To Elect or to Appoint ?

In this appendix we consider a procedure to assess quantitatively whether the expressive or strategic voting model is more appropriate for the state supreme court voting data used in this paper. As we remarked in the main text, for a given set of cases, the first-stage likelihood function for the votes is identical for both the expressive and strategic voting models, and in this sense the two models are observationally equivalent. Here, we bring additional data to bear which was not used in our main analysis, in order to discriminate between these models. Specifically, in the paper we only considered cases in which all members of the court vote. Because of this, we did not use the subsample of cases for which only k < n of the court members (an “incomplete court”) voted. This is the additional subsample which we use to compare the strategic and expressive voting models. In particular, using the estimates from the main analysis, we construct predicted voting probabilities for these “incomplete court” cases, under both the strategic and expressive voting hypotheses. We then compare the two models based on which one generates the higher likelihood according to these predicted voting probabilities. We proceed in the following sequence of steps.