Skip to main content

Improved Inference for the Instrumental Variable Estimator or You Can't Get Something for Nothing

The paper examines the hypothesis that, in a properly-functioning representational system with fully rational actors there should exist no relationship at the macro level between representatives' activities and demand for policies on the part of voters. Over time, a measure of the reaction of voters to public policy outputs should approximate a random walk in such a system. We examine the distributions in changes in election margins for a variety of elections to show that the prediction of fully endogenous representation does not hold and propose an alternative model based on boundedly rational decision-making processes.