Social-Circle Questions in Elections and Beyond

22 Oct 2025

“The imaginations which people have of one another are the solid facts of society[.]” — Charles Horton Cooley, Human Nature and the Social Order (1902)

The accuracy of most kinds of public opinion polls can’t be measured, but elections are an exception: ballots are counted, candidates win or lose, and margins are real. That makes election polling a good place to test and evaluate ideas we might also use to politically-relevant public opinion more generally, for example on issue prioritisation or policy preferences.

Social-circle surveying in election polls

In election polls, and in most polls, the default approach is asking people about what they themselves intend to do, often referred to as “own-intention”. An alternative approach, which is commonly known as social circle surveying, is to ask people about what they think other people will do. The exact wording of these kinds of social-circle questions varies, but in general they involve asking respondents what share of their friends, family, colleagues, or neighbors support a candidate.

Here is an example of the most basic kind of own-intention question, from an October 2024 poll in the US¹:

If the 2024 presidential election were held today, who would you vote for if the candidates were: Kamala Harris, the Democrat, [and] Donald Trump, the Republican.

And here is an example of a social-circle question, from a 2020 USC tracking poll² in the US:

Out of all your social contacts who live in your state and are likely to vote in the 2020 U.S. presidential election, what percentage do you think will vote for: Joe Biden (Democrat), Donald Trump (Republican), Someone else?

A forecast based on an own-intention question in the same USC poll missed the 2020 national popular-vote margin by roughly 5.5 points, while a forecast built from the social-circle question above missed the margin by just 1 point³. This improvement is notable given that a later polling industry report found that, on average, 2020 polls were off by 3.9 points⁴.

This isn’t a one-off. In a peer-reviewed study, a team from the Sante Fe Institute, University of Leeds, and University of Southern California (USC) fielded social-circle questions alongside standard questions in major national polls during the 2016 U.S. and 2017 French elections. In both settings, social-circle responses improved prediction accuracy relative to own-intention answers⁵.

In the U.S., they tracked vote shares more closely, called more state winners, and performed especially well in pre-defined swing states. Notably, one of the social-circle forecasts uniquely anticipated a Trump electoral-vote win. In France, social-circle questions outperformed own-intention ones in both rounds.

Now, one concern about applying social-circle surveying beyond elections is that many kinds of public public opinion research, such as studying issue prioritization or policy preferences, involve many more competing choices than two-candidate elections. However, a team from USC, Sante Fe Institute, Lund University, and Tilberg University found that social-circle questions still performed better than the own-intention questions in predicting two elections with many political parties: The Netherlands’ 2017 general election, which included 13 major parties, and the Swedish 2018 general election, which included eight major parties⁶.

Most recently, social-circle surveying gained attention⁷ when an anonymous French bettor reportedly commissioned a “neighbors” survey late in the 2024 U.S. cycle, asking swing-state respondents who they thought their neighbors would vote for⁸. Seeing a strong Trump advantage in those neighbor reports, he increased his exposure on prediction markets and later credited that signal for the win.

Why asking people about others can beat asking people about themselves

So, there is evidence that social-circle questions can forecast elections better than own-intention questions. Why might this be? There are three common arguments:

• More information per interview. Each respondent compresses dozens of contacts into one estimate. If those summaries are even moderately accurate, the survey captures more of the real distribution of views. This inclusion of additional contacts is often called the “implicit supersample”.
• Better coverage of hard-to-reach voters. Some people avoid polls or won’t disclose their own views. Asking what their friends or neighbors think can partially bridge into those networks and reduce blind spots.
• Less social-desirability pressure. Reporting on others is often easier than reporting on oneself, particularly on polarizing or stigmatized choices, which can bring estimates closer to realized outcomes. The “shy Trump voter” is a famous example of this phenomenon.

However, there may be a more fundamental reason that social-circle questions enable better election forecasting, and that’s because a person’s own views and behavior are in a dynamic relationship with their perception of those around them. For example, two results point to the same mechanism: people adapt their actions based on their perception of their social-circle.

From the study of the 2016 U.S. and 2017 French elections⁵ mentioned above:

Not all participants ended up voting for the candidate they announced as their favourite in the week before the election. For example, participants whose own intentions mismatched those in their social circles were less likely to eventually vote for their intended candidate. [O]ur overall results suggest that social-circle reports foretold a switch in voting intentions before it happened. Generally, changes in participants’ social circles over time predicted their later intentions to vote for specific candidates and to vote at all[.] This pattern of results was found for both Trump and Clinton voters, suggesting that participants’ perceptions of how social contacts would vote affected their own beliefs regarding the candidates.

Similarly, in public health, a national longitudinal study⁹ found that perceived social-circle vaccination coverage was associated with own vaccination in the same and subsequent flu seasons:

Our longitudinal analyses suggest that participants’ perceived social circle vaccine coverage was associated with their vaccination behavior as reported for the concurrent and subsequent flu seasons. […] Our findings suggest that participants who perceived greater social circle vaccine coverage followed the perceived social norm.

Together, these findings suggest that social-circle responses are doing double duty: they inform us about the population (because networks carry real information) and also about behavioral trajectories (because norms nudge choices). In elections, they can foreshadow late switching or actual voting; in health, they can predict future uptake.

As social creatures, we are sensitive to the norms we perceive in our communities, and we adapt our beliefs and actions in response to them. By asking about a person’s social-circle, we aren’t only learning about the person’s friends or colleagues, we’re learning about the social environment that shapes the person themselves.

Moving from elections to issues and policies

Now, we have seen how social-circle surveying works to improve election polling, and we can imagine how the same mechanisms may apply to other politically relevant opinion research, such as issue prioritization and policy preferences. We may be able to make simple adaptations of social-circle questions to surface what’s salient and how intense support is and do segmentation to reveal how attitudes are locally clustered or broadly shared, which are all crucial for agenda-setting and persuasion.

For example:

• Priority: “Of the people you’ve spoken with in the past 30 days, what percentage have discussed [issue]?”
• Support level: “Among your social contacts likely to vote, what percentage support [policy]?”

More concretely, these kinds of questions can help identify:

• issues that drive conversations in communities (high salience in social-circles),
• groups that are ready for mobilization (agreement within social-circles), and
• policies that can power broad-based campaigns (agreement across social-circles)

We are currently piloting projects in this area.

Our hope is that bringing new, validated approaches to issue and policy opinion research can transform the work of politicians, policymakers, journalists, and advocacy campaigns and improve democratic processes and outcomes by grounding them in accurate and actionable insights in the public and society.

❦

The teams at Athena Insights and Early Studies, based on their deep experience in policy campaigns and market research, have been collaborating on cutting edge public opinion research using social-circle surveying and other emerging methodologies. If you are interested in learning more, please reach out.