The idea behind early voting is that by making it easier for people to vote, you will ensure that more people do so. Somewhat surprisingly, in the real world, it may not be working out that way :
Our research, conducted with our colleagues David Canon and Donald Moynihan at the University of Wisconsin, is based on a three-part statistical analysis of the 2008 presidential election. First, we analyzed voting patterns in each of the nation's 3,100 counties to estimate the effect of early voting laws on turnout. We controlled for a wide range of demographic, geographic and political variables, like whether a county was in a battleground state.
Controlling for all of the other factors thought to shape voter participation, our model showed that the availability of early voting reduced turnout in the typical county by three percentage points,. Consider, as an example, a county in Kentucky, which lacks early voting. If we compared this to a similar county in neighboring Tennessee, which permits early voting, we would observe, other things being equal, turnout that was 3 points lower.
Next, we studied the data on more than 70,000 voters and nonvoters from the Census Bureau's Current Population Survey, which asks respondents whether they voted. Once again, we employed a statistical model to control for demographic variables like education and race as well as geographic and political factors. The model showed that an individual living in a state with early voting had a probability of voting that was four points lower than a comparable voter in a state without early voting.
Third, we took advantage of a useful feature of the census survey, which asks individuals whether they voted early or on Election Day. We examined the characteristics of voters and nonvoters, and found that the profiles of early voters and election day voters were mostly similar.
With one big exception: our model forecast that early voters had profiles that made them two percentage points more likely to vote than Election Day voters, whether there was an early option or not. Early voters were more educated and older and had higher incomes, all traits associated with a higher probability of voting. A probability difference of 2 percentage points may seem like a trivial figure, but when applied to populations of millions, it can shift national and state elections.
Even with all of the added convenience and easier opportunities to cast ballots, turnout not only doesn't increase with early voting, it actually falls. How can this be?
It turns out that early voting mitigates the impact of get-out-the-vote mechanisms, and it also diminishes the impact of "Election Day". Voting used to be a public act of civic engagement, with all sorts of activity focused around election day, from exhortatory news stories to social pressure from the sight of neighbors trudging to the poll. One way to think about it is that voting signals something about you to others in the community, but with the advent of early voting, that signal is no longer so powerful: someone who doesn't turn up on election day might not have voted, or they might simply have gotten it out of the way weeks before.
This isn't perhaps quite as surprising as it seems. Dan Ariely's book discussed a pre-school in Israel where a day care center started charging a fee for parents who were late picking up their children, only to see the number of late pick-ups rise--apparently once there was a fee, parents no longer felt guilty about being late. The day care then rescinded the fee, but the norm--"don't pick up your kids late" had already been broken; now parents simply concluded that there was no penalty at all, and the problem became even worse.
Economists like to say "incentives matter", but we need to be careful that we understand what peoples' incentives are before we try to change them.