Partisan Opinions, but Common Language: Similarities in Topic Use by Appellate Judges
Abstract: As the final word on thousands of legal matters each year, appellate courts make some of the most impactful decisions in modern society. Understanding partisan behavior by their judges is therefore critical for the rule of law. However, judicial language is technical, making partisanship challenging to objectively measure and creating a unique opportunity for natural language processing. Using fine-tuned language embeddings from transformer models, we leverage the random assignment of individual judges to three-judge panels, and of those panels to cases, to causally estimate how discussion of legal topics on U.S. appellate courts differs across partisan environments. We show that while Democratic judges write more dispersed opinions, judges of both parties agree on average about the important topics in each legal case. Further, we demonstrate that mandatory bipartisanship does not reduce the range of topics considered. Judicial partisanship is thus driven by disagreements within legal issues rather than disputes about which issues apply. These results provide a clearer understanding of the structure of judicial language and open new directions for natural language processing research and impact.
Paper Type: Short
Research Area: NLP Applications
Research Area Keywords: NLP tools for social analysis, quantitative analyses of news and/or social media, legal NLP
Contribution Types: Data analysis
Languages Studied: English
Submission Number: 310
Loading