Alexandre Dézé is a lecturer in political science at the University of Montpellier, researcher at the Center for Political and Social Studies of the same establishment and teacher at Sciences Po Paris. In his work, 10 Lessons on Political Polls (De Boeck Supérieur, 144 pages, 12.90 euros), he pleads for a drastic reduction in political opinion polls.
What criticisms do you make of political polls?
The main thing is the unprecedented importance given to them. Do we need nearly 600 surveys to cover a presidential election, as was the case in 2017? This overabundance is a problem. One can also ask the question of their role in the selection of candidates, all the more so in a situation where political parties are weakened. Never have there been so many polls when there are real weaknesses in their production, and opacity reigns.
Can these surveys, all the same, prove to be useful as tools for analyzing a political situation?
I’m not sure. At all stages of the process, we can identify problems that cast the greatest doubt on their reliability. The polls carried out six months before a presidential election, for example, are correct in one out of eight cases. What are they for ? They fuel a political debate regarding virtual power relations. We test undeclared candidates, we force respondents to complex intellectual exercises where they have to imagine several different scenarios. This also presupposes a minimum knowledge of the candidates and of politics, which not all respondents have. Also, there is a problem with online samples of self-recruited volunteers.
That is to say ?
We are far from the prerequisites of a classic sampling. A representative survey, that is to say without bias, is a random survey where people are drawn by lot. But, more broadly, everything is problematic in the production of these surveys: from the composition of the samples, to the formulation of the questions, through the adjustment of the raw results. The ideal would be to reduce the number of political polls and increase their quality. Because, when a survey is done well, it generates valuable knowledge.
If we do not focus solely on voting intentions, polls can also be used to deconstruct elements of the language of certain parties. For example, we see that the main concerns of the French are purchasing power, the Covid epidemic or health, ahead of security and immigration…
You have 52.1% of this article left to read. The following is for subscribers only.