AtlasIntel/Bloomberg’s latest survey gives Flávio Bolsonaro a nominal edge in a simulated runoff, but not a statistically meaningful one. The poll places him at 47.7%, against 47.6% for President Luiz Inácio Lula da Silva — a 0.1-point gap within its one-point margin of error.¹

The result reverses the numerical order from the institute’s previous round, when Lula held a 47.7% to 47.4% advantage. Yet both surveys describe the same central reality: a technical tie. In the first-round scenario, Lula leads 45.3% to 42.2%, but neither candidate reaches the threshold needed to win outright. AtlasIntel/Bloomberg surveyed 5,005 people online from September 23 to 28.¹

Eurasia Group reaches a different conclusion by looking beyond the snapshot of voting intention. Its electoral models show a 50-50 race, but the consultancy assigns Lula a 55% chance of winning, citing the advantages available to an incumbent and what it sees as remaining vulnerabilities in Flávio Bolsonaro’s candidacy.² “When we are in an election that really looks 50-50, we give a bias to the incumbent,” Eurasia’s Christopher Garman said, pointing to a president’s ability to make announcements and shape the closing campaign agenda.²

Garman also said Bolsonaro “has not been fully tested,” arguing that a second round would expose him to broader scrutiny. At the same time, he acknowledged that models emphasizing public security and corruption lean toward Bolsonaro, while those based on government approval favor Lula; intention-vote models, he said, indicate balance.²

A reposted Bloomberg report adds another potential complication for Lula: polling signs of weakening support in Brazil’s Northeast, a region he has long dominated. In such a close contest, even limited erosion there could matter.Paulo Figueiredo (8) (pfigueiredo08)³ The competing assessments therefore differ less on whether the race is close than on whether incumbent power or shifting voter sentiment will matter more at the finish.