Lula recovers ground in Minas and São Paulo, but Flávio Bolsonaro increases his advantage in Rio
Quick Look
- Published state surveys show Lula's recovery in Minas Gerais and São Paulo, with a technical draw in Minas and a reduction in the disadvantage in São Paulo, while in Rio de Janeiro Flávio Bolsonaro increases his advantage.
- The election is entering a phase of firmer choices, with less room for changing votes.
AI-generated summary
Why It Matters
State surveys released show the evolution of voting intentions for the 2026 presidential election in three states in the Brazilian Southeast: Minas Gerais, São Paulo and Rio de Janeiro.
Lula's (PT) recovery in the Southeast, signaled by the Quaest national survey of September 21, gains clearer contours in state surveys released this Wednesday (23).
Minas Gerais and São Paulo show positive fluctuations in the president in the second round and in the government's assessment. Rio de Janeiro heads in another direction, with Flávio Bolsonaro's (PL) advantage increasing. The Southeast is moving, but not as a block.
Minas offers the most important synthesis. At the beginning of the month, Flávio had numerically ahead in the second round, with 40% against Lula's 37%. Now, the president reaches 40%, while his opponent maintains the same percentage. If the state continues to function as a Brazilian thermometer, the temperature is that of an open and numerically tied election.
In the first round of Minas Gerais, both candidates advance. Lula goes from 31% to 35%; Flávio, from 27% to 30%. The sum of the two grows from 58% to 65%, while Augusto Cury (Avante) drops from 8% to 5% and Romeu Zema (Novo), former governor of the state, from 6% to 2%. Those undecided fall from 15% to 13%.
The movement suggests a concentration of preferences in the two main competitors, although the research does not allow us to say that it was exactly these candidates' voters who migrated.
SEE ALSO
The government's assessment reinforces the sign of recovery in Minas Gerais. In comparison with August, approval rises from 41% to 44%, and disapproval falls from 52% to 49%. The balance remains negative, but the distance between the two indicators drops from 11 to 5 points.
In São Paulo, Lula's improvement appears more clearly in the second round simulation. The president goes from 33% to 36%, while Flávio fluctuates from 41% to 42%. The senator's advantage decreases from eight to six points. Flávio remains ahead in the country's largest electoral college, but Lula reduces his disadvantage.
This improvement is not reproduced in the same way in the first round of São Paulo. Flávio advances from 31% to 34%; Lula, from 30% to 31%. Cury has 7%, Caiado (PSD), 4%, Renan Santos (Missão), 3%, and Zema, 1%.
In the government's assessment, in relation to August, Lula's approval went from 37% to 40%, and disapproval fell from 58% to 55%. There is relief for the president, but still within a largely unfavorable environment.
Lula (PT) and Flávio Bolsonaro (PL) — Photo: Ricardo Stuckert and Vittor Sales
Rio imposes a limit on the regional recovery reading. In the first round, Flávio goes from 34% to 37%, while Lula fluctuates from 32% to 30%. In the second, the senator goes from 43% to 44%, and the president, from 37% to 36%. The distance increases from six to eight points. The government's assessment also worsens numerically: compared to August, approval fell from 40% to 37%, and disapproval rose from 51% to 56%.
The variations need to be read in light of the margins of error: two points in São Paulo and three in Minas and Rio. The set of indicators allows us to identify regional signs and differences, without transforming each oscillation into a consolidated change.
There is, however, a movement common to the three states: choices are becoming firmer. Among voters who nominate a candidate, the proportion of those who consider their vote to be definitive goes from 71% to 81% in Minas, from 74% to 80% in São Paulo and from 77% to 82% in Rio.
The election is entering the phase where convincing someone to change becomes more difficult, and mobilizing those who have already chosen becomes more important.
It is in this scenario that three factors can make a difference:
The first is engagement: transforming preference into attendance. The winning campaign will probably be, as I have said since the beginning of this election, the one that gets voters out of their homes.
The second is the useful vote of those who today choose Cury, Caiado, Renan or Zema. These four candidates still account for 13% in Minas and 15% in São Paulo. Part of this electorate may make a strategic calculation in the final stretch, increasing polarization. But there is no automatic transfer or single destination for these votes.
The third is the Globo debate, next week, in prime time. At a time of greater attention and choices still being consolidated, confrontation can reinforce convictions, mobilize supporters and influence those who are still considering change.
Lula's recovery in Minas and São Paulo preserves the competitiveness of the dispute. Flávio's advantage in Rio shows the limits of this reaction. And Minas, with 40% for each side in the second round, leaves the message: the election remains open. What happens in the final stretch could define which side this balance will be broken.
What to Watch
AI outlook — possibilities, not facts
The Globo debate next week will influence voting intentions, especially among voters still consolidating their choice.
Likely · Within days
Part of the electorate of the minor candidates (Cury, Caiado, Renan, Zema) will cast a useful vote in the final stretch, increasing the polarization between Lula and Flávio Bolsonaro.
Possible · Within weeks
Open Questions
- What will be the effects of the Globo debate next week on voting intentions?
- How can the useful votes of minor candidates be distributed in the final stretch of the campaign?
- What will be the impact of electoral engagement on the comparison between preference and turnout?






