A man wears stickers after casting his ballot on Oct. 10 at the Hamilton County Board of Elections in Cincinnati. (John Minchillo/AP)

Prior to the last weekend before an election, campaigns contact voters by phone, text and in person to encourage or persuade them to vote for their candidates. In places where there’s early voting, campaigns may push voters to do so, to bank the votes, as they say. On that last weekend and through Election Day, though, the system shifts to one purpose and one purpose only: making sure that the right people cast their ballots.

This is “get out the vote” or GOTV. The rule of thumb is that a strong GOTV effort, good “field,” can make a difference of a few percentage points. If you’re polling at 49 percent and your opponent is at 50, the idea is that good field can make the difference. As with all rules of thumb, it’s worth taking that with a grain of salt, but it’s certainly the case that turning out more of your voters isn’t going to do you any harm.

Consider what turnout means in a real-world example. The New York Times has been partnering with Siena College to run polls in various House districts for months. That includes Virginia’s 10th District, where incumbent Rep. Barbara Comstock (R) trailed Democrat Jennifer Wexton by seven points earlier this month.

How big that gap is depends on how many people vote, of course. The Times offers different turnout scenarios in the district, which offer different results. They figure that 286,000 people will vote, giving Wexton that edge. If you look just at people who say they’re certain to vote, turnout spikes to 317,000, and Wexton’s lead grows to 10 points. If you consider an electorate like that of 2016, 385,000 people would vote, and Wexton’s lead drops to five points. It’s not just turning people out, it’s about turning out the right people, meaning, for the Wexton campaign, more women and nonwhite voters.

Note, by the way, that increasing turnout from 286,000 to 317,000 vote — gaining Wexton three percentage points — requires getting 31,000 more people to vote. From the standpoint of a campaign, that’s a lot of turnout increase. Figure you’re going door-to-door on Election Day encouraging infrequent voters to cast a ballot. Knock on six doors an hour and get two people to vote who otherwise wouldn’t, and you need 1,200 volunteers working all 13 hours that polls are open. That’s … unrealistic. (As is getting two additional voters per hour.)

Notice, too, that the Times doesn’t figure that all of those people who say they’re certain to vote are going to vote. There’s good reason for that: A lot of people who say they’re certain to vote don’t.

The Post and our polling partners at ABC News have polled in the October before each of the past three elections. As expected, more people in 2016 told us they were certain to vote than did in 2014, the last midterm election. The percentage saying they’re certain to vote this year is down from 2016, but way up from 2014 — an indicator of the increased interest in this election that’s been widely reported.


(Philip Bump/The Washington Post)

Then compare those predictions of certainty to vote with actual turnout, as calculated by the University of Florida’s Michael McDonald at the United States Election Project.

In 2014, 65 percent of voters said they were certain to vote; 37 percent did. In 2016, 85 percent said they were certain to vote, 25 percentage points higher than actual turnout.


(Philip Bump/The Washington Post)

Where that becomes particularly important is when looking at demographic groups that are important to particular candidates or parties. Younger voters, for example, broadly prefer Democrats and also turn out less heavily.

There’s a lot of data in these next two graphs, but we’ll pick out the important stuff.


(Philip Bump/The Washington Post)

The most important part of this chart is that turnout among those younger than 40 is poised to look far more like 2016 turnout than 2014 turnout — at least if certainty to vote is any indicator. The estimates of actual turnout use different age groupings than the Post-ABC estimate of certainty to vote (as, annoyingly, do the exit polls), but the pattern is obvious: Young people vote less often and more heavily Democratic.

The gap between predicted voting (certainty to vote in our polling) and actual turnout is about 20 points or so across demographics. If that holds for young voters in 2018, it would mean 50 percent turnout — low, but at almost presidential-election levels.

That if is a very big if.

Notice on the age graph that the low turnout in 2014 means that the shift from 2014 to 2018 is much more stark among younger voters. More voters who are 40 to 64 or 65 and older also told us they were certain to vote this year than in 2014, but the jump among young voters is much sharper.

The same holds for black voters. While the number of black respondents in our polls is small (resulting in a larger margin of error), there’s a big increase in reported certainty to vote this year. As with younger voters relative to older voters, black certainty to vote still trails white certainty — but that change from 2014 to 2018 is potentially important. (The percentage of Hispanic respondents in our latest poll is too small to report.)


(Philip Bump/The Washington Post)

For Democrats, in other words, the “right” voters are already saying that they’re more likely to vote than said they were in 2014. It is far easier to increase the number of your own voters who go to the polls if those voters are self-motivated. That’s how you get shifts of more than a couple of points in election results.

Incidentally, the Times poll in Virginia’s 10th included not only an estimate of the results should everyone cast a ballot who says they’re certain to vote but also a total adjusted for the historic tendency to overestimate your likelihood to get to the polling place. Or, as they put it, “People who say they will vote, adjusted for past levels of truthfulness.”

In that universe of voters, Wexton wins by 8 points out of 307,000 ballots cast.