After the 2009 referendum on a proposed change to the Danish Law of Succession, it was widely claimed that the early publication of exit poll results changed the rate of turnout and eventually the outcome. We investigate this claim and contribute to the wider debate on the implications of exit polls by setting up and analyzing a formal model. We find that the introduction of an exit poll influences the incentive to vote both before and after the poll is published, but the signs of the effects are generally ambiguous. The observation that exit polls influence the incentive to vote even before they are published is often overlooked. We show that this can lead to premature conclusions about the impact of exit polls on electoral outcomes. In particular, in cases such as the Danish referendum where it clearly appears that exit polls changed the outcome, it could well be that the outcome would have been the same had there been no exit polls.
