Brazil's polls had Lula winning the first round. They were wrong. On October 4, Senator Flávio Bolsonaro, the 45-year-old son of the imprisoned former president, finished first with about 47 percent of the valid vote to Lula's 45, according to figures from the Superior Electoral Court reported by CNN. The two now meet in a runoff on October 25, and the numbers behind it are tighter and stranger than the headlines suggest.

The upset the polls missed

Flávio's first-round lead surprised almost everyone. CNN reported that surveys days earlier had forecast him finishing second, and Reuters described his showing as exceeding expectations. That is why our pre-election preview of the vote, which reflected the polling at the time, now reads as a snapshot of what forecasters believed rather than what happened.

The miss was not new. In 2022, the first-round gap between Lula and Jair Bolsonaro was closer than opinion polls had suggested, and Lula went on to win the runoff 50.90 percent to 49.10, according to the record of that election. Polls overstated Lula's first-round margin then, as they appear to have done this time. Whether they repeat that error on October 25 is the most important unknown in the race.

The arithmetic of the remaining voters

Take the final first-round figures: Flávio 47.03 percent, Lula 45.16 percent. That leaves 7.81 percent of valid votes that went to other candidates. If turnout stays the same, Lula needs about 4.84 of those 7.81 points to draw level, which is roughly 62 percent of them. Flávio needs only 38 percent.

The leftover voters are not a neutral pool. Reuters reported on October 5 that two right-wing rivals, Ronaldo Caiado (2.18 percent) and Romeu Zema (0.27 percent), endorsed Flávio. Brazilian press reports said the centre-right federation of the Progressistas and União Brasil also announced support for him on October 6, even though both parties sit in Lula's cabinet and governing base. The writer Augusto Cury (2.89 percent) and Renan Santos (2.24 percent), who together took about 5 percent, said they would back neither candidate.

So the real contest is narrower than it looks. Lula needs to win most of that roughly 5 percent that declined to choose a side, and also persuade people who did not vote the first time to turn out for him. This calculation assumes the same turnout and valid-vote rules as round one, and in practice voters move and abstain, but it shows how little room for error Lula has.

Faith and geography

Datafolha's data show the culture-war split. Among evangelical voters, Flávio leads Lula 47 percent to 27, while among Catholics Lula leads 45 to 33, CNN reported. Religion has become one of the sharpest fault lines in the race.

Geography matters just as much. CNN pointed to Minas Gerais, a state that has backed the winner of every presidential election since Brazil's first direct vote after the military dictatorship. It is large, politically mixed and rarely lopsided, which makes it the closest thing Brazil has to a bellwether. Southern states lean toward Bolsonaro, and Lula's strength in the northeast was slower to be counted in the first round, NBC News noted, a reminder that early results on election night can mislead.

What the latest poll says

A Datafolha poll released on October 15 showed Flávio with 47 percent and Lula with 45, Reuters reported, narrowed from a 49 to 45 Flávio lead a week earlier. Excluding blank votes (6 percent) and undecided voters (2 percent), Flávio leads 51 to 49. The survey interviewed 2,520 people on October 13 and 14.

A two-point lead among all respondents sits near the edge of a typical margin of error. It is a technical tie. The direction of travel favors Lula, and the history of first-round misses argues for caution about the direction of error. Both can be true, which is why neither campaign is treating any number as settled.

What to watch

The campaign has centered on crime, corruption and the shadow of Donald Trump, according to CNN, with Flávio presenting himself as an ally of the US president. Beyond the headline numbers, three things will decide the result: whether Caiado's and Zema's endorsements actually move voters, whether the roughly 5 percent who rejected both finalists stay home or pick one, and whether people who skipped the first round return.

The honest forecast, eight days out, is that Flávio has the stronger arithmetic and Lula has the stronger recent trend. The first round showed that Brazilian voters can surprise the polls, and the runoff may do it again.

---