Are Exit Polls Accurately Measuring the Vote Choice of the Youth Electorate?

In the aftermath of the 2016 presidential election, exit polls came under fire for their perceived deficiencies in accurately surveying the American electorate. According to an analysis by Thomas Edsall of the New York Times, that year’s exit polls undersampled white working-class voters and oversampled college-educated voters, thereby biasing the data and overestimating Democratic turnout.¹ Furthermore, exit polls in key swing states like North Carolina and Pennsylvania underestimated Trump support by margins greater than the margin of error, spurring additional criticism and calling into question the sampling methodology.²

At CIRCLE, we rely on exit polls on and immediately after Election Day to provide insight into how and how many young people voted. Because vote choice is not recorded in publicly available voter file data, statewide exit polls and national surveys of the American electorate are key tools to help us understand and explore how young people are participating in the political process. To that end, we wanted to interrogate how accurately exit polls have captured the youth vote in the past decade of national elections—as compared to other national surveys. Overall, we found that exit polls have been, for the most part, reliably accurate measures of youth vote choice on the national level.

We compared the vote choice data from national exit polls, which are conducted by Edison Research during presidential and midterm elections, to the vote choice data from two independent post-election national surveys: the American National Election Surveys (ANES) and the Cooperative Congressional Election Survey (CCES). Both are fielded in presidential election years, but only the CCES is fielded after midterms. Though both polls sample tens of thousands of Americans, each uses its own methodology with regards to weights and sampling distribution.

During midterm years, we examined the national House vote choice among youth (ages 18-29); in presidential years, we looked at presidential vote choice among the same range of young voters.

In both 2010 and 2012, the exit polls closely matched the youth vote choice data in the post-election surveys. In 2010, the estimated Democratic and Republican House vote shares in the exit poll and in the CCES were within a tenth of a percentage point of each other. In 2012, all three polls measured vote choice for the Democratic presidential candidate, President Obama, within a tenth of a percentage point of each other, and Republican presidential vote shares (i.e. votes for Mitt Romney), within one percentage point.

The exit polls from 2014 and 2016 did deviate slightly more from the post-election surveys. The 2014 midterm exit poll estimated the youth vote choice for Democrats four points higher than the CCES estimate, though the Republican vote choice diverged by less than two points. In 2016, exit polls reported Democratic vote choice 3.5 points lower and Republican vote choice 4.5 points higher than in the CCES, although they tracked closer to ANES data.

Of course, we can also measure the accuracy of the exit polls (and of the post-election surveys) by comparing their vote choice estimates to the election results as reported by the Federal Election Commission—though we can only perform this analysis for all voters, and not just for young people.

In all four elections, the actual results fell within the exit polls’ margins of error. While some state exit polls did underestimate votes cast for President Trump by large margins in 2016, the national exit poll in 2016 was off by just 1.1 percentage points for all voters. In the three previous election cycles, both the exit polls and the post-election surveys measured the actual level of support for each party to within 1 percentage point of the actual results.

Overall, we conclude that there is consistency in 2010 and 2012 across these surveys, but slightly less so in 2014 and 2016. And even though some state exit polls missed by a relatively wide margin in 2016, the exit polls generally estimated the actual vote choice nationally with some precision. As we prepare to use exit poll data to make calculations about youth participation in the upcoming 2018 midterms, we will continue to bear in mind that exit polls, like other surveys of the electorate, are estimates of how people actually voted. While all methods have their own benefits and challenges, this analysis buttresses our confidence that, for vote choice, exit polls are a reliable, accurate source of data on which to base part of our work.

---

[1] https://www.nytimes.com/2018/03/29/opinion/2016-exit-polls-election.html

[2] For more analysis of state exit polls from 2016, see Theodore de Macedo Soares, http://tdmsresearch.com/2016/11/10/2016-presidential-election-table/