Policy, party and perceptual bias in American politics

This dissertation examines how partisanship, policy views and citizen perceptions combine to influence political behavior in the United States. Using a unique new survey design and ideal point estimation technique, I am able to directly scale the ideal points of ordinary citizens alongside the positions taken by senators and the President. Furthermore, I am also able to estimate respondents' perceptions of the positions of their senators' positions on this same scale. Using these estimates, I am able to answer several central questions in American politics. The first chapter of this dissertation examines the axioms of Downsian voting and finds that, while people appear to use spatial voting rules on average, when partisanship is taken into account, the story changes dramatically. While independents are shown to use voting rules that are virtually identical to Downsian prescriptions, partisans are heavily biased by their emotional attachments. Second, I show that partisanship and personal views have a large influence on citizen perceptions of their senators. Citizens perceive senators of their own party to hold the same positions as they themselves do. Furthermore, I find strong evidence of negative projection, in contrast to the findings of much recent work on the subject. There are very strong negative correlations between respondents' ideal points and their perceptions of opposite party senators. Finally, I propose and estimate a hierarchical model of state voter ideology distributions. In addition to providing direct estimates of median voter positions, the model is able to test the (usually implicit) assumption of many previous studies that respondents use ordinal ideology scales on average in the same way. I show that respondents' use of these scales depends heavily on their political environment, with citizens from more liberal states using the scale in different ways from those in conservative states. This provides further support for the new survey and scaling techniques used here to measure voter ideal points.