Testing Theories of Legislative Voting: Dimensions, Ideology and Structure

While dimensional studies of legislative voting find a single ideological dimension (Schneider 1979, Poole and Rosenthal 1985b), regression estimates find constituency and party dominant (Kau and Rubin 1979, Peltzman 1984), and ideology secondary (Kalt and Zupan 1984). This paper rescales the dimensional findings to show their improved classification success over the null hypothesis that votes are not unidimensional. With the rescaling, most votes are not explained by one dimension, and several dimensions are important Nevertheless, fewer dimensions are found than constituents' preferences suggest. Thus a model is developed where transactions costs of building coalitions reduce the number of dimensions. When legislative parties build internal coalitions to pass and defeat bills, voting on randomly drawn bills has a single party-oriented dimension. And natural ideological dimensions are reinforced if parties write bills and logroll along natural lines of cohesion.