Weighing the alternatives: Preferences, parties, and constituency in roll-call voting

Theories of parties and lawmaking typically require measures of legislators' preferences for empirical analysis (Aldrich and Rohde 2000; Cox and McCubbins 2005; Krehbiel 1996). However, existing methods for generating estimates of these preferences presume that legislators care only about their own policy preferences and not about their constituency or party position (Clinton, Jackman, and Rivers 2004; Poole and Rosenthal 1997), though substantive scholars have for decades hypothesized otherwise (Clausen 1967; Fenno 1973; Sinclair 1995; Smith 2007). Focusing on the US Senate, I develop a new statistical estimator to determine the weights legislators place or their preferences, party, and constituency in roll-call voting. Estimation is within a Bayesian IRT framework. The results help to explain the gap between estimated ideal points and legislators' true preferences and, thereby, have important implications for lawmaking theories, as well as theories of representation.