Brazilian President Luiz Inacio Lula da Silva leads every simulated first-round scenario for the October election included in a recent AtlasIntel/Bloomberg poll, but his advantage evaporates in a hypothetical second-round vote against Senator Flavio Bolsonaro, the survey showed on Wednesday.
The poll modeled several first-round matchups. In those scenarios, the leftist Lula garnered between 43% and 47% of voter support across five different configurations of opponents. By contrast, the right-wing challenger Flavio Bolsonaro was seen polling between 33% and 40% in four separate first-round simulations.
When respondents were asked to choose between Lula and Flavio in a potential run-off, the vote split essentially evenly: Lula at 46.2% and Flavio at 46.3%.
Those run-off figures represent a narrowing from a January poll in which Lula held a 49.2% to 44.9% edge over the senator. The January result itself had already declined from a 12-point lead recorded in December, according to the polling timeline presented.
Beyond the Lula-Flavio pairing, the poll indicates Lula would defeat six other potential opponents in head-to-head run-off matchups but would lose narrowly to Sao Paulo Governor Tarcisio de Freitas in a one-on-one contest.
Market and political context
AtlasIntel noted that markets have been sensitive to polling developments since December, when the endorsement of Flavio Bolsonaro by his imprisoned father, former President Jair Bolsonaro, coincided with a slide in Brazil's currency and equities. Investors had been expecting a different candidate to receive that backing, one seen as more market-friendly, and the endorsement changed market expectations.
The poll also underscores the electoral mechanics in Brazil: if no contender captures more than 50% of valid votes in the first round, the top two vote-getters proceed to a second-round vote - a step that has occurred in every presidential election since 2002.
Survey methodology
AtlasIntel conducted 4,986 interviews between February 19 and February 24. The poll carries a margin of error of plus or minus 1 percentage point.
Additional details on cross-tabulations, turnout assumptions, or regional breakdowns were not provided in the poll summary available with the results.