Brazil presidential poll average for 2026: Runoff, Flávio Bolsonaro vs. Lula
Every poll for Brazil’s presidential runoff on October 25 in one place: a history by date, a chart for each pollster and the average weighted by the number of respondents. Each figure comes with the pollster, the release date, the margin of error, the registration number with the TSE, Brazil’s electoral court, and the source where it was published.
Loading the polls…
Runoff poll average
Valid votes, weighted by the number of respondents. Polls registered and released from the first round onward (October 4) are included; it is a RaisLife calculation.
Loading…
Runoff map: simulation based on the poll average
This is not a poll. No state-by-state interviews were conducted and nothing was registered with the TSE: this is a RaisLife simulation.
Each state is shaded for the candidate with the most votes in the scenario; the bigger the gap, the darker the shade. The map starts from each state’s first-round result and applies the national poll average. It is a scenario, not a forecast. After the election, the map is compared with the actual result.
History by release date
Each dot is a poll, placed on the day it was released, with the margin of error shown as a bar. The line is the running average of the polls released up to that day. Hover, tap or use the keyboard to see the details.
Each pollster, in its own series
The same date and percentage axes in every chart, so you can compare each pollster's trajectory with itself.
Upcoming polls
Official runoff polls announced or registered with the TSE that have not yet been released. They appear on the chart as hollow markers and are not included in the average.
Exit poll
On election day, registered exit polls appear here, separate from voting-intention polls.
After the election: who got it right and who got it wrong
The poll numbers compared with the official result from the TSE (Brazil's Superior Electoral Court), pollster by pollster, and the projected map next to the actual result.
All runoff polls
From newest to oldest.
How we read the polls
- Basis: valid votes, which exclude blank ballots, spoiled ballots and undecided voters. When the pollster published only the total (Quaest, Vox Brasil), its figure appears first and the valid-vote calculation follows, labeled as the work of whoever did it (RaisLife; Poder360 in the case of Vox Brasil).
- Date: the axis uses the release date. The fieldwork period is shown separately in the table. Polls without a confirmed date are left out of the charts and the average.
- Average: runoff polls registered with the TSE and released from October 4 onward (first-round day), each weighted by its number of respondents. What counts is the release date: AtlasIntel was in the field from October 3 to 8 and is included, with a caveat. Simulations released earlier are not included.
- Statistical tie: RaisLife's criterion, not a pollster's: a gap less than or equal to twice the poll's own margin of error (for the average, the average margin). It flags when the numerical lead is not conclusive.
- Limits: the average does not correct for differences in method between pollsters. It describes the polls and does not predict the election result, which is the one reported by the TSE.
- Sources: each poll cites at least one news outlet. Any divergence between sources is noted in the row's remarks.
How to read 2026 presidential election polls
A presidential poll measures the voting intentions of a sample of voters over a short period and has a margin of error. It is a snapshot of the moment it was taken, not a forecast of the outcome: the official result is the one from the Superior Electoral Court (TSE), counted on election night. That is why the page gathers all the runoff polls between Flávio Bolsonaro and Lula, not just the most recent one.
Valid votes, total votes and undecided voters
Some pollsters report the percentage of all respondents, including blank, spoiled and undecided responses; others report it as a share of valid votes. To compare different polls, we use valid votes, which are the ones that decide the election. On each poll’s line we show the figure the pollster published and, when RaisLife converted it to valid votes, we note that the calculation is ours.
Margin of error and statistical tie
The margin of error shows how much a poll’s result may vary, up or down. When the gap between candidates is small relative to the margins, it is not possible to say with confidence who is ahead: this is what is called a statistical tie. The page applies its own criterion, RaisLife’s: a gap equal to or less than twice the poll’s margin of error.
Which pollsters run presidential polls
Datafolha, Quaest, AtlasIntel, PoderData, Ipespe, Paraná Pesquisas, Nexus and other pollsters register polls with the TSE. Each uses its own method and sample, so the numbers differ; the “Each pollster, in its own series” section shows each one's trend against itself, and the average combines them all, weighted by the number of respondents.
Also follow
See all the 2026 elections, the live vote count on election day, the governors runoff simulator and Track your politician.
Frequently asked questions
When is the 2026 election runoff?
The runoff is on Sunday, October 25, 2026. Flávio Bolsonaro (Liberal Party, PL) and Lula (Workers’ Party, PT) are competing for the presidency of Brazil.
Where do the numbers on this page come from?
From polls registered with the Superior Electoral Court (TSE) and released by polling firms and news outlets. Each row lists the pollster, the registration number, the dates and the sources consulted. The figures are those released by the pollster; when it released only the total of respondents (as Quaest and Vox Brasil did), its figure appears first and the conversion to valid votes follows, labeled as a calculation by RaisLife (or by the outlet that made it, such as Poder360 in the case of Vox Brasil).
How is the weighted average calculated?
The average combines runoff polls registered with the TSE and released from first-round day onward (October 4), with two sources from different news outlets and fieldwork finished by the release. The rule applies to the release, not the fieldwork: a poll may have started fieldwork before the first round, like AtlasIntel (October 3 to 8), and is included anyway, with a caveat on its row. Each poll is weighted by its number of respondents, based on valid votes (excluding blank, spoiled and undecided). Polls released before the first round are not included. The average is a RaisLife calculation, not a registered poll.
What does a statistical tie mean?
When the gap between the two candidates fits within the combined margins of error, the poll cannot say who is ahead. The criterion used here, which is RaisLife's and not a pollster's, is a gap equal to or smaller than twice the poll's own margin of error (and, for the average, the average margin). By that criterion, several polls in which Flávio is numerically ahead are in a statistical tie, and the page reports how many.
Does the page predict who will win?
No. A poll shows voting intentions at the time it was conducted and has a margin of error. The official result is the TSE’s, on the night of October 25.
What is the most recent presidential poll?
The most recent one appears at the top of the “All runoff polls” table, with the pollster, release date, fieldwork period, margin of error and TSE registration. The page is updated with each new poll that passes the registration and source check.
Which pollster is the most reliable?
This page does not rank pollsters. What gives confidence is the poll's registration with the TSE, the disclosure of its method, sample and margin of error, and comparison across several pollsters, which is why the page shows each pollster's series and the average of all.
Do Datafolha, Quaest and AtlasIntel show the same result?
Not always. Each pollster sets its own sample, data-collection method and fieldwork dates, so the numbers can differ even within the margin of error. Compare each pollster’s series in the “Each pollster, in its own series” section, along with the average of all of them.
Do election polls have to be registered?
Yes. Brazilian election law requires a voting-intention poll to be registered with the TSE, the Superior Electoral Court, before it is released, along with the registration number, the client, the method and the margin of error. Only polls that are registered and reported by at least two sources are included here.
What about exit polls?
Exit polls may only be released after voting closes, on October 25. If any polling firm registers and releases one, it appears in the Exit poll section, with the release time, until the count begins. In the first round, polling firms did not conduct this kind of poll, according to Correio.