Polling and Democracy: Do Polls Help Democracy?: The oldest argument, with new evidence

In brief

Election polls do two separate jobs, each with its own effect. One: they give politicians a check on their own guesses about public opinion, which matters because representatives misplace the majority on a policy 29% of the time and are no better at the top of the party. And two: they also give voters a truer picture of the race, which helps supporters of parties near a representation threshold work out whether a vote will win a seat. The cost is that polls can demobilise people holding a minority view, though the evidence measures willingness to act, not turnout.

How to use this

Before treating a poll as a verdict on what the public thinks, ask what job it is doing. If you advise a party, test the candidate's reading of their own electorate against data, and expect electorally safe representatives to carry less accurate pictures. When a poll lands mid-campaign, keep the two effects apart: a voter who updates what they expect to happen has not changed their vote. Before quoting a demobilisation finding, check whether it measured willingness to act rather than turnout. To read a party near a representation threshold, watch small day-to-day movements in its own numbers rather than the national leaderboard.

What the story is about

Start with a study that asked politicians a straightforward question about their own voters. Sevincer et al. (2026) emailed 1,599 sitting German politicians and asked what the public would accept on climate policy. Then they compared the answers with two samples built to mirror the German public. The politicians underestimated public support for climate taxes and climate laws, and for a willingness to contribute 1% of income. The underestimate was largest in the policy areas that would cut emissions most. Walgrave and Soontjens (2025) found the same pattern among local representatives in Belgium.

That could be one country's bad luck. Walgrave et al. (2023) checked across four countries, surveying 866 politicians and comparing more than 10,000 of their estimates with real public support. The average miss was 17.6 percentage points. Citizens who did the same task missed by 20.9, and a random guess would have been worse again, so politicians were doing better than nothing. The size of the remaining error is the point: they put the majority of the public on the wrong side of a policy 29% of the time. Otjes and Rasmussen (2026) compared roughly 10,000 estimates from representatives in Denmark and the Netherlands and found that "misperceptions are widespread and associated with a tendency for representatives to project their own opinion onto the public".

Perhaps the leaders at the top know better. Walgrave et al. (2025) tested that across four countries, comparing leaders with backbenchers on every definition of leadership they could measure, from party office to executive office, current and past. Their conclusion was that "leading politicians do not possess a special public opinion rating skill". The error belongs to the job rather than to the person holding it. What does change politicians is information about their own voters. Sevenans (2021) ran an experiment with members of the Belgian parliament and found that politicians told their position opposed a majority of their electorate moved their stated position.

The standing of a party in the polls does very little to how people say they will vote. Van der Meer, Hakhverdian and Aaldering (2016) tested this with 23,421 Dutch voters using real poll results and concluded that "simple polling outcomes by themselves do not affect subsequent vote intentions". Lammers et al. (2022) found that both things happen, in different people: some voters drift to the party the polls show winning, while others drift to the party the polls show losing. What polls do reliably deliver is a more accurate picture of the race. Stoetzer et al. (2024) supplied national poll numbers and past results to voters in Germany's 2021 election and found their expectations about local candidates became more accurate.

Numbers also work on people whose side is losing. Farjam and Loxbo (2024) gave US participants real money to donate to advocacy groups on seven contested issues, then showed them poll results. They report that "we demonstrate that seeing polls generally demobilizes people with minority views". Moderate voters holding the minority view abandoned it and moved to the majority; people at the ideological extremes did not budge. Note that this is willingness to act and give money, not turnout. Polls also help voters near a representation threshold work out whether a vote will produce a seat. Verdoes (2026) found in Norway's 2021 election that supporters of parties polling around the 4% national threshold defect when their party stands no chance locally.

That 17.6-point figure is a measurement, and the way you ask changes its size. Dias, Lucas and Sheffer (2026) asked politicians to draw the distribution of their constituents' preferences rather than give a single figure, and the size of the typical miss halved. The studies also come largely from one research programme, mostly one team, and mostly European: Belgium, Germany, Switzerland, Canada, Denmark, the Netherlands and the UK, plus an 11-country elite survey. Independent teams have replicated the central result, but the benchmark is a survey that carries its own error. Walgrave et al. (2023) recomputed accuracy from confidence intervals rather than single figures, and the inaccuracy survived.

The real-campaign record is thin. Barnfield (2020) reviewed the concept and settled on a definition: "an individual-level change in vote choice or turnout decision towards a more or increasingly popular candidate or party", brought about at first by that popularity. The definition matters, because a study that measures a change in what people expect to happen is not measuring a bandwagon effect at all, and a great deal of the public argument mixes the two. Unkelbach, John and Vogel (2023) followed Germany's 2021 campaign for 55 days against published polls, party by party, and reported that "we found no evidence for bandwagon effects". The most famous real-world version, momentum from an early primary win, also fails a careful test. Kamphorst et al. (2026) compared candidates just above and just below the victory line and found that "winning in Iowa or New Hampshire has no lasting impact on subsequent contests".

Taken together, election polls do two things at once. They give politicians a check on their own guesses about the public, and they give voters a clearer picture of where the race stands, which helps some coordinate and leaves others less willing to act on a minority view. Their effect on democracy is neither simply good nor simply bad. It depends on which of those two jobs is doing the work.

So what

Polls do one job on what politicians know and a different job on what voters do. No study has yet followed a published poll all the way through to a vote in a legislature; every measure stops at what politicians believe, say or post on social media. The evidence also comes from a particular slice of the world. It is mostly peer-reviewed and reasonably strong, but it is drawn from about a dozen democracies, all wealthy, mostly in Europe and North America. Nothing here covers Africa, Latin America, South or Southeast Asia, or any non-democracy.

For political parties

Accuracy follows incentives more than it follows information supply. Butler et al. (2025) found that "politicians have a better understanding of public opinion when they think the issue matters to voters", which is about belief rather than measurement. Hug et al. (2025) found that "representatives who are electorally safe hold less accurate perceptions of voters' policy preferences". The picture they carry is also skewed towards the well connected: Sevenans et al. (2025) found that "politicians' perceptions are closer to the preferences of the richer than to those of poorer people", and Marié (2026) traces part of that skew to a politician's own background and contacts. The test is whether the polling serves the decision or the answer the party already preferred.

For government

Whether a correct number changes anyone's mind depends on the context it arrives in. Go et al. (2026) ran preregistered experiments in Japan and South Korea with 7,000 people, supplying accurate information about gender disparities. It changed opinion in South Korea, especially among women, and barely moved it in Japan. The authors' conclusion is that "the effectiveness of informational interventions depends on contextual factors". Salience and the surrounding information environment decide whether a number lands. Nobody has measured the opposite case either: no study has taken published polling away from politicians and watched what happens to their accuracy.

Case studies

Germany's 2025 federal election shows what a poll number does for a voter near the line. Three parties were polling around the 5% threshold for entering parliament: The Left, the FDP and the BSW. Only The Left cleared it. Krause, Thomeczek and Wagner (2026) matched daily published poll scores to daily interviews and found voters responding to small day-to-day shifts in The Left's and the BSW's numbers, and not to the FDP's. The number was not pulling anyone towards a winner. It was telling millions of voters whether a vote for the party they preferred would produce a seat at all.

References

Barnfield, M. (2020) 'Think twice before jumping on the bandwagon: clarifying concepts in research on the bandwagon effect', Political Studies Review, 18(4), pp. 553–574. Available at: https://doi.org/10.1177/1478929919870691 (Accessed: 14 September 2026).

Butler, C., Walgrave, S., Soontjens, K. and Loewen, P.J. (2025) 'Politicians are better at estimating public opinion when they think it is more salient', Party Politics, 31(3), pp. 410–422. Available at: https://doi.org/10.1177/13540688241239625 (Accessed: 14 September 2026).

Dias, N., Lucas, J. and Sheffer, L. (2026) 'Beyond the mean: how thinking about the distribution of public opinions reduces politicians' perceptual errors', Political Science Research and Methods, advance online publication, pp. 1–16. Available at: https://doi.org/10.1017/psrm.2025.10069 (Accessed: 14 September 2026).

Farjam, M. and Loxbo, K. (2024) 'Social conformity or attitude persistence? The bandwagon effect and the spiral of silence in a polarized context', Journal of Elections, Public Opinion and Parties, 34(3), pp. 531–551. Available at: https://doi.org/10.1080/17457289.2023.2189730 (Accessed: 14 September 2026).

Go, M.H., Kweon, Y., Miwa, H. and Ono, Y. (2026) 'Correcting misperceptions across contexts: the political impact of gender inequality information in Japan and South Korea', Public Opinion Quarterly, 90(SI), pp. 1114–1143. Available at: https://doi.org/10.1093/poq/nfag030 (Accessed: 14 September 2026).

Hug, S., Varone, F., Helfer, L., Walgrave, S., Soontjens, K. and Sheffer, L. (2025) 'The role of politicians' perceptual accuracy of voter opinions in their reelection', Legislative Studies Quarterly, 50(1), pp. 107–122. Available at: https://doi.org/10.1111/lsq.12471 (Accessed: 14 September 2026).

Kamphorst, J., Davenport, A., Hagley, M., Dinas, E. and van de Rijt, A. (2026) 'Is winning the first primaries of primary importance? A regression-discontinuity approach', Political Science Research and Methods, advance online publication, pp. 1–17. Available at: https://doi.org/10.1017/psrm.2026.10096 (Accessed: 14 September 2026).

Krause, W., Thomeczek, J.P. and Wagner, A. (2026) 'Polling and parliamentary representation: strategic voting in the 2025 German federal election', Electoral Studies, 104, 103142. Available at: https://doi.org/10.1016/j.electstud.2026.103142 (Accessed: 14 September 2026).

Lammers, J., Bukowski, M., Potoczek, A., Fleischmann, A. and Hofmann, W. (2022) 'Disentangling the factors behind shifting voting intentions: the bandwagon effect reflects heuristic processing, while the underdog effect reflects fairness concerns', Journal of Social and Political Psychology, 10(2), pp. 676–692. Available at: https://doi.org/10.5964/jspp.9241 (Accessed: 14 September 2026).

Marié, A. (2026) 'Why do politicians overestimate richer voters' opinions? Social background and interest group influence', Swiss Political Science Review, 32(1), pp. 102–128. Available at: https://doi.org/10.1111/spsr.70021 (Accessed: 14 September 2026).

Otjes, S. and Rasmussen, A. (2026) 'Sources of elected representatives' (mis)perceptions of public opinion: the role of self-projection and interest groups', The Journal of Politics, 88(1), pp. 79–93. Available at: https://doi.org/10.1086/734528 (Accessed: 14 September 2026).

Sevenans, J. (2021) 'How public opinion information changes politicians' opinions and behavior', Political Behavior, 43(4), pp. 1801–1823. Available at: https://doi.org/10.1007/s11109-021-09715-9 (Accessed: 14 September 2026).

Sevenans, J., Marié, A., Breunig, C., Walgrave, S., Soontjens, K. and Vliegenthart, R. (2025) 'Are poor people poorly heard?', European Journal of Political Research, 64(3), pp. 1326–1350. Available at: https://doi.org/10.1111/1475-6765.12750 (Accessed: 14 September 2026).

Sevincer, A.T., Hostlowsky, L., Styhler, F. and Hofmann, W. (2026) 'Public support for climate action is underestimated in the German political domain', Communications Earth & Environment, 7(1)., 543. Available at: https://doi.org/10.1038/s43247-026-03721-7 (Accessed: 14 September 2026).

Stoetzer, L.F., Kayser, M.A., Leininger, A. and Murr, A.E. (2024) 'Voters' expectations in constituency elections without local polls', Public Opinion Quarterly, 88(2), pp. 408–418. Available at: https://doi.org/10.1093/poq/nfae015 (Accessed: 14 September 2026).

Unkelbach, F., John, M. and Vogel, V. (2023) 'Jumping on the bandwagon: the role of voters' social class in poll effects in the context of the 2021 German federal election', Politische Vierteljahresschrift, 64(1), pp. 51–78. Available at: https://doi.org/10.1007/s11615-022-00417-3 (Accessed: 14 September 2026).

Van der Meer, T.W.G., Hakhverdian, A. and Aaldering, L. (2016) 'Off the fence, onto the bandwagon? A large-scale survey experiment on effect of real-life poll outcomes on subsequent vote intentions', International Journal of Public Opinion Research, 28(1), pp. 46–72. Available at: https://doi.org/10.1093/ijpor/edu041 (Accessed: 14 September 2026).

Verdoes, A. (2026) 'District expectations and strategic defection in two-tiered proportional systems: the case of the 2021 Norwegian election', Electoral Studies, 100, 103061. Available at: https://doi.org/10.1016/j.electstud.2026.103061 (Accessed: 14 September 2026).

Walgrave, M. and Soontjens, K. (2025) 'Responsive nor responsible? Politicians' climate change policy preferences and public opinion perceptions', Environmental Politics, 35(3), pp. 522–546. Available at: https://doi.org/10.1080/09644016.2025.2492439 (Accessed: 14 September 2026).

Walgrave, S. et al. (2023) 'Inaccurate politicians: elected representatives' estimations of public opinion in four countries', The Journal of Politics, 85(1), pp. 209–222. Available at: https://doi.org/10.1086/722042 (Accessed: 14 September 2026).

Walgrave, S., Sevenans, J., Varone, F., Sheffer, L. and Breunig, C. (2025) 'Do political leaders understand public opinion better than backbenchers?', British Journal of Political Science, 55., e9. Available at: https://doi.org/10.1017/s0007123424000929 (Accessed: 14 September 2026).