A voter casts a ballot at the Hamilton County Board of Elections on Oct. 10, the first day of early voting in Cincinnati. (John Minchillo/AP)

This November’s elections set a voter turnout record: 49 percent of the voter-eligible population showed up at the polls, the highest midterm turnout seen since 1914. Why?

Most commentators have concluded it was because voters thought the stakes were sky-high. With congressional control in the balance at a time when politics is highly polarized, many Democrats and Republicans thought the outcome was all but life-or-death for democracy. Pre-election polls found that about two-thirds of Americans believed this election was the most important midterm in their lifetimes; 93 percent of voters in battleground districts said their vote mattered just as much as in a presidential election; and enthusiasm about voting was at its highest level in any midterm in more than two decades.

That interpretation is consistent with the findings in our new book, “Why Bother? Rethinking Participation in Elections and Protests,” due out in January. Although political scientists have often emphasized that many people do not turn out because of the “costs” of voting (time, effort and so forth), we focus on the costs that people pay by not voting — especially when they believe much is at stake. Seen through that lens, the high turnout this year is hardly surprising.

We find that while voting is indeed costly, abstaining can impose costs as well. Many people suffer subjective and psychological costs if they don’t vote, which we call costs of abstention. These costs are in the form of a psychic tension, or dissonance, when people fail (or think they’ll fail) to take part — but that’s true only when they care a lot about the outcome. When they care a lot and also see the race as close, these moods intensify, pushing them still more strongly to go to the polls. This same logic applies to other forms of collective political participation, like protesting, as we found in major protests in Brazil, Turkey and Ukraine.

Here’s how we did our research

To understand how costs of abstention influence voting, we conducted a series of experiments. One was with 1,020 U.S. citizens recruited through Amazon.com’s Mechanical Turk platform in spring 2016. We randomly assigned participants into four groups. Participants in each group read one of four hypothetical gubernatorial election scenarios.

Two of the scenarios’ introductions illustrated how the governor can significantly affect our daily lives. These two explained a governor’s powerful capacities and asked respondents to imagine having a clear preference for one candidate over the other, without mentioning any other characteristics. The other two scenarios didn’t mention governors’ functions and asked respondents to imagine that both candidates would do an equally good job in office.

In each of those four scenarios, we gave respondents a different idea of how close the result would be. In two, we told them that reliable polls indicated that the race was very close; in the other two, it was pretty clear which candidate was going to win.

Then we then asked them, on a scale of 1 to 5 — in which 1 indicates little emotion and 5, intense emotion — how they would feel if they knew they would not vote in that election. Participants reported significantly more distress, upset and guilt when considering not voting in an important election than in an unimportant one. On average, they reported 0.4-points, or about 20 percent, higher levels of distress and upset. This was amplified when the election outcome was supposed to be close. Respondents were also more likely to state that they were going to vote when the election was important, regardless of whether the result was expected to be close or lopsided. Close elections boosted turnout, but only when the election was also depicted as important.

We also looked at surveys with “verified” turnout — the researchers checked respondents’ claims to have voted or abstained against public voting records. In the wake of the “credit crunch” in the United Kingdom, for example, Britons who reported being angry about the economy during the 2010 general elections voted in greater numbers, even after accounting for an array of known boosts to turnout — being highly educated or married or a union member, for example. Anger increased turnout especially among those who saw the economy as the most pressing problem facing the country.

Believing there’s a duty to vote kicks in when someone thinks an election is especially important

Decades ago, political scientists noticed that many people feel a duty to vote, and those who do are more likely to turn out. Scholars saw this as a stable trait that people pick up from their families or in school. This helps explain why people vote at all, despite the costs and inconvenience. But duty as a personal conviction doesn’t explain why turnout is higher in some elections than others.

The answer is that when someone thinks an election is extremely important, that sense of duty kicks in more forcefully — and they’re more likely to vote. In another experiment conducted ahead of 2014 U.S. gubernatorial elections, we recruited 504 U.S. citizens again through the Mechanical Turk, and manipulated respondents’ perception of the importance of the upcoming election (again by reminding them the powerful capacities of the governor) and the salience of a duty to vote.

Subjects who were primed to see the election as important felt a heightened sense of civic duty, even though the treatment said nothing about duty. Only the subjects who were primed to see the election as high stakes were more likely to vote than those in the control group. These results suggest that the duty to vote, in fact, is one component of costs of abstention.

Although this month’s elections saw record-high spending, that is almost surely not the reason turnout was so high. Consider the fact that though presidential campaigns spend the vast majority of time and money in battleground states and ignore safe ones, the difference in turnout between battlegrounds and safe states is surprisingly small — usually a few percentage points.

Rather, turnout in presidential elections likely surges because most people see those contests as crucial — high stakes — for the country. That gets them to the polls. Unusually, that’s how many citizens felt this year. For many people staying home this year would have been too psychologically costly.

Erdem Aytaç is assistant professor of political science at Koç University in Istanbul.

Susan Stokes is Blake Distinguished Service Professor at the University of Chicago, director of the Chicago Center on Democracy, and a co-founder of Bright Line Watch.