[Paper Review] Three quantitative predictions based on past regularities about voter turnout at the French 2009 European election

Christian Borghesi|ArXiv.org|May 28, 2009

Opinion Dynamics and Social Influence16 references3 citations

TL;DR

This paper presents three quantitative predictions for the 2009 French European Parliament election turnout based on statistical regularities observed in 12 previous national elections. Using a logarithmic transformation of turnout ratios per municipality, it predicts the standard deviation of turnout heterogeneity, its temporal continuity, and its correlation with local population distribution—each within a two-sigma confidence interval, independent of election type or overall turnout level.

ABSTRACT

The previous twelve turnout rates of French national elections by municipality show regularities. These regularities do not depend on the national turnout level, nor on the nature of the election. Based on past statistical regularities we make three predictions. The first one deals with the standard deviation of the turnout rate by municipality. The second one refers to the continuity in time of the heterogeneity of turnout rates in the vicinity of a municipality. The last one is about the correlation between the heterogeneity of turnout rates in the vicinity of a municipality and the population in its surroundings. Details, explanations and discussions will be given in forthcoming papers.

Motivation & Objective

• To identify consistent statistical patterns in municipal voter turnout across diverse French national elections.
• To predict the 2009 European Parliament election turnout using regularities from past elections, regardless of turnout level or election type.
• To test whether turnout heterogeneity remains stable over time and correlates with local population structure.
• To provide quantitative forecasts for turnout variation with uncertainty bounds based on historical data.

Proposed method

• Define a logarithmic transformation τ^α = ln(N⁺_α / N⁻_α) to normalize turnout ratios and avoid divergence.
• Measure the standard deviation of τ^α across ~36,200 French municipalities (communes) for each of 12 past elections.
• Assess temporal stability of local turnout heterogeneity by computing σ^α₀, the standard deviation of τ^β in the vicinity of each municipality α.
• Correlate σ^α₀ with π₀^α, the population distribution in the surroundings of each municipality α, across all central municipalities.
• Use mean and standard deviation of historical values to predict future values with a two-sigma error bar.
• Apply binomial simulation as a null model to compare observed correlations with random expectations.

Experimental results

Research questions

• RQ1 Does the standard deviation of municipal turnout (measured via τ^α) remain consistent across diverse French national elections with varying turnout levels?
• RQ2 Is the local heterogeneity of turnout (σ^α₀) temporally stable across elections, independent of election type?
• RQ3 To what extent is the local turnout heterogeneity (σ^α₀) correlated with the population distribution (π₀^α) in the vicinity of a municipality?
• RQ4 Can statistical regularities in past turnout data reliably predict future turnout variation in the 2009 European election?

Key findings

• The predicted standard deviation of τ^α across French municipalities for the 2009 European election is 0.376 ± 0.019, with a 95% prediction interval of [0.338, 0.414].
• The predicted correlation between local turnout heterogeneity (σ^α₀) and surrounding population distribution (π₀^α) is 0.645 ± 0.026, with a 95% interval of [0.593, 0.697].
• The temporal stability of σ^α₀ across elections is confirmed by consistent values over 15 years, indicating no global shift in turnout heterogeneity patterns.
• The observed correlation between σ^α₀ and π₀^α (0.645 ± 0.026) significantly exceeds the null model based on binomial randomness, suggesting a non-trivial spatial dependence.
• The distribution of τ^α − ⟨τ⟩ is remarkably similar across all 12 elections, indicating a universal statistical behavior independent of election type or turnout level.
• The use of τ^α as a transformed turnout measure reveals regularities not explained by simple binomial models, implying deeper underlying mechanisms.