The weekend science ticket with Mathilde Fontez, editor-in-chief of the new scientific magazine, Epsiloon, evokes today the subject of the polls that appeared at the beginning of the 20th century.
franceinfo: There is a lot of controversy about the reliability of polls, but beyond that, can they influence voters? Science has trouble answering …
Mathilde Fontez: Yes… sometimes science stumbles. And there, it is indeed the case. Yet the influence of polls on voters’ votes is an old question. It has been studied since the appearance of polls at the beginning of the 20th century. And it arises all the more today, when we are in a frenzy.
There have already been more than 60 polls for the 2022 presidential election in France. There were 560, in all, for the 2017 presidential election. And this is the case in all western democracies. A study has quantified the phenomenon in Great Britain: since 1945, the number of polls has multiplied by 50!
What effects could they have on the vote?
Four main effects have been identified and are being studied. There is the so-called “bandwagon effect”: the good old tendency that we have to conform to the dominant opinion, and therefore to vote for the leading candidate in the polls.
But there is also its opposite: out of sympathy for the less prominent small candidates, one might tend to vote for them. There is the “useful vote” effect: you abandon your favorite candidate to choose another who is close and who has a better chance of winning.
And there is the effect on participation: the polls could discourage us from going to vote… Except that on the four effects, the studies give very weak results, which in fact mean nothing. Or contradictory results.
How is that explained?
In fact, it is very difficult to study the polls. We can compare them, after the fact, to the results of a vote, but it’s a bit of a fish biting its tail. And researchers find it difficult to develop experiments to study the phenomenon in the laboratory, in a controlled manner: the role plays in which we put participants are always too artificial, too caricature.
And the worst part is that they too give very low results, whereas normally, this type of test is known to accentuate the effects.
These tests do not conclude …
They conclude, but very weakly. For example, a large study carried out over 10 years, in 7 countries, on nearly 23,000 people, published last month, shows a very small effect in favor of the favorite in the polls, but it remains in the error bar. So scientifically, it doesn’t show anything.
Finally, the most salient effect is the one on journalists – yes, we do feel a bit targeted. So not on the vote, but on their way of treating the news: the polls push to cover the elections in the style of horse racing, to say who wins or loses ground, not to mention the bottom.
And there is still a proven effect that affects everyone. It is the conviction that the polls do not affect us. We all tend to think that they only have an influence on others.