Four days before Brazil’s presidential runoff, the newest polls are telling two stories at once. Lula’s rejection rate has fallen sharply in the latest Quaest survey, and Flávio Bolsonaro’s numerical lead in that poll has narrowed. Yet other surveys released on Wednesday, October 21, still give Bolsonaro a modest advantage — and one shows that advantage widening.
The apparent contradiction matters more than any single headline percentage. Brazil votes on Sunday, October 25, and the final stretch is not producing a uniform swing toward either candidate. Instead, it is producing different signals from pollsters that use different samples, fieldwork periods and methods.
Quaest: a smaller lead — and a bigger change in rejection
In the Quaest survey released on October 21, Flávio Bolsonaro received 45% of total voting intentions against 44% for President Luiz Inácio Lula da Silva. A week earlier, the same institute had put Bolsonaro at 46% and Lula at 44%. The gap therefore narrowed from two percentage points to one, but that movement alone is too small to demonstrate a statistically meaningful shift.
The more striking change was in the rejection figures. According to Quaest’s latest rejection findings, the share saying they would not vote for Lula fell from 53% to 47%. Bolsonaro’s rejection declined from 54% to 52%. That leaves Lula with the lower rejection figure in this survey even though Bolsonaro remains one point ahead on the prompted vote question.
These are different measures: rejection is not a vote share, and an improvement in a candidate’s image does not necessarily translate into the same-sized increase in support. It can, however, suggest that the pool of voters willing to consider that candidate has changed — a potentially important development in a close runoff.
Quaest interviewed 2,154 voters face to face from October 17 to 20. Its stated margin of error is two percentage points, with a 95% confidence level; registration: BR-08535/2026. The numerical lead falls within that margin. View the poll in Votendi’s database.
Futura and Meio/Ideia: Bolsonaro remains ahead
Another survey released on October 21 by Futura Inteligência / 100% Cidades showed Bolsonaro at 51.1% and Lula at 48.9% of valid votes, which exclude blank, null and undecided responses. That is a 2.2-point numerical advantage. In its previous wave, released October 19, Futura had reported 50.4% to 49.6% — a 0.8-point advantage. In other words, the Futura series moved in the opposite direction from Quaest’s latest headline gap.
Futura conducted 2,000 telephone interviews on October 19–20. The stated margin of error is 2.2 percentage points; registration: BR-06614/2026. See the Futura poll.
The Meio/Ideia poll, also published Wednesday, reported the same 51.1%–48.9% division of valid votes. Its figures came from 1,500 telephone interviews conducted October 15–19, with a stated 2.5-point margin of error; registration: BR-08251/2026. See the Meio/Ideia poll.
These two sets of valid-vote figures look identical, but that does not make them one corroborated estimate or prove the margin is fixed. They are separate surveys with different fieldwork dates and error ranges; in both, the numerical difference remains compatible with a close contest.
Palver adds another close result — but watch the denominator
A Palver survey conducted online October 18–20 among 5,000 respondents put Bolsonaro at 50% and Lula at 48% of total voting intentions, within its stated 2.4-point margin of error. These percentages should not be placed alongside Futura’s and Meio/Ideia’s valid-vote shares as though they were measured on exactly the same basis.
The same caution applies to Quaest: its 45%–44% figures include other response categories. Votendi’s Quaest poll record displays approximately 50.6%–49.4% after recalculating shares from the rounded headline candidate figures. That is a derived approximation, not an independently published, unrounded valid-vote result from Quaest. Normalization can help comparisons, but it does not erase differences in methodology or rounding.
What the Votendi polling average says
At the time of this analysis, the Votendi Brazil runoff tracker displayed a weighted polling average of 51.3% for Flávio Bolsonaro and 48.7% for Lula across 32 published polls. The tracker weights survey results by factors including recency, sample size and methodology; the unreviewed new Palver wave was not included in that published-poll count at the time checked.
That means the current cross-poll snapshot still gives Bolsonaro the advantage, even as Quaest produces a more encouraging signal for Lula. A weighted average is a summary of the available polls, however — not a forecast of the result on Sunday. Any late movement, turnout differences or survey error could change the final outcome.
The final question is whether changing perceptions become votes
The clearest new analytical question is whether Lula’s six-point improvement in Quaest’s rejection measure becomes visible in future voting-intention surveys. For now, there is no consistent cross-pollster evidence of a decisive late break in his direction. Futura’s own sequence, in fact, has moved the other way since its previous release.
The remaining campaign days will test both possibilities: Lula narrowing the gap among reluctant voters, or Bolsonaro converting his recurring modest polling edge into an election-day victory. With technical ties appearing in the latest individual surveys, neither outcome should be treated as settled. The next releases — and ultimately the votes counted on October 25 — will matter more than declaring a winner from a one-point movement.
Methodology note: Polls are identified by their release dates and fieldwork periods. All differences are percentage points, not relative percent changes. Published total-vote intentions and valid-vote shares are distinct measures. Figures and tracker count reflect the data available when this analysis was prepared on October 21, 2026.
Related Votendi coverage: Why undecided voters still matter · The earlier cross-poll snapshot · Live Brazil runoff polling tracker.
