Exit Polls and the Extrapolation Mistake

Talking about turnout in the 2014 election can look an awful lot like making excuses for the Democrats' loss, which I wouldn't want to do. Democrats don't need to feel better about what happened last Tuesday. They ought to feel bad, not just over how their party performed but about the very real consequences to people's lives that might occur as a result.

But now that the data are coming in, we're seeing just how it was Republicans won. It wasn't because they did such a terrific job of persuading people to support their dynamic agenda for change, it was because their voters came to the polls and the Democrats' voters didn't.

That was made possible by the fact that turnout overall was so abysmal. According to the United States Election Project, turnout this year was 36.4 percent of the voting-eligible population, the lowest of any election since 1942. Among those who did vote, exit polls showed that Republicans outnumbered Democrats 36 to 35 percent, with the rest calling themselves independent. While that might not seem like a huge deal, in 2012 Democrats outnumbered Republicans 38 to 32 percent, in 2010 they were tied at 35 percent, and in 2008 Democrats led 39 to 32 percent. In other words, Democrats won't win unless they have a significant advantage in the number of their partisans who make it to vote.

Which brings us to the real point I wanted to make. There's a hidden assumption in a lot of discussion of exit polls, in which this basic point about the electorate that showed up gets forgotten. For instance, I heard Michael Barone on the radio the other day note that Millennials only barely favored Democrats in this election, and therefore that generation, far from being overwhelmingly Democratic like we've all been led to believe, is now "up for grabs."

This is a good example of how to take a true fact—among voters under 30, Democrats won in 2014 by only 11 points, while in 2012 Obama won them by 23 points and in 2008 by 34 points—and completely misinterpret what it represents.

Barone's hope that Millennials are moving toward the GOP might be warranted if the group of Millennials that turned out in 2014, not to mention those that did so in 2008 and 2012, were all perfect random samples of the generation as a whole. But there's no reason to believe that's true. Voters under 30 were only 13 percent of this year's electorate, compared to 19 percent in 2012. While turnout across the board was down, turnout among the young was down even more.

So what's most likely is that not only did many fewer Millennials vote this year than do in presidential elections, the ones that did were more heavily Republican. It wasn't that their minds changed. In other words, your 24-year-old hipster cousin with all the gay friends didn't bother going to the polls, while his freshman roommate, the one who works for a bank in San Diego and has a picture of Ronald Reagan in his wallet, did.

That isn't to say that the electorate in a presidential year, either as a whole or in any particular subgroup, is a perfect representation of the broader public either. It isn't, but it's closer to being representative than the electorate in a midterm.

This is why we have to be careful, when we're looking at exit polls from this election (or any other), to make sure we're properly restricting our interpretation of the results. What happened among a particular group of voters this year might help you predict what that group might do in 2018, or give you things to look out for in 2016, but it doesn't tell you what that group as a whole believes.

And in a way, this understanding gives us a different way to look at the polls, pretty much all of which failed to predict the outcome of this year's election (as Sam Wang notes, they underestimated Republican Senate candidates' performance by about 5 points on average, which is a lot). They were "wrong" in that sense, but that doesn't mean they weren't accurately measuring what people were telling them. It's just that a lot of people who told pollsters they were going to vote didn't in the end. That's always true to a degree, but this year it was more true among Democrats than among Republicans.

Of course, what actually matters when it comes to who gets to govern and what policies they carry out isn't what all the public truly believes in their hearts, but what the people who actually vote believe. Nevertheless, we still shouldn't mistake one for the other.