduct similar proportions for both patent and copyright cases, and I would also perhaps add trademark cases in there as well. After constructing the various proportions, one could use t-tests to assess the difference between Federal Circuit (patent) and regional circuit (copyright or trademark) cases to see whether there is a difference between the proportions. In addition, one could run regressions with the proportions and weight measures as the dependent variable to assess whether patent cases differ from copyright and trademark cases. In a similar analysis, Pekerek and Epstein found that the Federal Circuit is treated more authoritatively on patent law than the regional courts are on copyright law and that courts rely on Federal Circuit precedent more in pro-patentee cases than anti-patentee cases.[29]

How Does the Federal Appellate Courts Induce Compliance?

Questions: How does a specialized appellate court like the Federal Circuit induce compliance with its decisions?

After examining just how the courts cite and use precedent, I would then analyze how district courts comply with Federal Circuit precedent. Many scholars have critiqued the Federal Circuit for its weak enunciation of precedent, leading district courts to not apply precedent in a consistent fashion.[30]Further, compared to other appellate courts, the Federal Circuit is rumored to have one of the highest reversal rates, particularly with respect to decisions in which patent law claim construction is at issue.[31]While scholars have made the qualitative assessment that the Federal Circuit is a weak transmitter of its own precedent, few have analyzed it quantitatively. This part of the project hopes to provide an answer quantitatively.

To answer this question, I hope to borrow from the literature studying how opinion clarity at the United States Supreme Court induces compliance among the lower federal courts. Among all appellate courts, the Federal Circuit is most like the Supreme Court in that it is the single appellate court hearing patent cases, so one can easily apply this literature to the Federal Circuit. Indeed, the Federal Circuit caselaw offers the ideal mechanism in which to test the extent to which a specialized appellate court can induce compliance among the lower federal court judiciary. Such research contributes to several normative debates concerning how district courts should review cases involving complex technical matters that may be beyond the scope of a lay jury. Moreover, such research lends insight into the mechanism by which courts achieve legitimacy in a separation of powers system by clearly setting forth the roadmap by which appellate courts can serve as political principals guiding lower court decision making so as to ensure uniformity and accountability.

Scholars studying lower court compliance with the Supreme Court have recently begun using the tools of textual analysis to come up with a measure of opinion clarity, opining that the clearer the opinion, the more likely that the Supreme Court can induce compliance in the lower court. In a working paper, Ryan Owens and Patrick Wohlfarth analyze the role that opinion clarity in Supreme Court opinions has in inducing compliance in the lower federal courts.[32] First, they used developing techniques in reading comprehension to come up with various “scores” of Supreme Court opinion clarity.[33] Use of reading scores has become increasingly common in empirical legal scholarship more generally with scholars using scores to measure such things as the complexity of federal statutes or party briefs. As an example, some scholars used the Coleman-Liau readability index which is a composite measure of the length of the words contained within the text measured by the number of characters and the number of sentences within the text.[34] Using Shepherd’s, Owens and Wohlfarth then analyzed how subsequent federal circuit courts “treated” the Supreme Court decision, such as whether they responded in a positive manner (positively following a case), a neutral manner (explaining or harmonizing a case with precedent) or a negative matter (overruling, criticizing, questioning or limiting a case).[35] They concluded that when the Supreme Court articulates its precedent clearly, it increased the propensity of the lower court to comply.[36]