ABA Journal Reports on Our Supreme Court Prediction Algorithm

July 29th, 2014

This morning my colleagues and I announced our new paper, discussing our model that was able to predict 70.9% of the votes of the Justices, in every case decided since 1953. This afternoon I spoke with Debra Cassens Weiss, who wrote a feature about the new project.

A South Texas College of Law assistant professor who developed a Supreme Court fantasy league says he and two colleagues have developed a computer model that can predict decisions of the court and individual justices.

Law professor Josh Blackman, writing at his blog, says his computer model, applied to cases since 1953, correctly identifies 69.7 percent of the court’s affirmances and reversals, as well as 70.9 percent of the votes of individual justices. A paper at SSRN has details.

The predictions are based on data that was available before the court’s decision. Ninety variables are used, including the party of the appointing president, the court era in which the decision is written, the court and justice’s ideological direction, and the agreement level of the court. The model compares predictions for each case to what actually happened, learning which variables work and which don’t.

“While other models have achieved comparable accuracy rates,” Blackman writes, “they were only designed to work at a single point in time with a single set of nine justices. Our model has proven consistently accurate at predicting six decades of behavior of thirty justices appointed by 13 presidents.”

Blackman says he will be hosting a tournament where fantasy SCOTUS players compete against the algorithm. “What IBM’s Watson did on Jeopardy, our model aims to do for the Supreme Court,” he writes.

Stay tuned for a lot more cool stuff.