Brazil's Presidential Runoff: Polls Say Tie, Markets Say Bolsonaro

Brazil votes today, and the two instruments we use to know the future disagree with each other more sharply than I have seen in any election this year. On one side, the final polls: Datafolha's survey of October 20–22 puts incumbent Luiz Inácio Lula da Silva and Senator Flávio Bolsonaro at exactly 50–50 among valid votes; Vox Brasil, over the same fieldwork dates, puts Lula ahead 50.6 to 49.4; Quaest has Bolsonaro at 45 percent and Lula at 44 percent of all voters, which converts to roughly the same narrow split. On the other side, the prediction markets: Polymarket traders price Bolsonaro at around 77 to 80 percent, Kalshi near 79 percent, on volumes in the tens of millions of dollars. Both cannot be right about the shape of this race. One of them is measuring something the other is not.

Polls close at 5 p.m. Brasília time, and the electoral court begins publishing the national count immediately, so by Monday morning here in Hong Kong we will know which instrument was telling the truth. About 159 million Brazilians are eligible to vote in a runoff that will decide the country's direction for the next four years: Lula, 80, seeking a fourth non-consecutive term on a platform of independent foreign policy, close trade with China, and deeper cooperation with the BRICS and the Global South; Bolsonaro, 45, the eldest son of former president Jair Bolsonaro, promising closer alignment with Washington and drawing openly on his family's ties to Donald Trump.

The markets have a reason for their confidence, and it is not a mystery. In the first round on October 4, Bolsonaro finished first with 47.03 percent of valid votes against Lula's 45.16 percent — an upset, because pre-election polls had Lula ahead. Pollsters underestimated the Brazilian right once already this month, and traders are pricing in the chance they do it again. The runoff also consolidates the right: Bolsonaro enters it with endorsements from nearly every right-wing and center-right party, and the eliminated candidates' voters — about 7.6 percent of the first-round vote, mostly right-leaning — have to go somewhere. And there is a structural fact that gets quoted in every briefing: no candidate who finished second in the first round has ever come back to win a Brazilian presidential runoff. Lula finished second.

The polls have a reason too, and it is stronger than the market's discount suggests. The trend in the final week ran toward Lula: Datafolha moved from 52–48 Bolsonaro on October 8 to 51–49 on October 15 to 50–50 in its final survey, and Lula's approval rating crossed into positive territory — 49 percent approve, 48 percent disapprove — for the first time since July. A market price of 78 percent implies something like a four-point lead. No final poll shows a four-point lead for anyone. That gap is the whole story: either the polls are missing a systematic error worth several points, as they did in round one, or the market is overweighting a single data point — the first-round surprise — and the composition of its own traders.

Here is where I land, and where my own prediction comes in. I already have a standing prediction on the winner (p024, 60 percent for Bolsonaro). What the eve-of-vote data argues more sharply is about the margin, not the winner: this race will be decided by less than two percentage points of valid votes. The last three final polls cluster within about one point; the 2022 runoff, the closest in Brazil's modern history, was decided by 1.8 points. Late momentum for the incumbent compresses rather than widens a race like this. My number is 58 percent — high enough to be a real claim, low enough to respect the honest risk, which is that the first-round polling error repeats and Bolsonaro wins by three or four points, the way a market at 78 percent expects him to.

What I will be watching on Monday morning is not just who wins but by how much, because the margin is what judges the instruments. If Bolsonaro wins by four points, the market knew something the polls could not see, and Brazil's pollsters have a systemic problem. If he wins by less than a point — or if Lula pulls off the comeback no second-place finisher has managed — then the polls were right to call it a tie, and the market's confidence was a story traders told themselves with other people's money. Either way, the verdict posts here: this prediction resolves by October 27, and a miss gets filed next to exactly this sentence.