I’ve gotten a lot of pushback on my post from Sunday night about early voting figures. Michael P. McDonald of The Huffington Post, for instance — whose work I referred to in my article – calls it “fatally flawed” and suggests I have erred by comparing early voting totals in each state to party registration figures. Instead, he thinks, the proper comparison is to early voting figures from past years.
My article consisted of essentially two different parts. About 600 words were devoted to critiquing the notion that the early voting data – particularly in the way that some other analysts are using it — tells us much of anything at all. The other 600 words (the part that Mr. McDonald criticizes) were devoted to a comparison of the early voting figures against voter registration data in each state, which I suggested revealed a small “enthusiasm gap” in favor of Republicans.
If you take just one point from yesterday’s article, I’d really prefer it be the former, more skeptical one. A lot of the analyses of early voting figures are quite flawed. I’ll take some blame here for having selected a poor headline, which did not emphasize this point enough.
Mr. McDonald, as I mentioned, suggests that the right way to view the early voting data is to compare it to early voting data from past elections — rather than, as I did, to voter registration figures in the each state.
Here is the problem with that. Depending on which past elections we compare it to, the early vote data might suggest anything from an impending catastrophe for the Democrats, to an outcome in which they’d radically outperform expectations on Election Day and hold both the Senate and the House (possibly by somewhat comfortable margins).
Let’s say that point of comparison is 2008, for instance. Using data from Mr. McDonald’s Web site, I found seven states in which there was early voting data both for that year and for this one. In each of these even states, Republicans are turning out in greater proportion than they did in 2008, often by wide margins:
This year, on average, the early vote has favored Democrats in these states by an average of 2 points. But in 2008, it favored them by an average of about 17 points — making for a 15-point swing against them.
If there were a 15-point swing toward Republicans in the partisan composition of the electorate, coupled with poor performance for most Democratic candidates among independent voters, that basically looks a lot like the Gallup “lower turnout” model, which could imply Democratic losses in the House in excess of 70 or even 80 seats — and which certainly implies the presence of a very large enthusiasm gap.
Prior to 2008, however — when Democrats put a major emphasis on early voting — the conventional wisdom held that early voting worked instead to the benefit of Republicans, as older voters (who are more likely to vote Republican) were more inclined to vote early. Early voting is also something that Karl Rove had emphasized when he was running campaigns for George W. Bush.
The evidence in these years is a little spottier, but there is one important data point that confirms Republican strength in these years: the National Annenberg Election Study found that George W. Bush won 60 percent of the early vote in 2004, and 63 percent in 2000. That is, he won the early vote by 20 or 25 points each year, even though the overall margins in his elections against John Kerry and Al Gore wound up close to even.
If that is the point of comparison, then this year’s numbers look really, really good for Democrats. So far this year, Republicans have something like a 6-point advantage in early voting in a typical state, as compared with their voter registration figures there. But if, in the past, Republicans had something like a 20-point edge instead — which is what they’d have needed to allow Mr. Bush to win the early vote so comfortably – Democrats are way overperforming. (Perhaps this even implies that there will be a turnout gap in their favor by the time all the votes are counted.)
You could also potentially look toward the last midterm year, 2006, as a point of comparison, but there is less in the way of comprehensive evidence there.
The Washington Post’s Karen Tumulty noted, for instance, that a Democratic consulting firm collecting early voting data “told its clients … that early ballots in the 17 states … looks very much like that in 2006, the year Democrats took back the House and the Senate.” Strategists at Organizing for America have conveyed similar sentiments.
I’m a little wary about the comparisons to 2006, because I haven’t seen any that make an effort to be especially comprehensive. The data that Democrats are sharing publicly highlights figures from some states, but not others, and those states may or may not be representative.
Even if it were true, however, that Democratic early voting is about on pace with 2006, it’s still not clear that it forms the inherently better comparison than 2008, or 2004, or 2000, and so forth. What we know is that early voting patterns vary a lot from year to year — and, in recent years, have both massively favored Democrats and massively favored Republicans at different times.
There is no rule of thumb about what early voting figures “should” be. The early voting advantage is presumably some function of: (i) demographics, i.e., older voters are more inclined to vote early; (ii) enthusiasm, which is peculiar to each particular election; (iii) the extent to which each party emphasizes early voting; (iv) whether or not the year is a midterm. Suppose we were trying to fit a four-variable regression model to predict early voting. You can’t really fit a four-variable model on only four data points (e.g. 2008, 2000, 2004 and 2006). It just doesn’t work, statistically. You can’t even hazard a guess.
In other words, the early voting data this year could be consistent with anything from a massive, Gallup-style Republican wave to the first sign of a major Democratic comeback. We really don’t know.
The only thing we do know, rather, is the fact cited Sunday night: in most states so far, registered Republicans are casting ballots at a somewhat higher rate than registered Democrats. Their advantage so far amounts to about 6 points, which for better or for worse, corresponds quite nicely to the roughly 6-point “enthusiasm gap” that most pollsters are seeing.
Now, it is possibly or even probably a coincidence that the turnout gap in the early voting data matches the enthusiasm gap in the polling data (a point I should have made more clearly in last night’s article). Perhaps, in a universe where the overall enthusiasm gap is 6 points in favor of Republicans, their advantage among early voters “should” be 20 points, or 2 points, or some completely different number.
We really don’t know. I mean, we really don’t know. Early voting is quite new in most parts of the country, and something most voters weren’t taking advantage of until quite recently. Perhaps by 2018 or 2020, we’ll have a better idea of how early voting operates under different sets of political conditions and will have developed better techniques for analyzing it.
The notion, however, that you can just plug in the 2006 early data, or the 2008 or 2004 numbers, and have some decent baseline to “read” what this year’s early voting numbers mean is highly dubious. Mr. McDonald suggests such comparisons are apples-to-apples. But it’s more like comparing an apple sitting on the kitchen counter in front of you to some sort of Schrödinger’s apple that might either be a Red Delicious, or a Granny Smith, or some kind of crabapple — and you have no way to tell which.