The Statistical Analysis Of Roll-Call Data:

Roll-call voting and congressional procedures are two of the most heavily studied aspects of the U.S. Congress. To date, little work has focused on the effect of procedures on the composition of the roll-call record. This article takes a step in this direction by demonstrating the effect of chamber rules and institutional constraints on House and Senate roll-call data, as well as on the inferences that scholars have drawn from the roll-call record. More specifically, I focus on recent efforts to measure party effects and ideological alignments, and I demonstrate that the composition of the roll-call record can affect these measures. Roll-call voting is one of the most heavily studied aspects of the U.S. Congress. The large number of actors combined with availability of data in “machine-readable ” form have led scholars to do extensive theoretical and empirical work with the congressional roll-call record. Measures derived from roll-call votes, such as the Rice index of party voting (Rice 1925), as well as more-sophisticated scaling analyses, have allowed scholars to discover patterns and alignments in roll-call voting. Further, the development of the spatial model of voting and advances in computer technology have enabled scholars to develop even more-sophisticated measures of patterns in roll-call voting through algorithms such as NOMINATE (Poole and Rosenthal 1997) and, more recently, Bayesian Ideal Point Estimation (Clinton, Jackman, and Rivers 2004; Martin and Quinn 2002)