University of São Paulo: Election polls are important, but need to be analyzed with caution

Opinion polls and voting intentions gain more and more prominence as the election approaches. This year alone, there are already more than 100 surveys and 50 institutes registered with the Superior Electoral Court (TSE). Brazil has never had as many consultations as in 2022.

These surveys involve a number of difficulties and require essential techniques to ensure that they are reliable. The data obtained needs to be treated with caution, but it provides important information for both candidates and voters.

The science of polling
One of the main objectives of a poll is to know the number of people who would vote for a certain candidate. The best way to obtain this data would be to interview all the voters in the country, but for logistical and cost reasons, this is not feasible.


A sample, that is, the group of people who will be consulted, needs to be defined. Professor Gilberto Paula, from the Institute of Mathematics and Statistics (IME) at USP, says that, in general, this sample is made up of around 2,000 people.

From these 2,000 interviews, the researchers seek to estimate the behavior of all voters. For this estimate to be more accurate, it is necessary that people with different profiles and opinions are consulted. “We cannot favor any group.”

One of the strategies adopted for this is geographic and socioeconomic stratification. “Voting is very geolocated. In some neighborhoods of the city you will have people voting much more for a candidate than in other neighborhoods”, explains Neale El-Dash, PhD in statistics from the Institute of Mathematics and Statistics at USP and founder of the website Polling Data , which aggregates and analyzes research published in Brazil.

In the country, these interviews are usually conducted in person or over the phone. According to El-Dash, there is no more adequate method, “but, in general, doing a telephone survey in Brazil is faster, cheaper and more accurate”. In addition to saving travel costs, phone calls can reach people who would not be included in face-to-face surveys.

Even with all these strategies, it is not possible to guarantee the total accuracy of the results, so there is the so-called margin of error. “It is a mistake that we make precisely because we are talking to some people and not to everyone”, explains the researcher.

This margin marks the maximum distance between the obtained number and the actual value. For example, if a poll has a margin of error of 2 points and points to a particular candidate with 25% of the voting intentions, it means that the real number is between 23% and 27%.


There is also the confidence level, which measures the chance of this interval happening. Professor Gilberto Paula says that, in general, this level is 95%, so it is possible that the result is not real. “This does not mean that the research is wrong, there is always a chance that the margin of error will not be met”, he recalls. To further increase this confidence, it would be necessary to interview more than twice as many people, which would significantly increase the costs of the survey.

In addition to the sample size, other aspects make it difficult to obtain accurate data. Those selected may refuse to participate or respond with misleading information. There are also cases of people who have not yet decided their votes, or even do not intend to vote. Therefore, the surveys tend to be more accurate as the election date approaches. Still, El-Dash points out that many people decide votes in recent days, which causes very large variations.

“Polls are important, they can help guide voters, candidates and parties, but they cannot be a determining factor in the decision to vote, there are many other variables that voters need to take into account.”

As the choice of votes is variable and depends on many factors, Gilberto Paula says that comparing different polls, at different times, is better than analyzing them in isolation. “Each candidate has a curve. From the moment you have this curve, you start studying what makes this candidate go up or down [in the polls].” Wagner Iglecias, sociologist and professor at USP’s School of Arts, Sciences and Humanities (EACH), also recalls that, for more complete analyses, it is important to know the institutes and funders responsible for the surveys.

The TSE adopts some measures to ensure the quality and reliability of these consultations. Institutes, in order to register their research in the agency, must inform the methods used, the number of interviewees and the period of execution. This information can be consulted on the Court’s website .

Given these requirements, Neale El-Dash positively evaluates the research conducted in Brazil. “We are well served by companies that are interested in getting it right,” she says, saying that, at the national level, the surveys are reliable. The researcher also comments that the variety of institutes and researches is important and helps to build more solid results, which can be compared.

Democracy, candidates and voters

Despite all these difficulties, polls are an important part of elections. Iglecias recalls that the democratic system is not restricted to the electoral aspect, but “they are part of this repertoire with rallies, motorcades and electoral schedules”, therefore, they are important for democracy.

With this data, candidates can increase their popularity, learn about voters’ preferences and plan their campaign strategies. In addition to voting intentions, polls usually indicate current topics, such as the economy, security, corruption or the environment. “This helps parties to define their speech, prioritize certain themes and choose certain opponents.”

The data are also important for the population to know the candidates for public office and to have a kind of thermometer of the electoral period. However, the professor highlights a negative aspect: some people use polls to define their vote.

Influence
“In the final stretch, polls can influence the vote in a logic that is very typical of a part of the Brazilian electorate, which is like this: ‘I will vote for whoever is ahead, who will win’. There are many people who choose this way because they imagine that voting for a candidate who is behind, even though the person likes the proposal or the profile of that candidate more, would be to lose the vote”, he explains. There is also a kind of herd effect, where people define the vote only to suit the behavior of the majority.

Iglecias points out that a very significant portion of voters study and decide their vote in advance. But the last-minute, poll-based decision can somehow positively influence the best-placed candidates and hurt the lowest-performing ones. “It’s very focused on the idea of ​​’I’m not going to lose my vote, there’s a certain very interesting candidate, but she’s got 2%, I’d love to vote for her, but I’m not going because she’s not going to win,’” he exemplifies.

The professor emphasizes the importance of political education, not to ideologize or build loyalty, but to “form citizens and citizens, people who know, for example, how a bill is made”. “This would be important because it would have consequences at the time of voting: ‘I will vote for candidate because he is the best for my city’, and not for emotional reasons or voting for [candidate] A because I don’t want [candidate] B to win . Polls are important, they can help guide voters, candidates and parties, but they cannot be a determining factor in the decision to vote, there are many other variables that voters need to take into account”, he concludes.