Brazil Is Up 20% Since July. What Now?

This piece is an update to my July note, check it out below. It is free!

Disclaimer: This article is for informational and educational purposes only. It does not constitute financial, legal, or geopolitical investment advice. The views expressed are based on public intelligence and analytical frameworks as of the publication date and are subject to change without notice. Readers should conduct their own research or consult professional advisors before making any investment decisions.

In the final weeks of Brazil’s 2026 campaign, the two men running for president spoke different languages.

Lula spoke in data. Unemployment at a record low. Inflation back inside the target band. GDP growing. He is 80 years old and running for a fourth term, and his message to voters was to be patient a little longer.

Flávio Bolsonaro held up a grocery receipt and compared the prices to his father’s term.

On the night of October 4, every poll was wrong. They had Lula ahead by three to five points. The count came in Flávio 47.0%, Lula 45.2%. The next morning the Ibovespa rose 7.7%, its largest single-day gain since the COVID crash of March 2020. The real gained 4%. EWZ, the Brazil ETF most U.S. investors hold, rose 12.5%.

This piece answers three questions. Why the polls missed. What the market is actually buying. And after a move like that, whether to take profits or add.

To answer any of them, one number has to come first.

13.75%

Brazil’s policy rate, the Selic, sits at 13.75%.

For a dollar investor, that number is the engine of the trade. The Fed hiked in September to a range of 3.75% to 4%. Brazil pays 13.75%. The gap is roughly ten points, and as long as the real does not collapse, an EWZ holder collects it passively every year. That was the first pillar of the position when it was built in July.

For a household in the outskirts of São Paulo, the same number means something else entirely.

The grocery receipt was real. Since 2020, food-at-home prices in Brazil are up 61%. Headline inflation is up 43%. Real wages are up 9%. Eight in ten Brazilians carry some form of debt. Debt service now consumes 26.6% of household income, the highest on record. Among low-income borrowers, the default rate has climbed from 4.1% to 5.5%.

Lula’s macro numbers are all accurate. Unemployment really is at a record low. Inflation really is inside the target. But in September’s Quaest survey, nearly half of respondents said the economy had gotten worse over the past year, against 19% who said better. 57% said their income had not kept pace with prices. His approval stood at 44% against 50% disapproval.

Aggregates describe the economy. Pain is experienced one household at a time. A family handing a quarter of its paycheck to creditors every month does not accept statistics as an answer.

The interest rate that makes this trade work is the interest rate that pushed voters toward the opposition.

That is not a moral judgment. It is the starting point for everything that follows. The position is long a political shift that is being driven by other people’s pain (debt payments).

Who the polls could not reach

That the polls were wrong is unremarkable. What is remarkable is that they were wrong in the same direction twice.

In 2022, Datafolha’s final poll underestimated Jair Bolsonaro by 7.2 points while missing Lula by 1.6. In 2026, it underestimated Flávio by 5.0 points while missing Lula by 0.2.

Random error is random. It lands on either side. Two consecutive misses of similar size in the same direction point to a population the sample systematically fails to reach.

Who are they?

The congregation. The 2022 census counted 47.4 million evangelical Christians, 26.9% of Brazilians aged ten and over, concentrated in the urban periphery. Between the 1950s and 1980s, millions migrated from the countryside to the edges of the cities, poor, isolated, often illiterate. They walked into the nearest place of worship. The churches gave them food baskets, literacy classes, sobriety groups, and a community where someone knew their name. One anthropologist’s description is exact: the church is an informal social safety net where the state is absent. So when a pastor signals a candidate, the congregation’s compliance is not manipulation. It is loyalty to the only institution that ever showed up. Phone sampling captures individual preference. It cannot capture a network that meets weekly, lends each other money, and watches each other’s children.

The indebted. Pollsters stratify by income, region, education, and religion. None stratify by what share of last month’s paycheck went to debt service. That variable is the strongest anti-incumbent predictor in this cycle.

The ones who will not say it aloud. Last October, Rio police launched an operation with 2,500 officers. Police counted 122 dead. The public defender’s office counted 132. It was the deadliest in Brazilian history. Subsequent polling found substantial public support for the operation’s severity. Picture someone telling a stranger on the phone that they approve. Most will not. But they will vote. The two former security chiefs who ran the operation were both elected to office this cycle. Shame is evidence of how strongly the preference is held.

One more finding cuts against intuition. Focus groups found evangelical voters unenthusiastic about Flávio. They see him as a diminished copy of his father. Many did not bother watching campaign broadcasts. In runoff simulations, they backed him 69 to 31.

Voting behavior has decoupled from enthusiasm. This bloc switched sides once, in 2018, and has not switched back. That is not mobilization. It is identity.

Polls measure temperature. Elections are decided by habit.

Act III: The president does not run Brazil

Now to what the market is buying. The common reading is that it is pricing a Flávio victory. That is a small part of it.

Brazilian politics has one defining feature: the president, acting alone, can do almost nothing.

Congress has two chambers, 513 deputies and 81 senators. Different acts require different vote counts. Ordinary legislation needs a majority of those present. Overriding a presidential veto needs 257. Amending the constitution needs 308. Impeachment needs 342.

The amendment threshold is the one that matters. Brazil’s 1988 constitution is extraordinarily detailed. Pensions, taxation, and fiscal rules are all written into it. In Brazil, “major reform” effectively means “constitutional amendment,” and an amendment does not require the president’s signature. Whoever controls 308 votes can rewrite the rules without him.

Where do those votes come from? This is the part foreign investors most often misread.