Evaluating Theories of Congressional Roll-Call Voting

Criteria are developed to evaluate recent theories of roll-call voting in the House of Representatives. Since tests of these very different theories find high levels of predictive success, we must decide how to choose among them. Baseline models are developed to show the extent to which the votes could be predicted with minimal information. The 80-90 percent of the individual votes correctly predicted by the theories is found to be little improvement over the baseline models predicting voting along with the House or party majority. Since the statistical criterion is found to be indeterminate, the importance of verisimilitude to the process being studied is stressed. Simulation studies have done a good job of portraying the process aspects, and they could be usefully combined with statistical studies of long-term forces and interviewing studies of short-term forces affecting the voting. Theory development in the field should move to a more longitudinal perspective, as well as combining the long-term and short-term elements into a single overarching theory.