Dimensions in Congressional Voting

While dimensional studies of congressional voting find a single, ideological dimension, regression estimates find several constituency and party dimensions in addition to ideology. I rescale several unidimensional studies to show their increased classification success over the null hypothesis that votes are not unidimensional. Several null hypotheses are explored. With these null hypotheses, 66%–75% of nonunidimensional roll call votes are nevertheless correctly classified by one dimension. After the rescaling, one dimension succeeds in correctly classifying 25%–50% of the votes, and second and third dimensions are important.