What explains the difference between the polls and election results in Brazil?

Analysts have pointed to respondents’ declaration of false information or “embarrassed voting” as a major factor in the difference between the polls and the results

October 04, 2022 by Brasil de Fato

The various voter intention polls for the Brazilian presidential election indicated different results than those presented at the polls this Sunday, October 2. Luiz Inácio Lula da Silva (Lula) of the Workers’ Party (PT) won 48.4% of the votes and Jair Bolsonaro of the Liberal Party (PL), won 43.2%. Simone Tebet of the Democratic Brazil Movement (MDB) had 4.2% and Ciro Gomes of the Democratic Labor Party (PDT), 3%. The other candidates together did not even reach 2%. However, Datafolha research released only hours before the presidential race predicted that Lula reaching 50% of the vote, indicating the possibility of a victory in the first round.

One of the factors that may explain the difference is the fact that a significant portion of the voters decide their vote on election day. According to a Datafolha survey conducted shortly before the 2018 presidential election, 12% of voters decided their vote on election day. These choices are left out of the polls.

But in this year’s elections, researchers say that another factor explains the difference. João Feres Júnior, a political scientist at the Institute of Social and Political Studies at the State University of Rio de Janeiro (UERJ) and coordinator of the Manchetômetro (a website that monitors press coverage for economics and politics), explains,“The difference was only in Bolsonaro’s voting intention. In Lula’s voting intention, the surveys got it right within and close to the margin of error.” 

“I think that the only possible answer, although it is a hypothesis, is that Bolsonaro’s voters are averse to polls, that is, they avoid answering polls or, when they do, they report false information.” The problem, says Feres Júnior, was “capturing the preference for Bolsonaro.”

To explain this difference, some analysts have been using the theory of the “embarrassed vote.” The thesis of the German political scientist Noelle-Neumann, in the book Spiral of Silence, is the following: the perception that there is an advantage for a certain candidate leads voters to actually vote for that candidate, without this vote being revealed in the polls.  

According to Márcio Moretto, coordinator of the Digital Political Debate Monitor at the University of São Paulo (USP), “the cases studied by the German researcher were votes in which, although they were very tight, there was a very clear trend in the perception of the electorate about who would win the election. In these cases, the vote went to the ‘favorite; well above the margin of error of the polls.”

“Notice, the polls got it wrong not because people lied or omitted to answer them, but because one of the camps was embarrassed to publicly defend their vote and that had an impact in the last hour.”

But by this thesis, the result obtained by former President Lula could be higher than it appeared on the polls’ radars. The reality, however, is that there would have been a “minimal effort to take people to the streets at times when it might make sense” on the part of the PT campaign, according to Moretto, and an “overly cautious posture of Lula in the debates,” which was put aside only in the last debate, on TV Globo. “The bet was on a strategy of ‘winning by standing still,'” says the researcher.

The strategy, however, “left room for Bolsonaro to occupy the streets and networks with the rhetoric that he was the real favorite.”

“Inspired by Neumann’s work, Bolsonaro’s entire campaign on the networks was to discredit the polling institutes and convince people to trust what they see on the streets and networks. The strategy worked. Bolsonaro gave a message to his base at the beginning of the year: ‘I don’t believe in polls’. His base didn’t answer any more polls and his name shrank in the polls,” says Moretto.

Ciro Gomes’ performance

Some political scientists also point to the migration of votes from Gomes to Jair Bolsonaro already in the first round, due to Gomes’ attacks on Lula and the Workers’ Party, in an attempt to win the votes of Bolsonaristas.  

“On September 28, we showed that Lula had 51%, and Bolsonaro 36%. Tebet 5%, and Ciro, 7%. On the 1st [of October], we released another poll showing Lula falling, with 49%, and Bolsonaro rising with 38%. That is, the last minute trend was already close,” explains the director of Quaest Consultoria, Felipe Nunes, in an interview with Universo Online (UOL).

“Lula ended up with 48%, within the margin of error, but Bolsonaro appeared with 43%. This means that he grew five points. Where does this vote come from? There were approximately three points from Ciro that went away, not to Lula, but to Bolsonaro. The posture that Ciro adopted in the final stretch of the campaign was a determining factor in the kind of impression he made to the voter,” said Nunes.

Mayra Goulart, professor of Political Science at the Federal University of Rio de Janeiro (UFRJ) and the Graduate Program in Social Sciences (PPGCS) at the Federal Rural University of Rio de Janeiro (UFRRJ), shares this opinion.

“Bolsonaro, as a far-right candidate, tends to have a final stretch of the campaign that involves fake news, aggressiveness towards his opponent. This makes the undecided votes on the right end up focusing on him. In this sense, the vote of Ciro Gomes went to Bolsonaro,” she said in an interview for Jornal Brasil Atual.

But according to Júnior, there is no proof of this. “It’s pure guesswork, because in the same polls, when they asked Ciro voters who they would vote for in a second round between Lula and Bolsonaro, the majority went for Lula, both Ciro and Simone Tebet. So it is not true that the deflation that there was from Ciro went to Bolsonaro. I don’t believe this,” he says.

The poll results

The results diverge from the Brasmarket Institute survey, released on Friday September 30 before the election. In the spontaneous survey, when the names of the candidates are not presented to the interviewees, Bolsonaro had 44.3% of voting intentions against 27.6% for Lula. Gomes had 3.8% and Tebet, 3.2%.

They also diverge from the survey released by the Veritá Institute in the last week before the election. In the spontaneous survey, Bolsonaro appears with 47.3%; Lula with 44.7%, Gomes Gomes with 3.4% and Tebet with 3.1%.

A survey by the Brazil Balance Institute, released on September 28, showed Bolsonaro with 46%, against Lula’s 41%. Gomes and Tebet had 5% and 4%, respectively, also in a spontaneous survey.

The three polls were widely publicized by the Bolsonaro networks for placing Bolsonaro ahead of former president Lula. But it wasn’t only these surveys that showed different rates than the ones that materialized in the polls.

In an Ipec (formerly Ibope) survey, released one day before the election, Lula had 51% of the valid votes and Bolsonaro had 37%. Gomes and Tebet were tied with 5%. On Datafolha’s poll, Lula had 50% of the valid votes, against 36% for Bolsonaro. Tebet had 6% and Gomes scored 5%.  

In Quaest’s survey, Lula had 49% of the vote, followed by Bolsonaro (38%), Gomes (6%) and Tebet (5%). Similarly, in a Paraná Pesquisas poll on the Friday before the election, Lula had 43.9%, Bolsonaro appeared with 37.3%, Tebet had 5.8% and Gomes scored 4.9%.

Bolsonaro’s takeoff

According to Mayra Goulart, the nationwide results, even if they had some difference from the surveys, are within the variation of the polls. “The statewide results erred quite a bit, but reduced samples have a greater chance of error, because they make a national prospection of something regional,” she says.

In São Paulo’s election for Senate, all polls were showing Márcio França (PSB), from former president Lula’s coalition, ahead of Marcos Pontes (PL), a Bolsonaro ally. On the Saturday before the election, the Bolsonaro candidate had 31% of voting intentions against 43% for França, according to the Ipec survey. By the end of the day, however, Marcos Pontes was elected senator with 49.91% of the vote. França came next, with a performance of 35.9%.

In the election for the São Paulo state government, former mayor Fernando Haddad (PT) led the voting intentions with 39%, while Tarcísio de Freitas (Republicans) appeared with 31%, according to Datafolha data also released on Saturday. With 100% of the polls counted by the Regional Electoral Court (TRE), however, Tarcísio had 42.3% of the vote and Haddad had 35.7%.

The uprooting of Bolsonaro’s voters, which was not captured by the polls, also reinforced the Bolsonaro agenda in the National Congress. President Jair Bolsonaro’s party, the PL, managed to elect the largest bench in the House of Representatives and the Senate. In total, 99 congressmen were elected, an increase of 23 congressmen in relation to the current legislature. In the Senate, eight senators were elected, bringing the total to 13.

This brings the total number of congressmen in the House to 13. With this, the PL managed to elect 112 congressmen in the National Congress.

This article was translated and adapted by an article written by Caroline Oliveira and originally published on Brasil de Fato.