Permanent no-excuse absentee voting, nonpermanent no-excuse absentee voting, and Election Day registration all have positive correlations with turnout.
If you move from one state to the next, you may find dramatic differences in election administration. Back in my home state of California, I registered as a permanent absentee voter. Prior to each election, I received my ballot in the mail, which I completed at my leisure and returned by mail. Here in Utah, I vote in person on those creepy Diebold machines after waiting in line for too long.
In a recent article, Larocca and Klemanski look to see whether these state-to-state variations affect turnout. They conclude that these variations do matter. They find that permanent no-excuse absentee voting, nonpermanent no-excuse absentee voting, and Election Day registration all have positive correlations with turnout. They find that universal mail voting (where everybody is an absentee voter) and voter identification requirements have marginal positive correlations.
I use the word “correlation” intentionally in the previous paragraph. Larocca and Klemanski speak as though these are causal relationships: If a state adopts one of these laws, its turnout will rise. But it’s not clear to me that they’ve really shown causality.
They look at self-reported survey data (in the Current Population Survey) from 2000, 2004, and 2008. I won’t comment on the well-known tendency of survey respondents to overreport turnout because there’s a bigger problem here.
For this approach to work, it must be the case that lots of states were changing their turnout laws between these elections. Ideally, we’d have lots of states adopting these laws and lots of other states repealing these laws, with adoptions and repeals scattered randomly across states and randomly over time.
Suppose the opposite were true, and no state changed its registration laws between 2000 and 2008. Then, in effect, Larocca and Klemanski are just doing a cross-sectional analysis. If a state has a political culture that values participation, then that state would be more likely to (a) adopt laws intended to boost turnout and also (b) have higher turnout than other states no matter what laws it adopts. Any correlation between registration laws and turnout levels would be entirely spurious.
So there are two extremes. At one extreme, adoption and repeal of these registration laws happens frequently and randomly. At the other extreme, there are no changes in registration laws in any state between 2000 and 2008. A weakness in Larocca and Klemanski’s paper is that there’s not a clear indication of where we are along that spectrum. They mention a handful of changes between 2000 and 2008 in passing (mainly in pages 78-80, 89, and 93), but it’s not clear from their discussion that LOTS of states changed their laws, which is what we would need for this analysis to work. They claim on page 78 (and again on page 85) that a table is available online showing which states adopted each reform in which year, but the link is unfortunately broken. An essential table is, therefore, missing.
I like the study and want to believe it, but the missing table would make it much easier to decide how much faith to put in the results.