Replicating and extending Martin-Quinn scores

Abstract

We set to replicate and extend the commonly used Martin and Quinn (2002) scores of judicial ideology for the entire universe of U.S Supreme Court justices from 1793 onwards. Our replication strategy consists of three steps. First, we replicate the constant ideal point scores with a novel variable distinguishing between liberal and conservative case reversals. Using this variable, we show that the replicated scores exhibit a moderate-to-strong correlation with the original scores. Second, we extend the replication of constant ideal points to specific policy areas and show that the variation in latent judicial ideology in the areas of criminal procedure, economics, first amendment and due process strongly predict the overall scores. And third, we replicate and extend dynamic scores to capture the temporal changes in the judicial ideology using repeated sampling Monte Carlo Marko Chain simulation without imposing prior distribution on case-level parameters. To allow for judicial ideology to trend smoothly over time, we employ a large-scale MCMC simulation involving roughly 1.9 billion iterations of the baseline latent judicial trait model, and a filtering parameter to separate the long-run ideology trend from its cyclical component. The replicated dynamic ideal point estimates without prior restrictions generally confirm the trends of dynamic Martin-Quinn scores.