Statistical Models of Senate Roll Call Voting

Abstract

This paper uses statistical analysis to consider what factors influence the way senators vote and how important these influences are. The answers are specific statements about individual senators' decision processes, and quantitative estimates of the weights applied to the different variables in these models. The form of the voting models and the variables in them were developed from hypotheses about individual decision-making and descriptions of the legislative process. Examples of the variables are the preferences of the senator's constituency, the opinions of his party leader, and the views of the President. Quantitative measures of the variables were obtained from state demographic characteristics and by Guttman-scaling the votes on passage of the amendments to specific bills considered in 1961 and 1962. Linear regression analysis of these Guttman scales was then used to test the hypotheses and estimate the coefficients in each senator's voting model. After the results from these analyses are discussed, the explanatory power of these models on individual bills is evaluated by comparing the vote they estimate with those predicted using two alternative models.