by Phillip Bump, Washington Post
There are few things more consequential than federal elections, and so many Americans — politics junkies in particular — spend the weeks leading up to every other November poring over information about what is likely to happen. It’s like the week before the Super Bowl, but with fewer Ford F-150 ads. (I hope this article runs with a Ford F-150 ad.)
Curious what will happen, we turn to augurs like Nate Silver to offer some insight. We look at poll data and try to assess which way the lines are likely to move. And we look at indicators such as early-vote patterns to try to discern some murky vision from deep within those tea leaves that will tell us something we’ll find out in two weeks anyway.
So let’s look at some tea leaves.
I was intrigued by a tool created by TargetSmart, a Democratic data firm, that shares its analysis of early-vote returns as they trickle in. Called TargetEarly, it includes a number of demographic splits for those ballots that have already been returned, the sort of overload of data that can keep up distracted political gurus for an unhealthy amount of time.
The data also allowed me to assess a question I had. In most elections, the people who vote are people who usually vote — the sort of statement that seems both insightful and obvious, and probably is each. In exceptional elections, people who don’t usually vote turn out, too, shifting things dramatically one way or the other. My question was straightforward: Are surges in early-vote totals a function of infrequent voters being motivated to go to the polls?
Data provided to The Washington Post by TargetSmart contained an answer.
No, not really.
TargetSmart breaks the voter universe into four groups, defined here. The top 20 percent of most-frequent voters are dubbed “super voters.” Those who fall somewhere in the middle are “frequent.” If you’ve voted once, “infrequent.” And then there are those who’ve never voted at all.
In every state for which there are data, save one, at least three-quarters of the ballots cast came from voters in one of the the first two, more heavily voting categories, according to TargetSmart’s analysis. (The exception was New York, and it missed the mark by only about a percentage point. In most of the states, at least 40 percent of votes cast came from “super voters.” In three states, half did.)
On average, about 17 percent of the votes cast came from infrequent or new voters. The state with the highest percentage of new voters is Nevada, at 5.3 percent. (We exclude Connecticut because there haven’t been many ballots cast there that are included in the analysis.)
Still, though: Nearly a fifth of votes came from new or infrequent voters? Isn’t that perhaps a sign that 2018 is shaping up to be something unusual? Well, on average, those groups make up about 46 percent of the electorate in a state, so they’re still broadly underperforming.
What’s more, the pattern in state after state looks a lot more like 2014 than 2016. TargetSmart ran the same analysis to this point in those cycles, and the general pattern is that more new or infrequent voters cast ballots in 2016. That’s not unusual, of course; presidential elections see higher turnout. But if we’re wondering how 2018 stacks up as an election with motivated voters, these data suggest it will look like the last midterm.
In nearly three-quarters of the states for which there are data, though, the density of less-frequent voters is up over 2014, by an average of about 4.3 points. The biggest increases year-to-date over 2014? In New York — and Georgia, where a lot of emphasis has been placed on restrictions that some voters face.
Remember: These are just tea leaves. It’s one analysis of one part of the election, so take it with a grain of salt.
It’s also only one way of looking at the data. In some states, the number of ballots cast is near or higher than the number of ballots cast to this point in 2016. That’s important, too.
But the question was whether those ballots were coming from people who would probably have voted anyway. And the answer, it seems, is overwhelmingly yes.