The Spatial Model and the Senate Trial of President Clinton

Abstract

Most accounts of the Clinton Senate trial are based on legal theories and description or suggest that the Senate trial was sui generis, unrelated to regular Senate business. Alternatively, we argue that presidential removal trials can best be explained through a spatial voting model, with ideological estimates for senators best predicting the trial outcome. To examine these hypotheses, we generate (a) ideal-point estimates using Markov chain Monte Carlo methods on roll-call data as well as (b) text-based (wordscores) point estimates scaled from public statements. Our results support the spatial hypothesis, but not legal or idiosyncratic accounts. Moreover, the significant discrepancy between text- and vote-based estimates provides support for Mayhew’s contention that position taking for constituency benefit is rather costless. We conclude that when asked to be judges, senators, faced with multifarious political incentives, are more likely to act like legislators.