Fábio Faria, Brazil’s communications minister, could not contain his irritation when fresh polling for the country’s presidential election last week showed incumbent Jair Bolsonaro still trailing by more than 10 percentage points.
“Enough of these absurd polls. The moment of truth is coming,” said Faria, who like many supporters of the rightwing president believes the nation’s handful of mainstream pollsters are biased in favour of race frontrunner, leftwing former leader Luiz Inácio Lula da Silva.
For months, an array of surveys have shown Lula ahead with about 45 per cent support of the electorate versus about 35 per cent for Bolsonaro. In recent days that gap has extended further, with some polls now putting the former union leader at 47 per cent, within the margin of error for a first-round victory.
If no single candidate wins more than 50 per cent in the first round on October 2, the race will go to a second at the end of the month.
Yet for Bolsonaro’s backers, these polls do not reflect the true desires of Latin America’s largest democracy. For them, the president’s ability to pack out large campaign events — such as a rally this month on Copacabana beach in Rio de Janeiro which attracted tens of thousands — is a clear sign of his primacy.
Bolsonaro himself last week said that he expected to win in the first round. If not, “something abnormal will have happened”, he said. The comments have again raised fears that if he loses, the former army captain could cry fraud and contest the results as former US president Donald Trump did after the 2020 US election.
“The population wants our government to continue. These polls are worthless,” he said.
Brazil’s handful of mainstream pollsters, including Datafolha, Ipec and Quaest, vehemently defend their methodologies and accuracy. They are regularly accused by Bolsonaro’s supporters of having failed to forecast his success in 2018, even though the results of that race were broadly in line with their models.
“Political supporters are like football fans. They have a key role during the 90 minutes of the game: to support, to believe and to encourage. The polls, as well as the final score, act as a bucket of cold water on this behaviour,” said Felipe Nunes, founder of Quaest.
“Fans of both Lula and Bolsonaro cheer and complain about the results of the polls, as football fans do with referees when a foul is called. But there isn’t any basis for believing that polls have been inaccurate. On the contrary, the serious institutes have described well the mood swings of public opinion.”
Some analysts have pointed out, however, that pandemic-related delays to the census mean that pollsters do not have a complete overview of the population, particularly in regards the weighting that should be given to different income groups. This could affect the results.
“The postponement of the 2020 census compromises the accuracy of household and electoral surveys, as neither the [official statistics agency] nor anyone in Brazil knows the reality, with precision, of the profile of the Brazilian population,” said José Eustáquio Diniz Alves, a demography researcher who worked for the Brazilian Institute of Geography and Statistics.
“For example, Lula has a higher proportion of voting intentions among the low-income population and Bolsonaro has a greater proportion of voting intentions among the higher-income strata. But what is the statistical weight of the different income strata? Nobody knows precisely.”
Those close to Bolsonaro also believe that he suffers in polls because respondents are reluctant to openly admit they will vote for him. The phenomenon is known as the voto envergonhado — the embarrassed vote.
Nunes from Quaest, however, denied that this played a role. He said that the company had conducted three separate studies using different methodologies to try to account for this phenomenon. In all three, they found that voters were embarrassed not about supporting Bolsonaro but rather Lula, given the ex-president’s association with a vast corruption scandal that rocked Brazil between 2014 and 2017.
Neale El-Dash, a statistician who runs a polls aggregator website, said that polling in Brazil was historically “quite accurate” and that Bolsonaro’s campaign against the pollsters was “more a strategy to try to get voters motivated. It’s not based on reality”.
Voting intention research was not without its challenges, he added. Phone interviews can’t reach the 10 per cent of Brazilians who don’t have access to a mobile or landline. Face-to-face interviews, meanwhile, can be difficult in gated apartment blocks or dangerous, particularly in the nation’s dense favelas, where organised crime is often present.
Bolsonaro’s supporters regularly cite Paraná Pesquisas, a small outfit, as the most reliable pollster. Research from the group has put Lula and Bolsonaro in a technical tie.
This week local newspaper Folha de São Paulo revealed that Paraná Pesquisas had received BRL2.7mn (US$500,000) from Bolsonaro’s Liberal party ahead of the election campaign. The pollster said it worked for many political parties, and that other companies had received similar funding.
Nara Pavão, a professor of political science at the Federal University of Pernambuco, said that a desire to be on the winning side often meant that voters tend to naturally align with candidates leading in the polls.
“The consolidation of Lula as a viable candidate will not make a Bolsonarista change his mind and vote for Lula. But it affects the undecideds a lot as well as the people who would vote for lesser known candidates,” she said.
“There is a study that refers to opinion polls as self-fulfilling prophecies, because, at the end of the day, they influence the intention of the vote and end up confirming the advantage of the lead candidate.”