"Does Judicial Ideology Affect Copyright Fair Use Outcomes?: Evidence from the Fair Use Case Law,"

especially strong. Empirical work in other areas of the law, however, such as tax 7 and securities 8 law, has less commonly found evidence that judges' ideological preferences affect outcomes. 9 As Sag, Jacobi, and Stych note, this has led to speculation that in these primarily economic areas of the law, judges' (and our own) underlying ideological beliefs-regarding, for example, the proper relation between liberty and equality or between the state and society-simply do not clearly favor one outcome over another. 1 0 Consider, then, Figure 1. Since the passage into law of the Copyright Act of 1976, the Supreme Court has heard four cases involving section 107 of the act, which establishes the affirmative defense of copyright fair use. 1 ' For each of these cases, Figure 1 arranges along a liberal-conservative continuum each justice's Martin-Quinn ideal point estimate 12 (a leading quantitative measure of Supreme Court justices' underlying ideological preferences) for the term in which he or she voted in the case; a greater ideal point estimate value indicates a more conservative ideology. The figure also shows whether the justice voted in favor of or against a finding of fair use-or otherwise did not address the fair use issue in his or her vote. Clearly, there is no relation in these cases between the justice's ideological position and his or her votes. In Sony v. Universal City Studios, Inc., 13 for example, Justices Marshall and Rehnquist voted together in dissent (something which, according to the Spaeth database, 14 they did exactly 31 times out 6,058 opportunities to do so), while in Campbell v. Acuff-Rose Music, Inc., 15 Justices Stevens and Thomas voted together along with the rest of the unanimous court (the Spaeth database suggests that the two vote together either in the majority or in dissent about 51% of the time Is copyright fair use, then, a non-ideological or at least "ideologically ambiguous"' 18 area of American law? Notwithstanding its importance to freedom of expression and the viability of the public domain, is fair use, as a matter of ideological judging, more akin to securities or bankruptcy law than to civil liberties or environmental law? Sag, Jacobi, and Stych have applied rigorous statistical analysis to Supreme Court intellectual property case law to show that judicial ideology does in fact affect outcomes in that case law. 19 But is copyright fair use an exception to this general finding, not only at the level of the Supreme Court but across the circuit and district courts? This brief paper seeks to answer that question empirically.

Part I describes the data set used for the study. Part II reports the study's results. It shows that judges' ideological preferences have no significant effect on their adjudication of the fair use defense.

In establishing the affirmative defense of fair use, section 107 of the Copyright Act sets forth four factors that judges "shall" consider in determining whether to find fair use. 20 For a previous study, 21 I developed a data set describing all 306 reported federal court opinions that made substantial use 22 of the section 107 fourfactor test in adjudicating a defense of fair use from the January 1, 1978 effective date of the Copyright Act through 2005.23 For this study, I have excluded from that dataset four opinions written by magistrate judges and an additional nine opinions that found outstanding issues of fact on the fair use issue or issued mixed rulings in which some uses were found to constitute fair use while others were found not to constitute fair use. This left a total of 293 opinions. Overall, including votes cast by judges who joined an opinion, these 293 opinions yielded a total of 454 votes either in support of a finding of fair use or in support of a finding of no fair use. Two of these votes were excluded from consideration because they were cast by International Court of Trade judges sitting by designation. This left a total of 452 votes.

For each judge casting at least one of these votes, I included in the new data set his or her values from various widely-accepted-but by no means noncontroversial 24 -indices of judicial ideology. Specifically, I included where possible: (1) the political party of the judge's appointing president, 25 (2) the Poole Common Space score of the judge's appointing president, as developed by Keith Poole and Howard Rosenthal, 26 (3) the judge's Judicial Common Space score, as developed by Lee Epstein, Andrew D. Martin, Jeffrey A. Segal and Chad Westerland, 27 building on work by Michael Giles, Virginia Hettinger, and Todd Peppers, 28 and (4) the judge's NIXONI score, as developed by Robert Howard and David Nixon.

The previous study reviewed various summary statistics relating to the original data set of 306 opinions. 30 I provide here only those details about the revised data set that may help the reader to evaluate the findings reported below.

Table 1 reports that of the 452 votes studied, 32 were cast by Supreme Court judges, 218 by circuit court judges, and the remaining 202 by district court judges. Of these 452 votes, 193 (42.7%) were cast in support of a finding of fair use, while 259 (57.3%) were cast in support of a finding of no fair use. These proportions do not vary significantly among the three levels of courts being studied.

As for the overall ideological distribution of the 452 votes studied, 260 (57.5%) were cast by judges appointed by a Republican president, while 192 (42.5%) were cast by judges appointed by a Democratic president. Figure 2 provides a more nuanced profile of the ideological distribution of the votes. For the 441 votes cast by judges whose NIXONI score is available, the figure shows the distribution of those votes by their judge's score. As with the Martin-Quinn ideal point estimate, a greater NIXONI score indicates a more conservative judge. Consistent with the distribution of votes by the party of their judge's appointing president, the mean 24. For a discussion of the controversy surrounding the measurement of judicial ideology, see Sisk & Heise, supra note 1.

25. On the efficacy of this measure of judicial ideology, see id. But see Orley Ashenfelter, Theodore Eisenberg & Stewart J. Schwab, Politics and the Judiciary: The Influence of Judicial Background on Case Outcomes, 24 J. LEGAL STUD. 257, 260 (1995) (arguing that presidential politics do not explain outcomes). NIXONI score of these 441 votes was a slightly conservative 0.073 (std. dev.=0.255). 31 Thus, though the votes studied are slightly skewed towards a finding of no fair use and their judges towards a conservative ideology, the data set is sufficiently balanced to allow for a relatively straightforward statistical analysis of the effect of judicial ideology on fair use outcomes.

Regardless of which measure of judicial ideology is used, the data show no significant relation between a judge's ideology and her adjudication of the fair use defense. Specifically, there is no relation between judicial ideology and a judge's likelihood of finding fair use or no fair use, nor is there any relation between judicial ideology and how judges treated various factor and subfactor considerations that, as the previous study showed, 32 typically drive the fair use analysis.

As an initial matter, Table 2 sets forth pairwise correlation coefficients for the relations among five variables in the data set: the four measures of a judge's ideological preferences and a binary variable indicating whether or not the judge found fair use. As expected, the various measures of judicial ideology correlated very strongly with each other. However, none show a significant correlation with judges' fair use findings.

Of course, correlation analysis does not control for the facts of the cases. Table 3 reports the results of logistic regression of the outcome of each of the 441 votes for which their judge's NIXONI score is available on various factual findings made by the judge (or the opinion the judge joined) and the NIXONI score of the judge casting the vote. Here again, judicial ideology appears to exert no significant influence over judges' willingness to find fair use or no fair use. Each of the other three indices of judicial ideology included in the data set similarly show no significant influence.

With respect to the 18 votes cast in dissent, of which 11 were cast in favor of a finding of fair use, none support the proposition that the court was split along ideological lines. Each of these 18 dissenting votes was cast by a judge who shared his or her ideological sign (-/+), as established by his or her NIXONI score, with at least one judge in the majority.

Finally, judicial ideology showed no significant relation with any of the numerous factor and subfactor findings made in the opinions studied. For example, judicial ideology did not help to explain whether a judge was likely to conduct a commercial use inquiry or whether a lower court judge was likely to ignore relevant Supreme Court precedent-a phenomenon explored in detail in the 31. For the 449 votes cast by judges for whom the Poole Common Space Score of their appointing president was available, the mean Poole Common Space Score was 0.094 (std. dev.=0.494). The comparable mean for the Judicial Common Space score was 0.013 (n-227, std. dev.=0.350).

It may be encouraging to learn that copyright fair use is not an area of the law in which judicial ideology appears to influence adjudication-and thus that copyright fair use is an exception to Sag, Jacobi, and Stych's more general findings that judicial ideology affects intellectual property outcomes, at least before the Supreme Court. 34 Yet one might tentatively observe that there is also something disturbing about these results. They are disturbing because copyright fair use should be ideological. As intellectual property scholars have long recognized, the stakes involved in fair use adjudication are immense. Fair use outcomes define the contours of the private and public domains of human expression and, in doing so, directly impact our capability for human flourishing. 35 Fair use is far more than an economic area of the law calling for the post-ideological balancing of costs and benefits; it goes to the core of what constitutes a good society. 36 Thus, while the "politics of intellectual property" 37 that James Boyle and others called for many years ago apparently has yet to make itself felt among the ranks of the federal judiciary, one might tentatively hope that some future study of this nature may yield different results.