Two national surveys published on October 9 offer contrasting snapshots of Brazil’s presidential runoff, with different methodologies producing notably different results.
Brazil’s presidential runoff between Flávio Bolsonaro and incumbent President Luiz Inácio Lula da Silva is entering a crucial phase ahead of the October 25 vote. Two new national polls released on Friday, October 9, illustrate why the race cannot be understood through a single survey.
AtlasIntel/Bloomberg reports 52.8% for Flávio Bolsonaro and 47.2% for Lula among valid votes.
Vox Brasil, however, reports 44.2% for Lula and 42.7% for Bolsonaro among all respondents. When those figures are converted to a two-candidate valid-vote basis, they correspond to approximately 50.9% for Lula and 49.1% for Bolsonaro.
The findings differ not only in the size of the gap but also in which candidate is numerically ahead.
Brazil runoff polls: the October 9 comparison
| Poll | Flávio Bolsonaro | Lula |
|---|---|---|
| AtlasIntel / Bloomberg | 52.8% | 47.2% |
| Vox Brasil* | 49.1% | 50.9% |
Both rows show valid-vote shares. Vox figures were recalculated by Votendi from the institute’s reported total-vote intentions. They are not an independently published valid-vote estimate from Vox.
AtlasIntel/Bloomberg: Bolsonaro at 52.8%
The AtlasIntel/Bloomberg survey interviewed 5,026 voters between October 3 and October 8, using random digital recruitment.
The study reports Bolsonaro at 52.8% and Lula at 47.2% of valid votes, a difference of 5.6 percentage points.
Among all respondents, Bolsonaro records 51.1%, Lula 45.7%, and blank, invalid or undecided responses account for 3.2%.
The survey has a reported margin of error of ±1 percentage point at a 95% confidence level and is registered with Brazil’s Superior Electoral Court (TSE) under BR-03663/2026.
One important detail is that the fieldwork began before the October 4 first round and continued afterward. The published aggregate therefore spans two distinct stages of the election campaign.
Vox Brasil: Lula at 50.9% after valid-vote conversion
Vox Brasil interviewed 2,100 voters through face-to-face household interviews conducted between October 5 and October 7.
Its published results among all respondents were:
Lula: 44.2%
Flávio Bolsonaro: 42.7%
Blank, invalid or neither candidate: 3.4%
Undecided or no response: 9.7%
Excluding the non-candidate responses and dividing each candidate’s share by the combined 86.9% supporting either candidate produces the following comparison:
Lula: 50.9%
Flávio Bolsonaro: 49.1%
These are Votendi’s arithmetic conversions, rounded to one decimal place. They allow a like-for-like presentation of the two candidates’ shares but should not be confused with a separate polling model or forecast.
The Vox survey reports a margin of error of ±2.15 percentage points at a 95% confidence level and is registered with the TSE under BR-09623/2026.
Why do the polls differ?
The surveys use different approaches to collecting responses.
AtlasIntel relies on digitally recruited participants, while Vox Brasil conducts in-person household interviews. Their sample sizes also differ substantially, and their fieldwork periods are not identical.
Differences in sample composition, weighting, interview methods, question wording and timing can all affect polling estimates. The available information does not establish which factor accounts for the observed gap.
There is also a distinction between the numerical result and its statistical interpretation. Vox’s narrow 1.5-point difference in total-vote intentions is smaller than its reported margin of error.
Neither survey is an election result, and neither should be treated as a prediction of what will happen on October 25.
Datafolha provides another reference point
A third major national survey, released by Datafolha on October 8, reported Bolsonaro at 52% and Lula at 48% of valid votes.
Datafolha interviewed 2,520 voters in person and reported a margin of error of ±2 percentage points.
Taken together, the latest surveys demonstrate meaningful variation in estimates of the same runoff contest. AtlasIntel and Datafolha report higher figures for Bolsonaro, while the Vox numbers produce a narrow numerical advantage for Lula after conversion to valid votes.
These differences underline the value of examining individual polls, their methods and their fieldwork dates rather than relying on one headline percentage.
What comes next for Brazil’s election?
The second round takes place on October 25, 2026, following an October 4 first round in which Bolsonaro received 47.03% of valid votes and Lula received 45.16%.
New surveys will provide additional measurements as the campaign continues, but they must be evaluated with attention to timing, population, methodology and whether their figures represent all respondents or valid votes.
Votendi will continue tracking published polling data and separating the second-round contest from the archived first-round results.
Follow the Brazil 2026 election tracker for polling data, election context and updates.
Sources: AtlasIntel/Bloomberg reporting – UOL, Vox Brasil – CNN Brasil, Datafolha – Folha de S.Paulo.
Published survey figures and methodological information current to October 9, 2026.
