I’m serious – don’t pay attention to early voting.

About this time in every election cycle analysts become starved for data to analyze. Of course, actual useful data is just around the corner, dropping on Election Day 2022. Some analysts, not content to wait, turn to the closest thing we have: early voting. It happens every time. So once again I will implore the site’s readership to just wait until Tuesday, because the analysis of early voting is a mug’s game in almost all circumstances.

The basic rationale for this thesis is straightforward. Consider one way of looking at the Democrats’ vote share in an election. It is determined by four factors: (1) the number of people who vote early; (2) multiplied by the percentage of those voters who voted for Democrats; plus (3) the number of people who vote on Election Day; (4) multiplied by the percentage of those voters who voted for Democrats.

We currently have data for a large portion of (1). If we live in a state with party registration, we can get some sense of (2). That still leaves us with two values that are almost entirely unknown. What does it mean if Democrats get 100% of the early vote in a state? We don’t know. If Republicans sweep the Election Day vote and more voters vote on Election Day than vote early, they would still win. If not, they’d lose, which isn’t terribly helpful.

Perhaps we can compare to turnout in previous years? This is always problematic, because parties change strategies and voters adapt their behaviors. Early voting in North Carolina was largely the traditional province of white Republicans. In 2008, however, the Obama campaign emphasized early voting, and it increasingly belongs to Democrats.

But it is especially difficult to analyze the early vote this cycle, because we are coming off of a highly unusual cycle. The 2020 elections offer a flawed baseline because of the COVID pandemic and unusually high prevalence of early voting and mail-in voting. 

Perhaps we could compare instead to the 2018 election? The problem is that the 2020 elections probably altered behavior somewhat, as people became increasingly aware of early voting and mail-in voting. The larger problem is that early/mail voting took on a partisan valence, as Republicans were subjected to a barrage of claims by party officials that early voting and mail-in voting were means of stealing votes. We might expect more Democrats to partake of early voting because of this party messaging. Or, maybe not? We just don’t know.

Because of this uncertainty, early voting analysis is ripe for “wish-casting.” One can see a defensible conclusion (we’d expect Democrats to do better in 2022 than 2018 because of the debates of 2020) and either reject it or accept it due to our partisan preferences. One can certainly look at the early voting and see good signs for Democrats. One can also, however, look at the early vote and see good signs for Republicans as well: Republicans are performing well in the early vote in Miami, for example. Nevada elections guru Jon Ralston (who actually has a track record of predicting elections well, because almost all of the votes in Nevada are cast early) has declared that Democrats are in trouble. The black share of the electorate has steadily declined in Georgia as the early voting period has progressed. Is this a warning sign for Democrats? Or is this a sign that white liberals are hyper-engaged?

How you react to these questions is probably dependent on your political affiliation, which isn’t a recipe for good analysis. To paraphrase Paul Samuelson, early voting has predicted five of the last two good Democratic years, because every year there are good signs in the early voting for Democrats, as well as good signs for Republicans.

What is particularly frustrating is that this is guaranteed to be a good night for Republicans, or a disappointing night for them. So someone who is currently analyzing early voting will be right. But they’ll be right in the sense that someone who adds two and two and gets five, and then adds three and gets seven ends up in the right place. Bad analysis can lead to solid predictions, but that’s the best one can hope for from early voting analysis (unless you are Ralston).

So as tempting as it may be, just ignore the early voting tea leaves. We’ll have actual results to fight about in five nights.