Previously posted on SSRN.

Underlying datasets available at


What do judges really care about? Scholars have used various methods to identify a judge’s policy preferences. The standard method in political science, called the Martin-Quinn score, counts a judge’s votes for conservative or liberal outcomes. But judges don’t just vote, they give reasons in written opinions. Reason-giving is not only part of the tradition of common-law decision making but is also central to rule-of-law ideals, concerns that are not the focus most empirical methodologies. What’s more, the reasons a judge gives for reaching a conclusion provide powerful evidence for what the judge herself cares about. That is especially the case when the judge writes a concurring or dissenting opinion, free of the demands of drawing together and speaking for a court majority.
This Article is the first to apply a novel empirical method—citation network analysis—to particular appellate jurists’ separate judicial opinions (e.g., concurrences, dissents) in an effort to provide a more detailed picture of a judge’s ideological preferences. It focuses on the separate opinions of Justices Scalia and Thomas through the end of October Term 2019: they served for a similar number of years, wrote separately at similarly high rates, and were one another’s closest ideological fellow-travelers, which somewhat controls for their Martin-Quinn scores.
The findings I report suggest a bright future for citation network analysis in legal studies. Despite the similarity of their Martin-Quinn score, the two justices had demonstrably different ideological priorities. Justice Scalia was much more focused on curtailing Eighth Amendment review of state death-penalty regimes and on deconstitutionalizing prisoner and sex-related liberty and equality claims. Justices Thomas, by contrast, has focused on vindicating a judge-run, formalist federalism that places the states’ police powers beyond federal oversight. No handful of cases from one or another doctrinal silo can fully capture these distinctive projects. No vote-tallying unidimensional metric can reveal it either. Network analysis, however, does.