Are these experimental polling questions pointing to a Trump victory?
Researchers discuss two poll questions that are intended to “harvest the wisdom of crowds” to predict the election outcome. (Composite: Letty Avila. Image Source: iStock/teddyandmia.)

Are these experimental polling questions pointing to a Trump victory?

Two new questions added to the USC Dornsife Daybreak Poll this year, reported separately from the main results, predict that the national popular vote for president could be much closer than most polls are predicting. They also suggest Trump will once again win the election in the Electoral College. [7¼ min read]
ByJim Key

This election season, the USC Dornsife Daybreak Poll is reporting predictions for the presidential election based on a few different methodologies.

The poll’s standard methodology is to ask a panel of participants what the chance is that they will vote for each of the candidates running for president. It’s known as the “voter intention question” — the question asked by most polls to predict the share of the popular vote that each candidate will receive.

In this year’s Daybreak Poll, researchers are asking participants two additional questions that are intended to, as they say, “harvest the wisdom of crowds” to predict the election outcome. The “social-circle question” asks respondents to report the percentage of their social contacts they expect to vote for each of the candidates. The other one, known as the “state winner question,” asks participants who they think will win the election in their state.

The researchers studying the data related to these “wisdom-of-the-crowd” questions are Wändi Bruine de Bruin of the USC Dornsife Center for Economic and Social Research and USC Schaeffer Center for Health Policy and Economics, Mirta Galesic and Henrik Olsson of the Santa Fe Institute, and Drazen Prelec of the Massachusetts Institute of Technology. They answered the following questions about these two new methodologies.

What’s the value of asking people which candidate they expect other people in their social circle to vote for?

From our previous research on social judgments, we learned that people seem to know their immediate social circles quite well. Their answers about the distribution of income, health status — even the relationship satisfaction of their friends, family and acquaintances — were often in the right ballpark. And when we averaged the data from their responses across a large national sample, it provided a surprisingly accurate picture of the overall population.

So, we’re hoping to learn whether it’s possible, from this question, to harvest this “wisdom of one’s own crowd” to predict the election as well as, or even better than, the standard questions about one’s own voting intentions.

Have responses to this question successfully predicted any previous election outcomes?

Yes, in all five of the elections in which we tested this question, the social circle question predicted election outcomes better than traditional questions about voters’ own intentions. These five elections were the 2016 U.S. Presidential election, the 2017 French Presidential election, the 2017 Dutch Parliamentary election, the 2018 Swedish Parliamentary election, and the 2018 U.S. election for House of Representatives. 

In both the U.S. elections, the social-circle question predicted national and state level results better than the “own intention” question in the same polls. In fact, data from the social-circle question in 2016 accurately predicted which candidate won each state, so it predicted Trump’s electoral college victory.

It’s important to acknowledge, however, that we’re still studying the predictive capability of this question and we don’t know if it will be successful this election cycle. It’s possible the pandemic could impact responses to this question because reduced interaction with friends and family might hinder the ability to predict how they’ll vote.

Another factor that might diminish the predictive capability of the social circle question this year is the experience with the 2016 election. Some people might overestimate the number of Trump supporters in their social circle and expect Trump to have a surprise victory again.

Why do you think the social-circle question has successfully predicted previous elections?

We believe there are three main reasons. By asking people how their social contacts will vote, we’re implicitly gaining a larger and more diverse sample of participants. 

Second, it can be less embarrassing or intimidating for someone to tell a pollster that their friends plan to vote for an unpopular candidate than to report their own intention to vote for that candidate. 

Third, we are all influenced by our social contacts. Even if we report an intention to vote for a candidate other than the one most of our friends support, there is a chance they will eventually persuade us to vote for their candidate. So, asking about the voting intention of someone’s social circle could offer a glimpse of that person’s future beliefs, sort of like a crystal ball.

Which candidate is predicted to win, based on responses to the social-circle question so far?

The social-circle question is predicting Biden will win the popular vote, but by a much slimmer margin than what’s being predicted by the standard voter intention question — in the USC Dornsife Daybreak Poll and most others — regarding how poll participants themselves plan to vote.

Graph shows predicted number of votes for President Trump (red line) and former Vice President Biden (blue line) based on poll respondents’ expectations of how their social contacts’ will vote.

When we calculate how many electoral votes each candidate could get based on state level averages of the own-intention and social-circle questions, it’s looking like an Electoral College loss for Biden. We should note that our poll was not designed for state-level predictions, and in some states we have very few participants. Even so, in 2016 it predicted that Trump would win the electoral vote

In fact, in the last presidential election the social-circle question was more successful than both the own-intention question and aggregate polls in predicting winners of four of five swing states that unexpectedly went to Trump (Michigan, North Carolina, Pennsylvania and Wisconsin). The own-intention question accurately predicted just three of the swing states and aggregate polling failed to accurately predict any of them.

What about the question asking poll participants how they expect people in their state will vote?

This state-winner question produces even more pessimistic Electoral College results for Biden. It’s possible, however, that respondents to this question may be even more susceptible to the belief that, because of Trump’s surprise victory in the last election, he’ll win again. They might have such a strong mistrust of polls that they believe Trump will win because most polls show that he won’t.

How likely do you think it is that the results of the 2016 election, which surprised so many people, are impacting the results of your wisdom-of-the-crowd questions this time?

We investigated two types of beliefs that might be influencing the answers to these questions by our poll participants. There’s the “shy voter belief” that some Trump voters are not being counted in polls, and there’s the “magic candidate belief” that one of the candidates, typically Trump, will achieve a last-moment surge in votes.

To evaluate the potential impact of the shy voter belief on the responses of poll participants, we asked them three questions: What percent of their social contacts might be embarrassed to admit to pollsters their opinions about Trump or Biden, what percent might fear harassment if they admit these opinions, and what percent might want to obstruct polls by misreporting who they will vote for?

On average, our participants believe that people in their social circle might be more reluctant to admit their support for Trump than for Biden. However, there are interesting differences based on the participants’ own voting intentions. 

Biden supporters think friends and family members might be embarrassed to acknowledge their support of Trump. Trump supporters think that other people in their social circle who support the president’s reelection might fear harassment or intentionally obstruct pollsters; they don’t think embarrassment for supporting the president is a strong factor.

To gauge the potential impact of the magic candidate belief on the responses of poll participants, we asked them three other questions: What are the chances that either candidate will achieve a last-minute unexpected lead, what are the chances that either candidate will cause a lower voter turnout, and what are the chances that a candidate will challenge the election results even if they’re clear?

We found strong partisan differences and more pessimism among Biden supporters. Those who support Trump have a much stronger belief in a last-minute lead for their candidate than do those who support Biden. One reason for that might be a perception among Biden supporters that their candidate already has a large lead. Biden supporters are also much more likely to think their candidate’s opponent will challenge the election results than supporters of Trump.

In sum, the results suggest that optimism among Trump supporters stems from their expectation of a “silent majority” of Trump voters and of an unexpected (“magic”) last-minute Trump lead. Biden supporters’ pessimism stems from their expectation of lower voter turnout and the possibility that Trump will challenge election results. Taken together, these shy voter and magic candidate beliefs might be one reason why wisdom-of-crowds methods predict such a close race.

What’s next for your study of the predictive capability of these questions?

Prior to the election, we will report our predictions based on our wisdom-of-the crowd questions, and of course, post-election we’ll analyze the data from the responses to gauge how effective they were in predicting the outcome.

We believe that each question contains an important piece of the election puzzle; data from the responses to all three might be the most predictive.