Moving Beyond Mavericks: Competing Conceptions of Individual Member Vote Misclassifications in the 75th to 110th Congresses.

Ideal point estimates are one of the most commonly used measures in studies of the U.S. Congress. While much has been written about the overall accuracy of measures such as NOMINATE, relatively little has been said about when these measures yield incorrect predictions of member voting behavior. Ideal point estimates derived from legislative roll-calls suggest that most voting behavior in the U.S. Congress can be understood by a one or two dimensional space. Perhaps the most convincing evidence of this point is that one-dimension correctly predicts over 80% of all votes over the last 60 years and over 90% in the last decade (Poole and Rosenthal 2007). While the relatively small number of misclassified votes is central to arguments about the low dimensionality of roll call voting in Congress, much of the information contained in these errors remains untapped. In this paper we focus on member level misclassification rates measured by the failure of ideal point estimates to predict actual votes. Conventional wisdom would suggest that high misclassification rates are associated with "maverick" members of the House and Senate. We move beyond this conception and consider a number of potential explanations (e.g. constituency pressures, realignment, the changing agenda, and responsibility within the chamber) for member misclassifications among representatives and senators over time. By understanding what NOMINATE scaling procedures fail to predict, we can get a better idea of what the first two dimensions truly capture about members of Congress beyond a maverick tendency. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]