What is behind the adjusted result of the Quaest poll?

The new Quaest poll places Flávio Bolsonaro at 45% and Lula at 44%. That one-point difference is within the poll’s margin of error, so it indicates a technical tie rather than a clear lead. The important issue is not only stated preference, but whether supporters will actually vote.

Flávio is stronger among people who voted for candidates eliminated in the first round. Among this group, 43% now prefer Flávio and 37% prefer Lula. Lula leads among people who did not choose a candidate, with 40% for Lula and 36% for Flávio, but that group includes many previous absentees.

The poll’s final interpretation could therefore change after a likely-voter calculation. Felipe Nunes said Flávio’s advantage could grow if Lula fails to bring more of those voters to the ballot box. However, the article stresses that this projection had not yet been released and does not guarantee a larger gap.

The candidates’ additional second-round support comes from two different voter pools. Flávio Bolsonaro has an advantage among people who voted for candidates eliminated in the first round. Lula leads among people who did not choose anyone, including those who stayed home or cast blank or invalid ballots.

Among voters of defeated candidates, approximately nine million people, 43% intend to vote for Flávio and 37% for Lula. Another 19% plan to abstain, vote blank, or vote invalid. Among people who did not choose a candidate in the first round, 40% prefer Lula and 36% prefer Flávio. The remaining respondents include nonvoters, blank or invalid voters, and undecided people.

This creates a turnout-sensitive contest. Lula’s four-point advantage among previous non-choosers falls within the six-point margin of error for that group. If many of these voters do not participate again, Lula may lose more potential support than Flávio.

Two large groups could matter in the runoff. Approximately nine million people voted for candidates who failed to reach the second round. Another 39.3 million Brazilians did not choose a candidate in the first round. Together, these figures represent about 48.3 million people across two categories that may influence the final result.

The groups behave differently. Among voters of defeated candidates, 43% intend to vote for Flávio, 37% for Lula, and 19% for blank, invalid, or no vote. Among those who did not choose anyone, 40% prefer Lula, 36% prefer Flávio, 21% again plan not to vote or to cast blank or invalid ballots, and 3% remain undecided.

The 39.3 million non-choosers included 33.4 million people who did not attend the polls. The article presents these figures as separate voter groups, not as a single official total. Their impact depends on both preference and actual turnout on Sunday.

A likely-voter model adjusts a poll by considering which respondents have the greatest chance of voting. It matters because an election result depends on ballots cast, not simply on preferences expressed in an interview. The model tries to distinguish potential support from support that is likely to become a vote.

In this poll, Lula leads Flávio 40% to 36% among people who did not choose a candidate in the first round. But 33.4 million people in that broader group had stayed home previously. If those respondents say they support Lula but are less likely to return, their answers may receive less weight in a likely-voter calculation.

Felipe Nunes said the distance could become larger for Flávio if Lula cannot bring more voters to the polls. The Saturday projection had not yet been published, however. The article therefore presents the possible effect, not a confirmed change in the candidates’ standings.

Poll support and election-day support are not identical. A respondent can tell a pollster they prefer Lula but still fail to vote. If that happens more often among Lula’s supporters than among Flávio’s, the final ballots may show a different balance from the unadjusted survey.

The article identifies the previous non-choosers as especially important. Lula receives 40% of their declared support, compared with 36% for Flávio. Yet 33.4 million people in the wider group did not attend the first-round vote. Their stated preference helps Lula in the poll, but their past absence raises questions about whether that support will become votes.

Felipe Nunes said Flávio’s distance could grow if Lula does not bring more voters to the polls. This is a possible scenario, not a guaranteed outcome. The final effect depends on participation and on the likely-voter calculation that Quaest had not yet published.

A poll measures what respondents say they intend to do, while an election counts who actually votes. This difference matters most when a candidate’s support is concentrated among people with a history of abstention. Their preference can lift the candidate’s poll result without producing the same number of ballots.

Lula illustrates this risk. He leads Flávio 40% to 36% among people who did not choose anyone in the first round. But the broader group includes 33.4 million people who did not attend the polls. If many of Lula’s supporters in this pool remain absent, his four-point stated advantage may not appear in the vote count.

Flávio has an advantage among voters who supported defeated candidates, a group more directly connected to first-round participation. Nunes said the gap could move toward Flávio if Lula cannot mobilize additional voters. The article does not say that this outcome is certain.

A poll surveys a sample rather than every voter. Quaest interviewed 2,154 people, so its percentages estimate the views of the wider electorate. The poll reports a two-point general margin of error and 95% confidence, meaning small differences should not automatically be treated as real leads.

Subgroups are less precise when they contain fewer respondents. The article gives a six-point margin of error for the group that did not choose a candidate. That is why Lula’s four-point advantage in that group does not establish a definite lead. The 45% to 44% overall result is also a technical tie because the difference is smaller than the general margin.

Turnout assumptions add another layer. The unadjusted poll includes stated intentions, while a likely-voter model estimates who will actually vote. If that model correctly identifies participation patterns, it may improve relevance; if turnout differs from its assumptions, the final estimate may shift. Polls inform expectations but do not determine results.