About
Supreme Court Report
Quantifying the U.S. Supreme Court’s behavior and making it legible for the public.
Supreme Court Report is built and maintained by Tommy Bennett.
Every page here is generated from the Court’s own public docket data. The daily dashboards, the conference reports, the argument navigator and a page for each of the roughly 55,000 dockets on file are all pre-rendered as static HTML — there is no server, no tracking beyond aggregate page views, and no paywall.
The cert-grant forecasts are estimates from a statistical model, not predictions about any particular case. How the model is built, what it gets right and where it is weakest are set out in the methods note.
The code that fetches the data, fits the models and writes these pages is open source: github.com/baldrige/ceRt.
Updates are published as Atom feeds, which any feed reader can follow: all updates carries new grants, the Court's order lists and decisions, conference reports and the daily docket; certiorari grants carries grants alone, which is a handful a month.
Subject areas
Case pages and the daily docket label each petition with a subject area: one of the 14 issue areas of the Supreme Court Database (Criminal Procedure, Civil Rights, First Amendment, Economic Activity, Judicial Power and so on), which scholars have used to classify the Court’s decisions for decades. The label is an estimate. A classification model reads the caption and the questions presented and picks the area; it is not the Court’s own description of the case, and the Database itself codes only the cases the Court decides.
Checked against the Database’s coding of 573 cases decided since 2017, the model agrees about 78% of the time, and about 81% on a hand-labelled sample of denied and pending petitions. Most of the disagreements are close calls, or cases the Database codes by how the Court disposed of them (as a question of standing, say) rather than by what the petition asked. A label is shown only when the model is reasonably sure of it — around 84% of cases — and not at all when the questions presented could not be read from the petition, as with many handwritten filings.
Found a problem?
Corrections and questions are welcome — email tbbennett@smu.edu. These pages are generated automatically, so an error in one is usually an error in many; reports are genuinely useful.