Should French pollsters be using Mister P?

2026-04-27 · Source: Statistical Modeling, Causal Inference, and Social Science · Field: Science & Research — Mathematics & Computational Sciences, Social Sciences & Behavioral Studies · Depth: Advanced, quick

Summary

An anonymous French statistics student developed a poll aggregator and applied it to the last 100 days of the past five French presidential elections (2002-2022). The student utilized a Kalman-RTS smoothing algorithm with cross-validated noise levels, developed by Aki Vehtari, and validated it on simulated data. The smoothed results, derived from Wikipedia pages of French polls, are publicly available along with the code. Analysis of the plots reveals periodic opinion movements, potentially due to non-response, and surprisingly large shifts in candidate support (e.g., 10% in under 100 days). The student advocates for the adoption of Multilevel Regression and Poststratification (MRP) in French polling, citing the industry's current practice of manually adjusting final results by one point based on "pifomètre" (nosemeter) and the potential for MRP to enable predictions for legislative and municipal elections, which are currently neglected due to cost and complexity with quota sampling.

Key takeaway

For French pollsters seeking to modernize their methodology and expand their predictive capabilities, adopting Multilevel Regression and Poststratification (MRP) is crucial. This shift would eliminate subjective "pifomètre" adjustments and enable accurate forecasting for legislative and municipal elections, which are currently too difficult and expensive with traditional quota sampling. Your organization should investigate integrating MRP to enhance data integrity and broaden market reach.

Key insights

MRP offers a robust alternative to traditional French polling methods, addressing issues like manual adjustments and limited scope.

Principles

• Opinion data can exhibit periodic movements.
• Manual adjustments (pifomètre) introduce bias.
• MRP can expand polling to complex elections.

Method

The student applied a Kalman-RTS smoothing algorithm with cross-validated noise levels to raw poll data, then used the smoothed results to analyze trends and advocate for MRP.

In practice

• Implement Kalman-RTS for poll data smoothing.
• Use MRP for granular election predictions.
• Cross-validate noise levels in smoothing algorithms.

Topics

• French Polling Industry
• Multilevel Regression and Poststratification
• Kalman-RTS Smoothing
• Presidential Elections
• Opinion Polling Reform

Best for: Data Scientist, Research Scientist, AI Student

Related on AIssential

• Survey Statistics: it is (still) the people — Synthetic past vote (X**) improves voter support estimation by addressing missing or inaccurate recalled vote data.
• Survey Statistics: dCV for MRP ? — Design-based cross-validation improves model assessment by respecting survey design elements like PSUs and strata.
• Use AI This Election — AI can effectively personalize complex political research, significantly improving voter decision-making.
• Plan Before You Trade: Inference-Time Optimization for RL Trading Agents — FPILOT enhances RL trading agents by integrating inference-time optimization with price forecasts, improving returns and risk metrics.
• Q&A: Do AI and bogus respondents threaten polling’s future? — The rise of AI-generated and faked responses in polling necessitates a return to rigorous, probability-based human sampling for trustworthy public opinion data.
• MA$^{2}$P: A Meta-Cognitive Autonomous Intelligent Agents Framework for Complex Persuasion — MA$^{2}$P is a meta-cognitive multi-agent framework improving complex persuasion by inferring latent states and adapting strategies.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Statistical Modeling, Causal Inference, and Social Science.