Professional Documents
Culture Documents
Predict Court Behavior
Predict Court Behavior
Predict Court Behavior
A general approach for predicting the behavior of the Supreme Court of the United States
Summary
A generalized, consistent and out-of-sample applicable model based on time-evolving
random forest classifier is developed to predict the decisions (case outcome, justice vote) of
the Supreme Court of the United States
The model achieves 70.2% accuracy at the case outcome level, 71.9% at the justice vote level
(1816-2015), and outperforms a baseline model by 5% over the past century
The model can be applied out-of-sample to the entire past and future of the Court
representing an important advance in the filed of quantitative legal prediction
Three criticisms
Although the model is able to achieve a promising prediction accuracy in both case outcome
and justice vote, it still relies on experts to explain how the decision is made
The performance of the author's model only outperforms one in-sample optimized null
model by nearly 5%, so we need more comparison with judicial prediction models that use
advanced machine learning technology such as deep neural networks
The concluded data in the article is not visualized as graphs making difficulties for readers to
draw their own conclusion.