Abstract:
High electoral turnout is seen as a key determinant of a successful democracy. One reason for this is that turnout is believed to affect electoral outcomes. This has given rise to a flourishing literature studying the effect of increased turnout on various outcomes. It is well known that variation in turnout is not random, so a plethora of identification strategies are applied to assure causal estimates. In this paper I argue that this endeavor is futile. I first discuss theoretically how we should define increased turnout. Next, I show that with the various identification strategies, estimates are based on the behavioral response of a biased group of voters.