Brazil’s presidential runoff is effectively tied nationally in Datafolha’s October 22 poll, but the headline conceals a pronounced divide by household income. Among voters reporting monthly family income of up to two minimum wages, President Luiz Inácio Lula da Silva receives 58% of valid-vote preferences against 42% for Flávio Bolsonaro. Above that income threshold, the pattern reverses: Flávio records 58% to Lula’s 42% in each of the two higher-income groups reported by the institute.
The contrasting results, released ahead of the October 25, 2026 runoff, illustrate how an evenly divided national electorate can contain very different voting patterns. Crucially, the figures describe subgroups within one survey. They do not mean Lula has a 58% national voting share, nor do they constitute separate polls that can be added to a national polling average.
A 50–50 national split hides a 58–42 income divide
Datafolha’s nationwide survey gives both candidates 50% of valid votes when rounded to whole percentages. The institute separately reports 47% for Flávio and 46% for Lula among all interviewed voters, with 5% indicating a blank or null vote and 2% undecided. The difference is explained by the denominator: valid-vote estimates exclude blank, null and undecided responses.
The income breakdown paints a different picture in each category:
| Monthly family income | Lula (PT) | Flávio Bolsonaro (PL) | Difference |
|---|---|---|---|
| Up to 2 minimum wages | 58% | 42% | Lula +16 points |
| More than 2 to 5 minimum wages | 42% | 58% | Flávio +16 points |
| More than 5 minimum wages | 42% | 58% | Flávio +16 points |
| National electorate (valid votes, rounded) | 50% | 50% | Tied in rounded figures |
Source: Datafolha survey fielded October 20–22, 2026; income breakdown reported by Folha de S.Paulo and CNN Brasil. Figures are the valid-vote shares within each group, not the share of the whole Brazilian population supporting that candidate.
The 16-point gaps are comparisons within income categories. They should not be added together or interpreted as a swing of 32 points between survey waves. Nor is each category equally large: the total national result depends on how many voters are in each group and the survey’s weighting, as well as preferences in those groups.
What the numbers reveal—and what they do not
The survey documents a clear association between reported household income and presidential preference. It does not, by itself, establish why individual voters support either candidate. Economic policy, perceptions of government performance, religious affiliation, region, education and other factors may overlap within these groups. Isolating their independent effects would require additional analysis rather than a reading of three percentages.
There is also a crucial statistical distinction between the full national sample and any subgroup. Datafolha reports a national margin of sampling error of ±2 percentage points at a 95% confidence level. Smaller demographic samples have wider uncertainty. CNN Brasil reports a margin of approximately three points for the income comparisons in its summary; the methodological documentation should be consulted before using a single subgroup margin as applying uniformly to each category.
The current income results are a snapshot from October 20–22, not a verified measure of changes since the previous week. Comparisons over time must keep the relevant income bands, the valid-vote definition and the survey methodology consistent.
Gender and religion show additional differences
Income is only one axis of difference. The same Datafolha release puts Lula at 53% of valid votes among women against 47% for Flávio, while among men it puts Flávio at 54% against 46% for Lula. The six-point gap reported among women falls within the survey’s stated subgroup uncertainty.
The distinction is larger across religious categories: Flávio receives 73% among evangelical voters, with Lula on 27%, while among Catholic voters Lula records 59% and Flávio 41%. The poll also reports Lula at 68% in the Northeast, against 32% for Flávio. Those demographic and regional categories overlap and should not be treated as separate or independent voting blocs whose totals can simply be summed.
Other characteristics matter as well. For example, Datafolha reports 57% for Lula and 43% for Flávio among voters with only elementary education. This is a descriptive result, not a judgment about voters or an explanation of the causes of their choices.
Why the distinction matters two days before voting
Across Brazil, the presidential second round takes place on Sunday, October 25. A national 50–50 reading does not erase the contrast across income brackets. But the reverse is equally important: a candidate’s strong lead among one subgroup must never be repackaged as a lead across all Brazilian voters.
The latest findings are consistent with the broader picture of a close contest illustrated in Votendi’s October 23 runoff polling analysis. Separate national polls, using different methodologies and fieldwork periods, should be evaluated on their own terms rather than combined by selectively choosing demographic subsets.
For a broader timeline of candidate-level national polling, see Votendi’s Brazil presidential runoff tracker. The new demographic breakdown is context for the existing Datafolha survey—not an additional nationwide poll observation.
Sources and methodology
- Folha de S.Paulo, October 23, 2026 — subgroup breakdown by income, gender, age, religion, education and region, based on the Datafolha release.
- CNN Brasil, October 23, 2026 — detailed valid-vote subgroup figures and interpretation of denominators.
- Datafolha, October 23, 2026 — institute’s report on national voting intention, survey dates and methodological description.
Methodological note: Datafolha interviewed 2,520 eligible voters aged 16 and over face-to-face in 154 municipalities from October 20 to 22, 2026. The survey was commissioned by Folha de S.Paulo and TV Globo and registered with Brazil’s Superior Electoral Court (TSE) under BR-09709/2026. National maximum sampling error: ±2 percentage points; 95% confidence level. Subgroup estimates have their own sampling uncertainty. These are survey measurements, not official election results or a Votendi election forecast. Editorial cut-off: October 23, 2026.
