Reconciling Theory and Estimation: Testing Theories of Lawmaking Using Ideal Point Estimates * Joshua D. Clinton Department of Politics Princeton University Princeton, NJ 08544-1012 clinton@princeton.edu Version 1.1: Under Active Revision December 1, 2005 Abstract Theories of lawmaking generate equilibrium predictions about the set of acceptable policy outcomes. Tests of theoretical predictions often use independently estimated ideal points based on roll call votes to estimate what appear to be theoretically relevant quantities. This paper demonstrates that such measures are invalid because of the endogeneity of the roll call record; predictions regarding equilibrium outcomes necessarily yield predictions about the agenda used to enact the outcomes. I demonstrate how divorcing the statistical model from the theoretical model results in the misleading use of ideal point estimates and the potential for incorrect inferences. Using the 103rd Congress and the predictions from prominent gatekeeping and veto theories, I illustrate how to integrate theoretical predictions within the statistical model used to estimate ideal points and reconcile the problematic disconnect between theory and estimation. In demonstrating how to account for the endogeneity of roll calls when theory-testing, I highlight the considerable complexity that results from testing predictions from models assuming perfect spatial voting using probabilistic voting model estimates and I argue that the properties of commonly used measures are not well-understood. * The author would like to thank Keith Krehbiel, David Lewis, Nolan McCarty, Adam Meirowitz, Keith Poole and Eric Schickler for helpful comments. 1