” No, but have you seen the latest polls? Le Pen came back up fucking phew! Admit it, you’ve heard this sentence between one and fifteen times in the last few days. Le Pen goes up, Zemmour is approaching the level of shame, Pécresse has hit rock bottom, but is still digging, Macron is plummeting, but remains in the lead, Arthaud has not passed the 2% mark: OK. All that is fine, but the “polls” are a bit vague for us, in fact. We see them everywhere, but we don’t understand everything! So hey, nice as we are, we’re going to try to answer a few questions, so as to see things more clearly in the grub, a few days before the elections.
1. What is a survey?
By definition, a survey is a statistical survey aimed at giving a quantitative indication, on a given date, of the opinions, wishes, attitudes or behaviors of a population by questioning a sample (some of the inhabitants). Opinion polls are not new! The first date back to the 1850s. In the USA, at the same time as the circulation of the opinion press swelled, the newspapers organized “straw polls”. They then ask residents to return a coupon with their intention to vote and complete their panel by going directly to the street.
2. What is it for in the context of presidential elections?
More than 150 years ago or today, the objective is the same (contrary to the method, which has evolved): to carry out electoral consultations before the official political deadline, in order to draw trends, to report on the majority opinion by questioning only part of the population concerned. We would tend to say that it is ultimately not much use to try to predict the results, but in reality, in an electoral context, the stakes are enormous for the candidates! Published in all the media, these surveys can influence the votes, by disqualifying a candidate presented as little supported, or by giving strength to another, displayed good first. According to Anne Levade, jurist at the University of Paris 1 Panthéon-Sorbonne, there is also an issue of transparency. She told Ouest France “ This data almost contributes to the transparency of the electoral process. A ban on polls would make the campaign a kind of black box where you don’t know what’s really going on “.
3. Who constitutes the famous “sample”?
In short: “Who is questioned”? Well, this sample is supposed to be as representative of the population as possible. Still within the framework of the presidential elections, the idea is to reconstitute a sort of “miniature France” with the same percentage of men, women, wealthy people or, conversely, in financial difficulty, etc., as those to France. For example, out of a sample of 1000 individuals, 137 people would have to be between 18 and 29 years old, since 13.7% of French people belong to this age group. You follow ? Each institute actually has a panel, from which it draws to create its samples. Note that to be interviewed, these citizens must volunteer and be available at the time of the survey (which, as a result, can inevitably create a bias). To know who to trust: it is estimated that in the context of voting intentions, you must question at least 1,000 people. Below 500, the results would be too imprecise.
4. Who makes them?
In France, there are nine main polling institutes, including Ifop, Ipsos and CSA. In these societies, and contrary to what one might think, opinion polls are not the main source of income! It is estimated that the latter contribute, on average, to the tune of 5% to 15% of the total turnover of these companies.
5. Where is it done?
No, in the context of the presidential elections, it is not an employee, pen in hand, who randomly questions you in the street! Opinion polls are done by telephone, and even online. Although this presence on the Internet has advantages, such as ease of access and the fact that respondents are no longer afraid of judgments (see last point), it also encourages new biases. For example, Luc Bronner, journalist for Le Monde, explained how he had been able to respond several times to the same survey by multiplying false identities, without ever being unmasked.
6. How many presidential polls are taken before an election?
No, because we seem to see new ones every day these days… Ah? Is that largely the case? During the presidential period, the polls multiply and the number of surveys carried out grow from five years to five years (or from seven years to seven years, if we go back further in time). The Polling Commission estimates the number of polls carried out in the context of the 2017 elections at 560, 409 in 2012, 293 in 2007 or even 193 in 2002. In 50 years, this number has multiplied by approximately 40!
7. Does the survey accurately reflect all respondent responses?
Well no. Not really. Some institutes discard responses from people who are unsure and certain of their voting intention. Clearly, some pollsters ask the respondent to estimate the certainty of his vote on a scale of 1 to 10. All those who do not answer 10 are not counted.
8. How are polls supervised during the election period?
They are very structured. The institutes are subject to strict obligations (listed in the following point). They must also be declared to the Survey Commission, and those, before any publication. Among the prohibitions: publish, distribute or comment on any opinion poll, regardless of the means, the day before and the day of the election. The law passed in 1977 fixed this ban at one week before the trip to the polls, but was reduced to these two days in 2002.
9. What information must appear on a survey?
The law modernizing the rules applicable to elections and the presidential election of April 25, 2016 extends the list of obligations set by the 1977 law. From now on, any first publication or broadcast of a poll must be accompanied by the name of the survey organization, the buyer or potential sponsor, the number of people questioned, the date of the interviews, the full text of all the questions asked, a statement specifying that any survey may have a margin of error, the margin of error itself, and finally, a statement indicating the right of everyone to consult the notice. Elements to check, therefore, before considering the results presented.
10. What is the “margin of error”?
As the name so aptly suggests, “the margin of error” corresponds to the level of uncertainty in a poll. This margin is normally between 1 and 3. Inevitably, the greater the margin of error, the less sounding and reliable, and vice versa. Moreover, this margin of error only considers those related to sampling. It must then be kept in mind that other elements (see next point) can also bias the results.
For example, if a candidate is credited with 20% of the voting intentions, with a margin of error of 2 and a level of confidence of 99%: it is necessary to understand “this candidate has a 99% chance of receiving between 18% and 22% of votes.
11. Bonus: is it reliable?
I’m assuming that if you’re wondering, it’s not, right? The polls are intended to be a representative maximum, but there are a multitude of biases that can distort the results, and therefore, the forecasts. For example, as mentioned in point 2, the question of respondents: the fact of volunteering and not of being chosen at random is a non-negligible variable. In the same way, the respondents are humans who can answer what they want: this is less the case since the process is done online, but all the same, it happens that some inhabitants lie about their intentions out of fear. to be judged. The list of reasons for which a poll is not 100% reliable is very long, but if we had to illustrate it with a single example: Le Pen, in the second round in 2002, yet announced loser by all the polls.
We will end with this sentence from Loïc Blondiaux, political professor at France info: “ In no case is the survey capable of anticipating a reality that will occur. Even one or two days before the deadline, there is no guarantee that a last minute movement will not come to contradict it. In sum, a poll is a kind of snapshot of intentions at a given moment, but is not a real prediction. It is clearer ? Yes ? Top. Now, all you have to do is choose your candidate… Yeah, that’s complicated too! If you are a bit of a crowd, know that we have summarized their programs for you in a few words.