Big Data Finds Supreme Court Becoming Increasingly Idiosyncratic


March 31, 2017 – The Supreme Court of the United States is drifting away from its judicial roots, according to a computer analysis of texts used in court opinions. The research shows evidence of the nation’s top court becoming more distinctive in the language that it uses, compared to lower courts, a trend that could undermine the court’s legitimacy over time.

The analysis, conducted by researchers at the University of Virginia School of Law and Dartmouth College, examined the differences in language between Supreme Court and U.S. Court of Appeals opinions dating from 1951–2007. 

The research comes in the backdrop of United States Senate hearings on Judge Neil Gorsuch, President Donald Trump’s nominee to succeed Justice Antonin Scalia on the court. The study is available on the Social Science Research Network and is to be published in the Arizona Law Review.

“Similar to the judicial robe and other traditions, the Supreme Court’s writings are among its most defining characteristics. By using language that is less judicial in character, the court could compromise the special place it holds in American society,” said Michael Livermore, associate professor at the University of Virginia School of Law.

By comparing close to 300,000 court texts, the research uses big data and machine text analysis to demonstrate that the Supreme Court “may be edging away from its roots”. According to the report, the growing distinctiveness in the language of Supreme Court opinion could cause lower courts to be less likely to cite those opinions over time.

“The Supreme Court’s opinions are central to its legitimacy. These findings raise practical questions about the continued prestige of the institution at a particularly fraught time in its history,” said Livermore.

The study found that:

  • Supreme Court writings have drifted away from their judicial roots by becoming more distinctive in recent decades;
  • the Supreme Court is selecting historically similar cases for review, but writing about them in an increasingly idiosyncratic manner;
  • Supreme Court opinions are recognizable and distinct from appellate court opinions at a basic level;
  • opinions written by the Supreme Court are more characteristic and easily identifiable than in the past.

In the research, a computational and statistical analysis was able to near-perfectly distinguish between Supreme Court and appellate court decisions. Generally, the Supreme Court tended to use words associated with constitutional rights and statutory interpretation, while the appellate courts used more words associated with time and dates, testimony and evidence, proof and appeals.

“The Supreme Court tends to avoid the nitty-gritty and focus on its own preoccupations, even when analyzing and writing about the same cases as the lower courts,” said Livermore. 

Researchers used topic model analysis – an approach that focuses on the extraction of groups of words that tend to occur together throughout the hundreds of thousands of written opinions. This method allowed the review of over a half-century of opinions at a speed and level of specificity unmatched by the traditional method of relying on human reading.

In the study, researchers analyzed the extent that opinions conform to, or depart from, the writing of other courts – which they refer to as the “judicial genre” – by reviewing the content of the writing. Their findings show that over time, a “Supreme Court genre” has emerged within the larger genre of judicial opinions. Researchers also studied case selection history and the distinctiveness of language used in opinions.

“Technology amplifies tradition in this work, by assisting with the reading of such a large body of text,” said Daniel Rockmore, Dartmouth College’s William H. Neukom 1964 Professor of Computational Science, who co-authored the study along with former Dartmouth Neukom Postdoctoral Fellow Allen Riddell. “By doing so, big data has opened up the complex world of judicial opinions to allow us to understand one of our most important institutions in ways that human reading and hand-coding simply cannot.”

According to the researchers, the analysis could also lead to a better understanding of how the nation’s top court selects cases to review. The findings can also lead to more specific analysis of the public perception of the court as well as any opening legitimacy gap.