When polling potential voters, determining whether they will actually vote on Election Day is just as important as their candidate choice. In new survey research, Jaroslav Tir and Shane Singh examine the effect of a potential influence on poll respondents’ voting claims – news about potential threats to US security. They found that survey respondents who had been exposed to such news were more likely to claim to have voted in a past election, if they had not actually voted, than those who had not. They also find that those who say they voted less often in the past were more likely to say they will do so in the 2024 election if exposed to news about security threats.

With the 2024 US presidential election rapidly approaching, opinion polling organizations are working in overdrive to assess the chances that various contenders will win their intended offices. But the accuracy of such predictions depends on whether those who are polled end up voting. Political campaigns understand this very well, as they spend vast amounts of money not only to persuade individuals that they should support their candidate, but that they should also actually turn out to vote for them. A high-profile example of a dramatic misestimating of voter turnout intentions was in the lead up to the 2016 election which saw former Secretary of State, Hilary Clinton leading in pre-election polling and then losing to Donald Trump in key swing states. The reported closeness of the upcoming presidential election underscores the importance of turnout assessments as a key component of providing accurate predictions of the election outcome.

Threats of terrorism can influence claims about voting

In our work, we illustrate how opinion polls may produce misleading assessments of citizens’ claims about election turnout. If, for example, some respondents systematically tell pollsters that they intend to vote but then they do not end up doing so, projections about winners of various races will be inaccurate. Building on research identifying respondent-level correlates of such misreporting, we show that contextual factors may also motivate poll respondents to exaggerate their voting claims. Our recent cross-national study shows that individuals who have been exposed to the threat from terrorism are more likely to claim to have voted in a past election than those who experienced no such threat—even if they did not. This results in what is known as vote overreporting.

We expand on these insights in the context of the upcoming 2024 election. Over the past several months, Americans have been inundated with news about threats to US security and its interests from the likes of Russia, China, Iran, and North Korea, sometimes in connection with the ongoing conflicts in Israel, Lebanon, Gaza, Ukraine, and Taiwan. A problem for polling firms is that they have no control over what kinds of related news (if any) poll participants consume prior to responding to their queries. Following the logic from our study, exposure to news about threats to US security would lead poll respondents to exaggerate their level of participation, potentially skewing election projections.

We embedded a survey experiment in an April 2024 YouGov poll conducted with the participation of about 2,000 eligible voters. Some survey respondents were randomly assigned to a news-style vignette referencing threats to US security, while others were assigned to an innocuous control condition referencing US Geological Survey map production. We subsequently asked the respondents whether they intended to vote in the upcoming 2024 election and if they did, which candidate.

Looking at responses that selected a candidate versus those that declared no intention to vote, and adjusting for individual-level attributes to increase precision, shows that referencing threats to US security has a small and marginally statistically significant positive effect on the claim of future voting participation. This is shown in Figure 1.

Figure 1 – Estimated effect of foreign threat exposure

Threats cause people who say they voted less often in the past to say they will vote

A deeper dive into these results reveals some troubling trends. In the same poll, YouGov inquired about general voting participation, asking respondents to rank themselves on a scale of how often they vote when there is a presidential election.

Importantly, YouGov asked this question before our survey experiment, meaning that we can condition the effects of referencing threats to US security on responses. Using the general self-reported participation history variable, we see that respondents were more likely to say they would vote in 2024 as a result of exposure to threat if they were less likely to have said that they voted in the past, or even not at all. The impact of referencing threats to US security ranges from about four to seven percentage points and is statistically significant for all but the “always vote” past participation level. These findings are shown in Figure 2 below.

Figure 2 – Estimated effect of foreign threat exposure on intended 2024 turnout

News of security threats may mean people overstate the likelihood they will vote

Our findings raise questions about the accuracy of inferences about a candidate’s likelihood of victory made with reference to opinion polls, including those pertaining to the upcoming American presidential election. Achieving an election victory depends not only on a citizen’s preferences over various candidates but also, critically, on whether the citizen will actually be a voter come November. Our results suggest that exposure to news about threats to US security and interests can systematically bias respondents to overstate their intent to turn out to vote. The failure to correct for this bias when making opinion poll-based projections about probable election winners will lead to inaccuracies—especially in closely contested races like the upcoming US presidential election.

“’Cause some days you just have to have #” (CC BY-NC 2.0) by ltenney1225

Correcting for this bias is unlikely to be straightforward, however. Pollsters cannot control whether respondents may have been exposed to news about threatening events in the news prior to participating in a poll. This makes it difficult to ascertain which individuals may have been “treated” by real-world circumstances and which ones remain in a more tranquil state resembling our control condition. Asking respondents if they have recently seen threatening news is also unlikely to help, as this very question may trigger the effect we describe here. And while many polling firms already employ algorithms that make projections based on “likely voters,” the recruitment of reluctant voters, who were missed by pollsters, is sometimes credited with helping propel Donald Trump to his unexpected victory in 2016. And the 2024 Trump campaign is banking on being able to identify and turn out citizens who have not previously voted. Accurately gauging the effectiveness and impact of such efforts in pre-election polls is difficult, yet the accuracy of projections ultimately depends on it.

How news about threats affects vote choice

We also looked at whether exposure to news stories about threats to US security also affects claims about intended vote choice. In a Joe Biden versus Donald Trump contest (recall that our data are from an April 2024 survey, before President Biden dropped out), referencing threats has no significant effect on claimed intentions to vote for Biden. But it does depress, by about four percentage points, the probability of claiming an intention to vote for Trump. This effect is statistically significant. See Figure 3.

Figure 3 – Estimated effect of foreign threat exposure on vote choice

Discerning why the stated intention to vote for Donald Trump is lessened by referencing foreign threats is beyond the scope of this piece. Respondents may genuinely trust the former President less on foreign policy issues, or they could be providing what they perceive as a socially desirable response. Regardless, these findings raise the possibility that pollsters may underestimate the support for a candidate based on what news, if any, poll respondents may have consumed prior to participating in a survey. Current polling methods do not account for this type of challenge to the accuracy of inferences about both intended turnout and vote choices.

• Please read our comments policy before commenting.  
• Note: This article gives the views of the author, and not the position of USAPP – American Politics and Policy, nor the London School of Economics. 
• Shortened URL for this post: https://wp.me/p3I2YF-erP