A previous version of this article incorrectly stated that no Brazilian presidential candidate had won a majority in the first-round vote in three decades. Fernando Henrique Cardoso won in the first round in 1994 and 1998. The article has been corrected.
Lula did win a plurality of votes, 48.4 percent, but he did not win a majority needed to secure a first-round victory. Bolsonaro, who in some polls was predicted to win a dismal 34 percent of the vote, ended up with 43.2 percent.
To many inside and outside Brazil, Bolsonaro’s surprise showing is about more than just one election: It’s evidence that the far right is undervalued by polls globally, echoing claims in other parts of the world.
“Polls are broken. They are undercounting right-wing support. And it’s vital this be fixed to maintain credibility,” the Brazil-based journalist Glenn Greenwald, a firm critic of Bolsonaro, tweeted on Monday.
“THE SILENT MAJORITY IS BACK!!!” former president Donald Trump wrote on the right-wing social network Truth Social on Sunday evening. He later wrote that Bolsonaro had beaten “inaccurate early Fake News Media polls.”
Besting the work of professional pollsters has long been a badge of honor for the former U.S. president. In 2016, before he was elected, Trump dubbed himself “Mr. Brexit” — an apparent reference to not only the incendiary politics surrounding the British referendum to leave the European Union but also the widespread idea that polls had missed the outcome of that vote.
Trump did indeed beat the pollsters in 2016 — and again in 2020. He lost the latter election but it still prompted something of a reckoning in the polling industry. One industry panel later said the surveys ahead of the 2020 presidential election were the most inaccurate in 40 years.
But with his post on Sunday about a “silent majority,” Trump was referring to a common theory among U.S. conservatives that right-wing ideas are actually more popular than they appear to be in polls.
In Britain, pollsters have also spoken of the “shy Tory” factor that suggests right-wing voters are less likely to admit their preferences to pollsters. Other theories have suggested that right-wing voters may be harder to pick up in mainstream polls, as they are more likely to be rural and less likely to use the internet, among other factors.
The full picture, however, is more complicated. Brexit may have been a political earthquake, but for some, it was a widely expected one. For weeks ahead of the vote, polls showed that support for leaving the E.U. was gaining among British voters.
More recently, electoral wins for the far right in Sweden and Italy have largely tracked with pre-election polls. And in France, the far-right leader Marine Le Pen underperformed her polls.
Even in Brazil, Lula, the left-wing candidate, actually overperformed his polling average by 2 percent. (Although Bolsonaro outperformed his by 8 percent.)
For practical reasons, polls are conducted with small sample sizes that represent only a slice of the entire population — with complicated equations on how to use this survey data to represent the national mood accurately. If those equations are off, so are the final results.
In 2018, a study published by the social and natural sciences journal, Nature Human Behaviour, looked at more than 30,000 national polls from 351 elections in 45 countries between 1942 and 2017. The study found that there did not appear to be any systematic decline in accuracy over that time period — and that there was “no evidence to support the claims of a crisis in the accuracy of polling.”
But Christopher Wlezien, a professor at the University of Texas at Austin and co-author of the study, said in an interview that he had not analyzed more recent data to see if the conclusion still stood up.
In 2015, Israeli polling firms failed to see that Prime Minister Benjamin Netanyahu was on a path to reelection (notably, the polling errors were so significant that even exit polls, conducted after the vote, were off).
Brian Winter, editor in chief of Americas Quarterly, noted on Twitter on Monday that there had been a noteworthy number of polling missteps recently, with major polls off in the United States, Argentina and Chile.
For Brazilians, polling has now become a political issue in itself. CNN Brasil reported Monday that Bolsonaro’s allies were now seeking a “broad investigation” into the work of polling companies ahead of the election to see if there is criminal liability.
If so, it would be just the latest round of attacks by Bolsonaro on Brazil’s election process. He has already claimed without evidence that the country’s electronic voting system is compromised. In the event of an electoral loss, Bolsonaro is widely expected to refuse to accept the results. It would be similar to what Trump did after his own 2020 electoral defeat — but this time it might work.
But concern about the polls won’t be limited to Bolsonaro and his supporters.
Winter tweeted that he no longer planned to share or discuss Brazilian polls ahead of the next election. “It’s clear the models are broken, respondents not being truthful, or some other problem,” Winter wrote. Other political analysts say that the polling companies need to take action before the controversy consumes them.
“It is very serious because the institutes are under attack and they are unfounded attacks. Because there are a lot of serious people in the institutes trying to do their best, but they were wrong,” Pablo Ortellado, a professor of public policy at the University of São Paulo, said in an interview with BBC News Brasil.
Brazilian polling firms will soon have another chance to prove themselves. Ahead of the second and final round of the election in four weeks, there will be more polls and more scrutiny. But companies will try to learn from their mistakes and tweak their methods to ensure higher accuracy.
It could be something easy to fix. But some polling inaccuracies have stumped top experts. Many, including Wlezien, heavily scrutinized the polling misses during the 2016 U.S. presidential election. In response, firms updated their methodologies and weighting — only for the polls to be off again in 2020.
“I don’t understand how it could hurt them,” Wlezien said of the idea that far-right parties overperformed their polls. “And the sense is that the people running these polls don’t have a strong interest in getting it wrong — in any direction.”