ESRA 2025 Preliminary Program
All time references are in CEST
Validating Elections Polls |
| Session Organiser |
Professor Peter Selb (University of Konstanz)
|
| Time | Wednesday 16 July, 14:00 - 15:00 |
| Room |
Ruppert 005 |
Election polling offers ample opportunities to validate survey estimates. The population parameters of primary interest (such as vote shares and turnout rates) become known after the election. The availability of numerous similarly designed polls for a single election brings us closer to observing not only the population parameters but also the sampling distribution of their estimators. When focusing on voter turnout, linkages with voter files even allow us to observe the criterion variable at the respondent level.
This panel welcomes methodological contributions that leverage these opportunities to assess the accuracy and precision of election polls, individual error components, and their variability across polls and elections. We invite both case studies and comparative work that explore these themes.
Keywords: election polling, survey methodology, validation
Papers
Projecting and estimating the results of the European Elections 2024
Mr Nicolas Becuwe (Verian Group ) - Presenting Author
Mr Alban Versailles (Verian Group)
Publishing transparent and reliable information regarding turnout and election results as soon as possible on election day is an important democratic requirement for institutions, and a methodological challenge for researchers. Using a combination of pre-election polls, exit polls, and partial or preliminary results, Verian has built a model to deliver reliable estimates during the 2024 European Elections.
The European elections took place from the 6th to the 9th of June 2024 in the 27 Member States of the European Union. It is the largest transnational democratic election in the world and the second largest democratic election in terms of number of registered voters. For this event, the European Parliament contracted Verian to prepare and deliver the estimates on turnout and seat projections before the closure of voting stations in all the EU. These estimates and projections were published on a bespoke dedicated website built by Verian team for the European Parliament.
We are very happy to report that both the turnout estimate, and the seat projections matched very closely with the official results. The turnout estimate was 51% and the actual turnout was 50.74%. Whilst the first projection in terms of seats published by 8PM was close to the final composition of the European Parliament during the constitutive session in July 2024.
This paper will describe the methodology applied before and during the night to achieve precise estimates and results and evaluate the model performance during the 2024 elections. Finally, it will discuss the broader implications for the use of election polls data by researchers and institutions to provide transparent and reliable information to the public.
Vote choice bias in Swiss public opinion surveys: Underestimation of support for the Swiss People’s Party (SVP)
Ms Nursel Alkoç (University of Lausanne) - Presenting Author
Public opinion surveys are valuable instruments for analyzing political behavior, including voter turnout and vote choice. Their estimates are prone to various errors and biases, however. The literature on turnout bias is extensive, whereas the research on vote choice bias remains relatively sparse. Previous research has identified two different forms of vote choice bias: the ‘winner bias’, where respondents inaccurately overreport their support for winning parties, and the ‘spiral of silence’, where individuals underreport their support for controversial or socially undesirable, typically right-wing, parties. The present study offers the first thorough examination of vote choice bias for the winning right-wing Swiss People’s Party (SVP) across three major ongoing probability-based cross-sectional surveys conducted in Switzerland, namely, the Swiss Election Study (Selects), the European Social Survey (ESS), and the Measurement and Observation of Social Attitudes in Switzerland (MOSAiCH). The findings suggest that all three surveys underestimate the level of support for SVP, thereby providing evidence not of a winner bias, but rather a spiral of silence for this party. Comparatively, ESS predicts a significantly higher SVP vote share compared to Selects and MOSAiCH, thus aligning more closely with the actual SVP share. In the absence of voter verification data, the implications of these results on the accuracy of these surveys' predictions of political behavior are discussed.
