In the first round of Brazil’s closely watched elections this month, the polls were off the mark. They significantly underestimated the support for the far-right incumbent, President Jair Bolsonaro, and other conservative candidates across the country.
Many on the right were furious, criticising the pollsters as out of touch with the Brazilian electorate.
That response was expected. What happened next was not.
At the urging of Bolsonaro, some of Brazil’s leaders are now trying to make it a crime to incorrectly forecast an election.
Brazil’s House of Representatives has fast-tracked a bill that would criminalise publishing a poll that is later shown to fall outside its margin of error. The House, which is controlled by Bolsonaro’s allies, is expected to vote and pass the measure in the coming days.
The bill’s final shape and fate is unclear. House leaders have suggested they may soften the legislation, and its prospects in the Senate, where opponents of Bolsonaro are in the majority, appear far less certain.
Still, whatever the measure’s fate, the proposal and other efforts to investigate pollsters for their recent miscalculations are part of a broader narrative pushed by Bolsonaro and his allies, without evidence, that Brazil’s political establishment and the left are trying to rig the election against him.
As Brazil prepares to vote in a presidential runoff Oct. 30, the surveys continue to show Bolsonaro trailing his left-wing rival, Luiz Inácio Lula da Silva, a former president, although the race seems to be tightening.
For his part, Bolsonaro has taken to calling the polling firms “liars,” claimed that their mistakes swung up to 3 million votes to da Silva in the first round, and has advocated for the firms to face consequences. “Not for getting it wrong, OK? An error is one thing,” he said. “It’s for the crimes they committed.”
He has not said what crimes he believes were committed.
The Brazilian Association of Pollsters said in a statement that it was “outraged” at the attempts to criminalise surveys that turn out to be inaccurate.
“Starting this type of investigation during the runoff campaign period, when the polling companies are carrying out their work, demonstrates another clear attempt to impede scientific research,” the group said.
Polling firms added that their work was not to predict elections, but to provide a snapshot of voters’ intentions at the time a survey is conducted.
The bill in Congress is not the only effort to target pollsters. Following a request from Bolsonaro’s campaign, Brazil’s justice minister ordered the federal police to open an investigation into polling firms over their surveys before the first election round. And Brazil’s federal antitrust agency opened its own inquiry into some of the nation’s top polling institutions for possible collusion.
Alexandre de Moraes, a Supreme Court justice and Brazil’s elections chief, quickly ordered both of those investigations halted, saying that they lacked jurisdiction and that they appeared to be doing the president’s political bidding. In turn, Moraes ordered Brazil’s election agency to investigate whether Bolsonaro was trying to use his power over federal agencies inappropriately.
Moraes has emerged as the top check on Bolsonaro’s power over the past year, drawing criticism at times for measures that, according to experts in law and government, represent a repressive turn for Brazil’s top court.
Among other moves, he has jailed five people without a trial for posts on social media that he said attacked Brazil’s institutions. On Thursday, election officials further expanded his power by giving him unilateral authority to suspend social media platforms in Brazil that do not quickly comply with his orders to remove misinformation.
Moraes and Brazil’s Senate appear poised to protect polling firms from measures that target their surveys.
Yet repeated claims that pollsters are corrupt could further weaken their ability to provide the best possible gauge of public opinion. Some of Bolsonaro’s top advisers have urged his supporters to ignore survey takers in order to sabotage their results.
“Do not respond to any of them until the end of the election!!! That way, it’ll be certain from the start that any of their results are fraudulent,” Ciro Nogueira, Bolsonaro’s chief of staff, wrote on Twitter. “Was their absurd screw-up criminal? Only a deep investigation will tell.”
The top polling firms had forecast that Bolsonaro would receive roughly 36% of the vote in the first round. He received 43.2%, a seven-point gap that was outside virtually all polls’ margins of error.
Their performance was even worse in many down-ballot races. In Rio de Janeiro, the polls showed that the conservative candidate for governor was ahead by about 9 percentage points. Instead, he won by 31 points.
In Sao Paulo, some polls showed that a left-wing candidate for Senate was ahead of his opponent by 14 percentage points heading into the first election round. Instead, a right-wing candidate won by nearly that same margin — a swing of 28 percentage points from what the preelection polls had found.
The polling firms have blamed a variety of factors for their flawed forecasts, including outdated census data that hampered their ability to survey a statistically representative sample of voters. The firms said their polls were also undercut because a larger-than-expected wave of voters switched their ballots to Bolsonaro from third-party candidates at the last minute.
Some polling firms also said they believed that many conservative voters were unwilling to answer their surveys.
The share of older voters far exceeded expectations, potentially because of a government announcement this year that voting was a new way to establish proof of life and keep retirement benefits active. Polls on the eve of the election showed that older voters supported Bolsonaro over da Silva.
Brazil is far from the only country where polls struggle to give an accurate picture of the electorate, particularly the strength of conservative support.
In 2016, polls in the United States did not accurately forecast the support for Donald Trump, and the firms gave similar reasons for the miss, including that some right-wing voters were unwilling to answer surveys.
The credibility of Brazil’s polling firms was damaged after the election’s first round, and some journalists have become more hesitant to share surveys before Sunday’s runoff.
Ricardo Barros, a conservative congressman who is helping to push the bill to criminalise faulty polls, said the legislation would force polling companies to be more careful with their findings. Under the proposed law, only polls that err outside their margin of error would face liability.
“If you’re not sure of the outcome, then place a margin of error of 10%,” he said. “It loses credibility, but it doesn’t misinform voters. The problem is that today it’s always being presented as an absolute truth.”
Lawmakers in both the House and Senate have also gathered enough signatures to open congressional investigations into the polling firms, although the leader of the Senate is expected to move to block that chamber’s investigation.
Alexandre Cordeiro Macedo, the head of Brazil’s federal antitrust agency and an appointee of Bolsonaro, tried to go further than Barros in taking aim at polling firms.
Before Moraes intervened and stopped the inquiry, Cordeiro Macedo had accused top polling companies of collusion based on what he said was the statistical improbability that they all had underestimated Bolsonaro’s support by such a significant margin. He claimed that the scenario was about as likely as winning the lottery several times.
But Alexandre Patriota, a statistics professor at the University of Sao Paulo, disputed that, saying proving collusion based solely on that single measure would be nearly impossible.
“Even if all the institutes got it wrong in the same way, this is not an indication of a cartel,” he said. “To have a hint of malice, you need something more than numbers.”