Results of three quantitative predictions based on past regularities about voter turnout at the French 2009 European election

Twelve turnout rates of French national elections by municipality have shown statistical regularities, neither depend on the nature of the election, nor on the national turnout level. Three quantitative predictions about voter turnout at the French 2009 European election were made in arXiv:0905.4578. Here, we give the results of these three predictions. Each one is confirmed by real measures.

💡 Research Summary

The paper investigates whether simple statistical regularities observed in French municipal‑level turnout data across twelve past national elections can be used to predict the turnout characteristics of a future election, namely the 2009 European Parliament election. The author defines three quantitative indicators: τ α, a measure of turnout variability for each municipality α; σ α0, another variability metric derived from the same data; and π α0, a measure related to the average turnout level. Using the historical dataset, the author first calculates the overall standard deviation of τ α across all municipalities, obtaining a mean value of 0.376 with a standard error of 0.019 (95 % confidence interval 0.338–0.414). The second regularity concerns the correlation of σ α0 between any pair of different elections, denoted C_{t_i,t_j}(σ 0). The average correlation across the twelve past elections is 0.567 ± 0.058 (range 0.451–0.683). The third regularity is the correlation between σ α0 and π α0, which is 0.645 ± 0.026 (range 0.593–0.697).

Based on these three regularities, the author makes explicit predictions for the 2009 European election: (1) the standard deviation of τ α should be close to 0.376; (2) the average correlation between σ α0 for 2009 and each of the twelve previous elections should be about 0.567; and (3) the correlation between σ α0 and π α0 for 2009 should be near 0.645. The actual election data, released by the French Ministry of the Interior, show a turnout of 41.4 % among 4 241 106 registered voters. The measured values are: a τ α standard deviation of 0.360, an average C_{t_i,2009}(σ 0) of 0.577, and a σ–π correlation of 0.657. All three observed figures fall within the two‑sigma prediction intervals, confirming the author’s hypotheses.

Methodologically, the study treats each municipality as an independent statistical unit, leveraging the large sample size (tens of thousands of municipalities) to obtain precise estimates of means, variances, and correlations. The use of two‑sigma (≈95 % confidence) intervals provides a transparent benchmark for assessing prediction success. The close agreement between predicted and observed values suggests that the turnout variability captured by τ α and σ α0 is governed by underlying structural factors that are stable across election types (presidential, legislative, European) and independent of the overall national turnout level.

The paper’s contribution lies in demonstrating that a physics‑inspired statistical approach can yield robust, quantitative forecasts for social phenomena traditionally studied qualitatively. By showing that simple aggregate statistics derived from past elections retain predictive power for a new electoral context, the work supports the broader agenda of “statistical physics of social dynamics.” However, the author does not provide the explicit mathematical formulation of τ α, σ α0, or π α0, nor does the paper detail the estimation procedures (e.g., handling of outliers, weighting by municipality size). This limits reproducibility and the ability to test the model on other countries or election systems. Moreover, while the predictions succeed for the French case, the universality of the identified regularities remains an open question; cross‑national validation would be required to claim a general law of voter turnout.

In summary, the study confirms three empirically observed regularities in French municipal turnout data and validates them with the 2009 European Parliament election results. The findings illustrate that statistical regularities can be harnessed for accurate, a‑priori predictions in electoral studies, bridging the gap between physics‑based modeling and political science. The work encourages further exploration of simple, data‑driven models for forecasting social behavior, while also highlighting the need for greater methodological transparency and broader comparative testing.