Why pollsters failed and what to watch for in Brazil’s presidential runoff

In a competitive runoff between President Bolsonaro and former President Lula, known unknowns like turnout, misinformation, and sudden shifts in public opinion might play a decisive role. This analyst gives a 50% chance to both candidates.

Marcelo Camargo/Agência Brasil

On 2 October 2022, left-wing former president Luiz Inácio Lula da Silva garnered 48.43% of the vote in the first round of the presidential election, while incumbent far-right president Jair Bolsonaro got 43.20%. Leading polling companies in Brazil failed to predict the tight result.

On average, pollsters estimated support for the Workers Party’s (PT) Lula correctly, but failed to capture many voters’ preference for President Bolsonaro. For example, Datafolha and Ipec, two of the more-established pollsters, used face-to-face interviews that had been considered the gold standard in the Brazilian market. Datafolha gave Lula a 14-point lead over Bolsonaro two days before the election. Ipec predicted an even larger 17-point lead. A far cry from the actual 5.23% difference.

Big polling companies’ failure to capture Bolsonaro’s support in the first round should lead analysts to debate polling methods rather than to deny polls are a valuable instrument for our work.

Why many pollsters failed to predict the result

Brazil’s polling discrepancy could be down to a pro-Bolsonaro “ashamed vote.” Andrei Roman, the CEO of pollster Atlas Intel, citing data from another surveyor, notes that men said they would vote for Bolsonaro more often when their interviewer was also a man rather than a woman. The negative implications of this finding on Datafolha and Ipec’s in-person polling methods are clear.

Roman’s Atlas Intel was the pollster that best predicted the first round. It forecast 48.3% for Lula and 41% for Bolsonaro. It did this through internet surveys rather than face-to-face interviews.

Another factor for Atlas Intel’s better forecast may have been the database it used to get a representative population sample for its polling surveys. Atlas Intel opted for the latest household survey (the PNAD Contínua, from 2021), while other companies worked with older samples, such as 2018 data from election authorities or adjusted data from the latest available national census from 2010.

Some of these older database samples appear to have over-represented poorer Brazilians who tend to vote for the left-wing PT and Lula. For example, Ipec used survey data in which 57% of respondents earned less than twice the minimum wage, while pollster Quaest only sampled 38% of citizens from this same income bracket. Quaest’s final forecast — an 8% lead for Lula — proved closer to the mark than Ipec’s 17%.

Religion, not just income, has also emerged as a reason why some pollsters underestimated the Bolsonaro vote. As the 2010 national census is outdated, polling firm Ipespe chose not to differentiate by religion. Ipespe used income groups in its samples. However, many poor evangelical christians tend to vote differently from poor, non-evangelicals, favoring more conservative, right-wing candidates, such as Bolsonaro.

Roman, of Atlas Intel, suggests that these different methodological caveats resulted in pollsters underestimating “Bolsonaro’s structural force” over the past two years, misleading many observers of Brazil’s politics.

What to watch for ahead of the runoff

Atlas Intel poll shows a toss-up between Lula and Bolsonaro in the battleground state of Minas Gerais, the second most populous in the country. Nationwide, on 24 October, Atlas Intel released a poll giving Lula a 6% lead. As of 25 October, the polling aggregators PollsterGraph, CNN Brasil/Locomotiva, and Polling Data also showed a close race, with Lula leading with a 1% to 3.4% margin. If these latter figures are correct, this is a tie in practical terms, since poll results must be read within a margin of error. In fact, Polling Data shows Bolsonaro 1% ahead of Lula when considering only polls conducted by telephone interviews.

This analyst gives a 50% chance of victory to both candidates. A tight election result is a realistic scenario. This takes into account the close results of Atlas Intel and polling aggregators, as well as three known unknowns: last-minute changes in public opinion, fluctuations in turnout, and misinformation.

Regarding turnout, fewer Brazilians have tended to vote in presidential runoffs than in the first round. In this year’s first-round election, abstention was highest among less-educated voters — a proxy for lower-income voters, who tend to support Lula. The question is what degree of abstention will occur on the runoff on 30 October.

Misinformation is another important factor and one which may have already left its mark on the race. In the 48-hour period before the first vote, news outlet O Antagonista reported that one of the jailed bosses of the São Paulo-based drug cartel Primeiro Comando da Capital (PCC) expressed support for Lula. Google Trends shows a spike in searches querying Lula’s link to the criminal organization between Friday night and midday Sunday, election day.

The Electoral Supreme Court (TSE) ordered the removal of O Antagonista’s story, considering it disinformation. That was after a complaint by the PT campaign arguing the media report was based on a screenshot of a police investigation report that could not be verified and did not display the Federal Police insignia.

It is impossible for pollsters to detect these sudden viral moments, and thus changes in opinion, right before the election. However, last-minute developments like these may partly explain why voters who originally planned to vote for third party candidates (Simone Tebet and Ciro Gomes) in the first-round switched their preference to Lula and — to an even greater extent — Bolsonaro, in the final days before the election.

Both sides frame this election as a defining moment. The Bolsonaro camp claims it’s a fight for liberty against socialism, and has spent heavily on ads on Google and Youtube in these final days. The left frames the Sunday vote as a final chance to avoid Brazil becoming the next chapter of “How Democracies Die”. In such a competitive race where last minute bursts of misinformation and tiny differences in turnout can swing the vote, both a win for Lula or Bolsonaro is just as likely.

Whether it’s Brazil or elsewhere in Latin America, Southern Pulse has the experience, network, and relationships to simplify this challenging region with honest, direct answers to your most complicated questions. Want to learn more? Let’s chat.




Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Southern Pulse

Southern Pulse provides strategic advisory services to help businesses operate successfully in Latin America.