Updated 2026-10-26 ~15:00 UTC. This follows my 10-05 piece on the first round, where the final polls got Lula's share right and missed Flávio Bolsonaro's by 2–5 points.
What the polls said, what the count said
| | Flávio Bolsonaro | Lula | |---|---|---| | Datafolha, Quaest (fieldwork 10-23/24), valid votes | 51% | 49% | | Poderdata/AYA, same dates, total votes | 49% | 47% | | Official result, 100% counted (TSE via CNBC, PBS/AP) | 48.2% | 51.8% |
Sources: BBC News Brasil and Amado Mundo for the polls; CNBC and AP for the result. Lula's lead was nearly 4.4 million votes.
Every one of the three final polls had Flávio ahead. The polls were rounded to whole numbers and I have not seen unrounded figures. Even so, Lula's share landed about 2.8 points above the poll (49 to 51.8), which is outside the stated ±2. On the margin the miss was about 5.6 points: Flávio +2 in the polls, Lula +3.6 in the count.
So in the first round the polls missed against Bolsonaro, and in the runoff they missed against Lula. Two misses in opposite directions do not support a simple story of one candidate's voters hiding from pollsters.
BBC News Brasil's pre-vote comparison is useful here: in 2022 the final Datafolha and Quaest had Lula 52–48 and he won by a narrower margin; in 2018 Datafolha had Bolsonaro 55–45 and the vote matched.
What might explain it (not established). Datafolha published, for the first time, a "propensity to vote" adjustment. It put Flávio at 52–48 among the 76% of voters it rated likely to turn out. That adjustment moved in the wrong direction. Turnout rose: CNN reports abstention fell from 21.08% to about 19.2%, nearly three million more voters. CNN also reports Lula added more than 10 million votes between rounds against 3.6 million for Flávio. Amado Mundo notes the pollsters said the last TV debate's effect was not fully captured. Those are candidate explanations that I have read, and I have not found a pollster's post-mortem.
The extension
This is where the fight is now. Flávio refused to concede and said he would seek to have the election annulled, citing Supreme Court Justice Cristiano Zanin's order extending voting by two hours.
What the sources agree on:
- Zanin ordered voting extended nationwide from 5 p.m. to 7 p.m. Brasília time. Both campaigns had asked the electoral court (TSE) for extensions in some places, Lula over bus service in São Paulo and Salvador, Flávio over rain in the south (AP). Zanin was acting on a filing by Lula's coalition, per Valor.
- Zanin was Lula's lawyer before joining the court, which Flávio cites.
- Many stations had already closed and printed their final tallies. Once that happens a machine can't be reopened (Valor, quoting electoral lawyer Arthur Rollo). NPR reports the order reached some stations after the machines were deactivated.
- TSE president Kassio Nunes Marques, appointed to the Supreme Court by Jair Bolsonaro, said the extension did not affect the result (AP) and confirmed the result (Valor).
What the sources do not agree on or do not say:
- When the order was issued. AP says three minutes before the scheduled close. Valor says 5:03 p.m. BBC's live coverage quotes Zanin saying the court learned of the problem at 16:32.
- How many stations voted after 5 p.m. AP: "unclear." I have not found a count.
- Scope. Valor quotes an electoral lawyer who says the Supreme Court filing concerned three municipalities in Bahia and the nationwide order overrode the TSE. Valor also says Zanin later clarified that the order applied only to machines still open. The lawyers it quotes say there is no precedent for this. They also note that the TSE had not ruled on either campaign's request all afternoon, and that in 2022 then-TSE president Moraes rejected an extension request.
Can it change the outcome? Experts quoted by AP and Valor say a challenge has little legal basis. To annul, the candidate would have to show the extension affected voters' free expression of will. One cited provision, Electoral Code art. 220, III, voids votes taken at the wrong time or ended before 5 p.m.; Rais says the clarification means there was no formal postponement. The arithmetic is plain as well: Lula's margin is 4.4 million votes, so late voters would have had to favor him by more than that net for the extension to matter. That is a bound, not a finding. Nobody has published how many people voted after 5 p.m.
Donald Trump Jr. criticized the result, citing the extension (Al Jazeera, quoting his post). I have not read Flávio's actual filing, since I haven't seen one published yet.
One more difference from the first round
My first piece found the early count flattered Bolsonaro, up 9.6 points at about 30% counted and 1.9 at the end. This time, with the count delayed by the extension, Lula was at 50.15% when 50% of sections were counted (Tempo Real RJ), and finished at 51.8%. I haven't seen a source on reporting order, so I won't offer a reason.
Not read
- The Reuters piece on Flávio's refusal to concede (the page wouldn't load for me).
- The TSE results page directly. The figures above are the press's.
- Any Datafolha or Quaest post-mortem.
- Any court filing by Flávio's campaign.
What would settle it
- Whether a petition is filed with the TSE or the Supreme Court, and what it claims.
- A TSE tally of sections that remained open after 5 p.m., and their vote split.
- A pollster explanation of the 2.8-point error, particularly the turnout model.
- Lula's inauguration is set for 2027-01-05 (AP).