Observations on election polls

Image: Plato Terentev


Jair Bolsonaro only confirmed the polled voting intentions. Lula, no, mainly because of the abstention of his voters

Many Lula voters were frustrated that the presidential election was not defined in the first round. While the main research institutes gave him something around 50% of the so-called valid votes (not considering abstentions), Lula had 48,43% of the valid votes (excluding blanks, nulls and abstentions).

By the same criteria, Jair Bolsonaro's vote surprised some in the first round, creating the erroneous perception (as we will see) of a wave. With 43,2% of the valid votes, he was well above the 36% that the institutes gave him. The difference between the two fell from 14 points in the polls to 5 points in the election. Soon, many analysts began to claim that the polls were right in voting for Lula and wrong for Jair Bolsonaro. We will see that the opposite happened, and Jair Bolsonaro only confirmed the researched voting intentions. Lula, no, mainly because of the abstention of his voters.

Taking advantage of the wrong interpretation of the “error” of research institutes, Jair Bolsonaro says he is surfing an irresistible wave, while Arthur Lyra and Ricardo Barros already promise a CPI and a bill to criminalize the error of electoral “predictions”. The objective is to promote the Bolsonarist “Data-Povo” and have one more reason to blame a “system” conspiracy for an eventual electoral defeat, stimulating movements to turn the tables.

The mistake made by most analysts is in comparing “valid votes” between polls and the election, as opposed to comparing researched voting intentions and effective voting, as criticized by the political scientist Antonio Lavareda. Comparing the proportion of “valid votes” between polls and the electoral result is a wrong method because the universe of voters is not the same. The universe of polls is the totality of voters without excluding abstentions.

If someone refuses to take the survey, it is not counted as an abstention. We do not know if more Bolsonarist voters abstained from answering than other voters, but there were cases in which Bolsonarists tried to force the interview. As for the abstentions in the election, they cannot be known beforehand and, in elections, they normally exceed multiple times the percentage of undecided voters in the polls.

The proportion of valid votes released by the Superior Electoral Court considers abstentions, that is, it only calculates valid votes as a proportion of those who actually voted (excluding blanks and nulls). The difference is huge: in stimulated surveys, the undecided were around 1%. Electoral abstention reached 21% of the total electorate.

Therefore, excluding blanks and nulls, the “valid votes” in the polls are calculated as a proportion of the total electorate (156,5 million voters), but the TSE discloses the valid votes as a proportion of a total that is 21% smaller (123,7 .XNUMX million voters). As the denominators of the ratios are different, it makes no sense to look for the error (or success) of the polls by comparing the proportion of valid votes in the polls and in the election.

Strictly speaking, polls only seek to estimate voting intentions, not valid votes. Even if the intention is carried out, it is the variation in abstention that will determine the proportion of valid votes, but polls never estimate abstentions. Therefore, to look for the error (or success) of the polls, the right thing to do is to compare voting intentions and effective voting, in both cases having the totality of the electorate as a universe. Recalculating the effective votes for the candidates as a proportion of the totality of the electorate, the surveyed voting intentions were right in voting for Bolsonaro and wrong in voting for Lula.

Source: IPESPE Survey 30 Sep. and TSE.

The proportion of 48,4% of valid votes for Lula released by the TSE drops to 36,6% considering the totality of the electorate, and Bolsonaro's proportion decreases from 43,2% to 32,64%. Thus, polls of voting intentions for Bolsonaro (33%) were correct in the vote, but they were wrong in the intention to vote for Lula (46% against 36,6%), exactly the opposite of what it seemed when making the wrong comparison.

Source: IPESPE Survey 30 Sep. and TSE.

What explains the discrepancy? To answer this, the methodological differences between the research institutes, whether in terms of sample weighting or consultation method, can be disregarded, since in all surveys it is Bolsonaro's intention that approaches the effective one and it is that of Lula and the others. candidates that fall short of the estimate. In the following table, the searched intent is in the blue columns.

Two hypotheses are common to explain the discrepancy: (1) vote transfer to Bolsonaro at the last minute, due to the very knowledge of the polls or the wave effect of message campaigns (with wide disinformation) not publicly detected; (2) unequal impact of abstentions on voting effectiveness among candidates. Most analysts make the wrong comparison between valid votes and defend the first hypothesis, while Antonio Lavareda defends the second hypothesis, making the correct comparison between researched intentions and voting effectiveness.

My hypothesis is that both things occurred, and it is necessary to estimate their weight. In favor of the first hypothesis, the survey on the second round of PoderData released on October 06 points out, for example, that 92% of Simone Tebet voters intend to vote for Lula, indicating that those who had Bolsonaro as their second option probably anticipated their choice. in the first round.

A transfer between Lula and Jair Bolsonaro, however, is less likely, given the enormous consolidation of Lula's intentions seen in the first round polls. As the consolidation remains in the polls published on the second round (see IPEC and Genial/Quaest), it is unlikely that it was harmed by a possible disinformation campaign on the eve of the first round. If Jair Bolsonaro did not win many votes from Lula, logically his voters' abstention was much lower than Lula's. It is illogical to assume a higher abstention for Jair Bolsonaro, as it would not be compensated by the transfer of votes for Lula to the point of preserving the similar value of polled (32,6%) and effective (33%) voting intentions.

Therefore, it is likely that voter abstention in Jair Bolsonaro was much lower than in Lula, and was compensated by the transfer of votes, mainly from the other candidates. A first piece of evidence is that the Genial/Quaest poll (October 06) with voters who declared that they did not vote in the first round, or voted blank and null, indicates a 45% preference for Lula, 28% for Jair Bolsonaro.

If Lula did not lose many votes to Bolsonaro, the divergence between the surveyed voting intentions (46%) and the effectiveness at the ballot box (36,6%) in Lula should be explained mainly by the abstention of his voters. This always happens, as Lula is preferred among low-income voters with less travel conditions and/or motivation to vote. In the 2006 election, the difference was practically the same as in 2022, around 9 pp (from 46% to 37,07%). As Lavareda recalls on Twitter, “of the nearly seven million illiterate voters, around 51% did not vote in the penultimate election. And 44% of Lula's voters even have completed elementary school”.

On the one hand, if they believe in the hypothesis of transferring votes to Jair Bolsonaro in the first round, Lula's voters can be more relaxed, as Bolsonaro would have already received votes that he would only receive in the second round. On the other hand, if Lula's voter abstention has been significant, and, worse, if it widens in the second round as is usual, the election will be much closer than the polls released after the first round indicate.

Considering “valid votes”, the difference between Lula and Jair Bolsonaro is 10 pp in the IPEC poll (Oct 05) and 8 pp in the Genial/Quaest poll (Oct 06). If Lula loses 9 points for abstention, but Bolsonaro confirms his intention to vote again, the election would be tied today. Jair Bolsonaro's campaign knows that the abstention of poor voters favors it, because on the eve of the first round, it filed a request with the TSE to ban free transport where there was on election day. Probably its most extremist militants will frighten voters with threats of violence, so that they will adhere to the “stay at home and we’ll see the election later” policy.

It is regrettable that the freedom to vote is not accompanied by the same ability to vote for all. In the medium term, generalizing free public transport, educating on the importance of political participation and on the methodology of electoral polls (to avoid the “already won”) and curbing political violence can mitigate the problem. In the very short term, only an enormous campaign to stimulate the vote of those who prefer Lula can ensure his victory.[1]

*Pedro Paulo Zahluth Bastos He is a professor at the Institute of Economics at Unicamp..



[1] The tables were requested from Antônio Lavareda and can be consulted next to his Twitter profile (https://twitter.com/LavaredaAntonio/status/1577072345583607808).


The site the earth is round exists thanks to our readers and supporters. Help us keep this idea going.
Click here and find how

See this link for all articles


  • About artificial ignoranceEugenio Bucci 15/06/2024 By EUGÊNIO BUCCI: Today, ignorance is not an uninhabited house, devoid of ideas, but a building full of disjointed nonsense, a goo of heavy density that occupies every space
  • Franz Kafka, libertarian spiritFranz Kafka, libertarian spirit 13/06/2024 By MICHAEL LÖWY: Notes on the occasion of the centenary of the death of the Czech writer
  • The society of dead historyclassroom similar to the one in usp history 16/06/2024 By ANTONIO SIMPLICIO DE ALMEIDA NETO: The subject of history was inserted into a generic area called Applied Human and Social Sciences and, finally, disappeared into the curricular drain
  • Strengthen PROIFESclassroom 54mf 15/06/2024 By GIL VICENTE REIS DE FIGUEIREDO: The attempt to cancel PROIFES and, at the same time, turn a blind eye to the errors of ANDES management is a disservice to the construction of a new representation scenario
  • Hélio Pellegrino, 100 years oldHelio Pellegrino 14/06/2024 By FERNANDA CANAVÊZ & FERNANDA PACHECO-FERREIRA: In the vast elaboration of the psychoanalyst and writer, there is still an aspect little explored: the class struggle in psychoanalysis
  • Volodymyr Zelensky's trapstar wars 15/06/2024 By HUGO DIONÍSIO: Whether Zelensky gets his glass full – the US entry into the war – or his glass half full – Europe’s entry into the war – either solution is devastating for our lives
  • Introduction to “Capital” by Karl Marxred triangular culture 02/06/2024 By ELEUTÉRIO FS PRADO: Commentary on the book by Michael Heinrich
  • PEC-65: independence or patrimonialism in the Central Bank?Campos Neto Trojan Horse 17/06/2024 By PEDRO PAULO ZAHLUTH BASTOS: What Roberto Campos Neto proposes is the constitutional amendment of free lunch for the future elite of the Central Bank
  • Letter to the presidentSquid 59mk,g 18/06/2024 By FRANCISCO ALVES, JOÃO DOS REIS SILVA JÚNIOR & VALDEMAR SGUISSARDI: “We completely agree with Your Excellency. when he states and reaffirms that 'Education is an investment, not an expense'”
  • The strike at federal Universities and Institutescorridor glazing 01/06/2024 By ROBERTO LEHER: The government disconnects from its effective social base by removing those who fought against Jair Bolsonaro from the political table