Search
Browse By Day
Browse By Time
Browse By Person
Browse By Mini-Conference
Browse By Division
Browse By Session or Event Type
Browse Sessions by Fields of Interest
Browse Papers by Fields of Interest
Search Tips
Virtual Exhibit Hall
Change Preferences / Time Zone
Sign In
X (Twitter)
Political scientists generally believe that if legal doctrine influences decision-making on the U.S. Supreme Court at all, it does so only through the form of monolithic decision structures known as "jurisprudential regimes" (Richards and Kritzer 2002; Scott 2006; Luse et al. 2009; Bartels 2009; Pang et al. 2012; Bartels and O’Geen 2015). On this view, such regimes function as relatively stable institutional constraints on ideological voting, and evidence for their influence can be detected by measuring macro changes in predetermined divisions of case outcomes over long periods of time.
In this paper, we argue that this perspective is theoretically impoverished and methodologically unsophisticated. We introduce a new theory of regimes -- conceptualizing them as linguistic, modular, and evolutionary -- and deploy the tools of machine learning to provide evidence for their existence and effect. Specifically, using word embeddings that we train on a corpus of the Court's own opinions, we first estimate the term-level strengths of 53 First Amendment free expression regimes over the period of 1946 to 2018. Then, using a combination of random forests and classical regression techniques, we assess the predictive and causal effects of these candidate regimes on the Court's voting behavior. Our findings suggest that jurisprudential regimes do in fact exist, but in a way that is more complex than conventional regime theory is capable of recognizing.