French elections and the limits of polls

The recent French elections' results surprised many, as they contradicted the pre-election polls. Alberto Bitonti, lecturer of political communication at USI, discussed this topic in an interview with Corriere del Ticino, examining the predictive mechanisms used in elections.

Bitonti highlights that polls reflect voting intentions at the time they are conducted but cannot accurately predict election results: "Demographic surveys capture voting sentiment at the time of the survey and not definitive predictions about elections."

Bitonti also adds that "many voters of parties considered socially unacceptable or unpopular tend not to publicly express their real preference, later expressing it in secret at the ballot box." Variations in turnout can also impact the results. In the French case, "the risk of having the Rassemblement National (RN) in government mobilised voters who generally abstain."

Bitonti points out that polls worked better in the first round than in the second due to different voting patterns under French electoral law. "The law mobilises voters in different voting modalities. In the first round, citizens express their preferences; in the second round, only the most voted candidates gain access, and voters choose strategically." This makes precise predictions impossible, even with well-constructed representative samples. In addition, polls must account for variables and margins of error. "Polls from reputable demographic institutes should explicitly state their methodological choices and must have a large number of respondents, extensive sample stratification, and representativeness." A similar example was recently seen in England, where the Labor Party, polling at 40 per cent, got just over 33 per cent. "In majoritarian systems, you have to think in terms of absolute votes and not in aggregate, as is often done instead."

The USI professor concludes by pointing out how an effective poll requires a representative sample, methodologically refined over time. Although 20th-century polls suffered from interviewer bias, the methodology has improved, allowing more reliable results to be produced. Representative groups are formed by "selecting as large and stratified a number of people as possible, distinguishing them by gender, social classes, educational level and other sociographic categories, which allows for real reliability of the survey."

The full interview, published in the Tuesday, 9 July edition of Corriere del Ticino, is available by downloading the attached PDF. (Italian only)