Problems with a Methodology Used to Evaluate the Voter Initiative

Abstract

This note identifies problems with a methodology that has been used to test whether policy is more or less responsive to public opinion in states with voter initiatives. The methodology is to regress a policy variable on a measure of constituent preferences, and compare the coefficients for states with and without voter initiatives. The states with the largest coefficients are said to be more responsive to public opinion. Such an inference is shown to be invalid.