The Reconquest

I am writing from Brazil, on the eve of the vote. I came to see with my own eyes what the numbers can only count, and the first thing I learned is that here the election doesn't show on the billboards. It shows on the street corners.

In Rio I took Avenida Brasil north. Around Parada de Lucas begins the territory the gang christened the Israel Complex. The Star of David they hung up high is gone: the police took it down in March 2025. The name stayed. And on the walls you still read Bible verses where, in another decade, you read the initials of the Comando.

What changed in the favelas is not the poverty. It is the map of faith. Where there used to be one Catholic chapel per neighborhood, now there is a church on every street: a metal roll-up door, plastic chairs, a keyboard, a long name painted by hand. Assembly of God. God Is Love. Foursquare Gospel. They open when the parish is closed and they stay when the state has already left.

In São Paulo I went at night to an Assembly of God service in the East Zone. Whole families arrived, the men in pressed shirts, the women with Bibles in hand, the children up front. From the pulpit nobody said Lula or Flávio. There was no need. They spoke of family, of the son who left with the gang and came back, of the addiction left behind. At the end, those who wanted to pray for Brazil came forward. Almost the whole church came forward.

I walked out into the street with an idea no poll had given me. In these neighborhoods the church is the only institution that arrives before the narco and stays after the state. Tomorrow that network votes. And it votes together.

On August 10, with Brazil tied within a point, I wrote that Brazil was not what was really at stake. What was at stake was the last foundation of the project born in a São Paulo hotel in 1990. And I said the result would hit Mexico before next year's election did.

At that moment the Nexus poll for BTG Pactual had Lula at 46 and Flávio Bolsonaro at 45. The big firms still had Lula ahead. The talk in São Paulo, in Brasília and in the newsrooms of Mexico City was something else: the name was burned, the father was in prison, the son lacked the stature, the Banco Master scandal was going to sink him.

On October 1 I went further and put numbers on it. Flávio wins the runoff with somewhere between 52 and 55 percent. And I warned of what almost nobody wanted to hear: there was a two-point error hiding in the polls, and two points of voting intention are four points of margin.

On Saturday, October 3, Datafolha published its last reading before the first round. Lula 45, Flávio 42, in valid votes. The next day, one hundred twenty-five million Brazilians voted.

Flávio, 47.03. Lula, 45.16. Two million two hundred twenty-four thousand nine hundred sixty-five votes apart, with one hundred percent of polling sections counted by the Superior Electoral Court.

The country's most respected poll missed the margin by almost five points. In the same direction as always.

And it was not just the presidency. Flávio's Liberal Party elected 121 deputies, the largest single-party caucus since the 1988 Constitution. It won 19 of the 54 Senate seats in play, a new record for the party, and it will reach 2027 with 28 senators. Even Randolfe Rodrigues, the Lula government's leader in Congress, lost his seat in Amapá.

What tomorrow decides is not whether the right wins Brazil. In Congress, it already has. What tomorrow decides is who holds the key to the Planalto Palace, and with it the power to use that Congress.

Tonight Brazil goes to sleep tied. Tomorrow the polls open at eight and close at five in the afternoon, Brasília time, and before dinner everything is known. Electronic voting doesn't let anyone go to bed in doubt.

The polls from the final week say the same thing in different voices. Datafolha, fielded October 20 to 22, has Flávio at 47 and Lula at 46: in valid votes, fifty-fifty. AtlasIntel for Bloomberg gives Flávio 51.1 against 48.9. Quaest has them dead even. Indexa, 50.6 against 49.4. Alfa, published yesterday, 52 against 48.

But there is a detail in the Datafolha series that nobody is talking about. On October 8 it had Flávio four points ahead in valid votes. On the 15th, two ahead. On the 22nd, zero. Lula closed in two weeks running.

That is the real suspense. Not the tie. The slope.

And here is the question that separates the analyst from the fan: who do I believe, Datafolha's slope or Datafolha's error? It is the same firm that three weeks ago handed Lula a three-point lead that did not exist. If the same bias repeats tomorrow, today's fifty-fifty is a 52.4 for Flávio. Right inside the range I signed.

There is another way to reach the same place, and it is the one I like best because it does not depend on any poll. It is arithmetic. Flávio came out of the first round with 47.03. To win, he needs a little under three points. Ronaldo Caiado, Romeu Zema and Augusto Cury added up to 5.34, and all three declared their support, Cury with the caveat that it was not an unconditional endorsement. Renan Santos, with 2.24, backed no one, but on the night of the 4th he said he saw Flávio's victory as practically settled. Lula has to find almost five points on ground where he has no allies left.

Move the dials yourself. Put each defeated candidate's voters where you think they will go tomorrow. You will see that for Lula to win, you have to assume two hard things at once: that the right that did not vote for Flávio in the first round stays home, and that the polls, for the first time in this election, underestimated Lula and not his rival.

To understand why Washington looks at Brazil the way it does, you have to lift your eyes from the electoral map and look at the whole map of the next 12 months that nobody has explained to you, and at how Mexico enters the equation of Marco Rubio's quietest shift.