Polling methodology under scrutiny

As the presidential election approaches, the use of opinion polls to measure voter intentions has become a central element of political commentary. Months before the election, nearly all institutes use the same question to gauge public sentiment: "If the first round of the presidential election were to take place next Sunday, for which of the following candidates would you be most likely to vote?" This standard phrasing is often presented as a convenience for data collection, but it is based on a hypothetical scenario that permanently clouds democratic debate.

Hypothetical scenarios in polling

The methodology of using a hypothetical scenario is problematic. A hypothetical proposal is based on a premise explicitly contrary to facts, as the election does not take place next Sunday except during the week of the election. This setup imposes a double forcing: a temporal forcing, where citizens are asked to act in an artificially urgent situation, bypassing the long-term deliberation, official campaign, and debates; and a political offering forcing, where the institute must provide a closed list of personalities, even though official candidacies are only validated by the Constitutional Council a few weeks before the election.

Limitations of hypothetical polling

The analyst does not measure a real behavior or disposition in the present world, but the result of a mentally constrained simulation, as imaginary as the hypothetical scenario. Thus, an unverified (and often invalidated) bet is made that the decision rule adopted by an individual in this framework is the same as the one they will adopt on the real day of the election. This bet is not unique to electoral polls. Experimental economics has documented, under the name of "hypothetical bias," the systematic gap between what individuals say they would do in a hypothetical scenario and what they actually do once the situation becomes concrete.

Methodological reductions in polling

To understand the magnitude of this artifact, it is necessary to open the methodological hood of the notices deposited by the institutes with the Polls Commission. How can one publish a chart with bars totaling 100% of expressed opinions months, or even years, before an election? Through a triple reduction mechanism of the sample.

Filter of absolute certainty

The first mechanism is the filter of absolute certainty: institutes assess the probability of voting, often on a scale from 1 to 10. To establish voting intentions, institutes restrict their sample to only those who respond 10/10, that is, those who say they are "certain to vote." This is a drastic filter. Asserting absolute certainty of participation months before an election is a constraining posture, which mechanically excludes a large part of the electorate. The magnitude of this first cut was highlighted before the National Assembly inquiry commission on March 26, 2025.

Illusion of urgency

The second mechanism is the illusion of urgency: to those surviving this first filter, the famous hypothetical forcing ("If the election were to take place next Sunday...") is imposed, accompanied by an arbitrary list. This fictional framework requires an artificial projection into immediacy, forcing an unnatural stance. A methodologically sincere approach would instead respect the reality of this distant deadline: it would ask voters for whom they would be most likely to vote on the actual election date, rather than forcing them to simulate an imaginary vote.

Invisibilization of undecideds

The third mechanism is the invisibilization of undecideds: despite the urgency of this hypothetical situation, a significant proportion of these voters (between 15% and 25%) who are "certain to vote" refuse to choose a name, declaring they will not make a statement (NSP) or want to vote blank. What does the polling mechanism do? If the institutes defend themselves against ignoring these profiles by asserting they keep them to analyze the general evolution of electorates, all of them exclude them in fact during media publication.

The role of "exploitability"

The technical argument often used to justify this mechanical cleaning is that of "exploitability." For commentators seeking rankings, the blur is perceived as a disturbing noise, a residue that needs to be purged. The hypothetical and the practice of "adjustment" (modifications of the raw data of a survey to make the sample more "representative" of the general population) serve as compressors of uncertainty. However, when a large part of the voters do not know how to answer or are waiting for the campaign, it is not a failure of measurement: it is the real state of the system.

Sociological critique of polling

Sociologist Pierre Bourdieu already formulated this criticism in 1973, in his founding text The public opinion does not exist (1973): the institutes, by assuming that everyone has an opinion and that all opinions are equal, produce a statistical artifact where it would instead be necessary to restore the power relations and the actual degrees of investment of the respondents. The temporal hypothetical described here is nothing more than an updated and technically refined version of the same mechanism: it no longer just makes as if all opinions are equal, it makes as if the absence of opinion itself does not exist.

Impact on political dynamics

By eliminating this area of blur to create smooth scores, the poll produces an opinion artifact. This presentation has direct effects on the political game: it prematurely freezes the landscape, dictates "strategic voting" according to fictitious power relations and transforms the campaign - supposed to be a space of persuasion - into a simple verification of a self-sustaining forecast.

Proposed reform of polling questions

A requirement for rigor must apply from the design of the questioning protocol. The questionnaires must be changed. The question currently favored by polling institutes, "If the first round of the presidential election were to take place next Sunday, for which of the following candidates would you be most likely to vote?" should no longer be used. It should be replaced by a question rooted in reality, a scientifically and democratically sincere measure, which allows for uncertainty and deliberation.

Respect for the voter's right to decide

Respect for universal suffrage begins with respect for its temporality. A poll should not force the voter to mimic the gesture of a fictitious vote to satisfy the continuous rhythm of news and produce premature rankings. Recognizing the right to doubt and to observe does not weaken the value of statistical information: it gives it back its rightful place and scientific and democratic dignity.