Luiz Inácio Lula da Silva won Brazil’s 2026 election on Sunday, 25 October, securing 51.79 percent of valid votes against Flávio Bolsonaro’s 48.21 percent. The result contradicted most pre-election polls, which had projected a tighter race.
Discrepancy Between Polls and Results
Thirteen of the fourteen surveys released the night before the election showed Flávio Bolsonaro leading. Datafolha, Quaest, AtlasIntel, BTG/Nexus, CNT/MDA, and PoderData all placed the Liberal Party senator ahead. One Datafolha poll even gave Flávio a one-point lead, with a two-point margin of error.
When comparing the two candidates, the effective margin of error doubles to about four points. Only three surveys—Ipespe, Vox Brasil, and Futura—had their final results fall within this wider confidence interval. Vox and Futura came closest to the actual outcome. Ipespe matched Nexus and Palver’s result only because it declared a larger margin of error of 2.6 points. Eleven other surveys were outside the range. Datafolha, for example, estimated a two-point lead for Flávio, while the official tally showed Lula ahead by 3.58 points—a difference of 5.58 points, exceeding the applicable margin.
Conversely, most first-round polls had shown Lula ahead. Eight of the twelve institutes predicted a lead for him. Quaest, for instance, adjusted for turnout and estimated a one-point advantage for Lula, within a four-point margin. Yet that lead did not match the first-place finisher. Four institutes, Vox Brasil, Futura, Gerp, and Palver, put Flávio ahead, a position he confirmed with 47.03 percent of valid votes against Lula’s 45.16 percent. By this measure, seven of the twelve polls were outside the interval, and all seven favored Lula.
Read Also: Brazil’s Tax Reform Triggers Infrastructure Contract Overhaul
Standard practice in election cycles raises questions about why polling failed to capture the shifts observed at the polls. Polls aim to track voter behavior and trends, yet the data suggests they missed these movements. Calls for a Senate inquiry and a Federal Police investigation have already surfaced, recalling the controversies of 2022. Outside the institutes, no one knows the cause of the discrepancies. Without access to the data, hypotheses remain untestable.
Transparency and the Path Forward
In 2022, polling errors were similarly contentious. Only two of the 28 institutes that registered national polls with the Superior Electoral Court released microdata: Datafolha and Ipec. Both deposited their findings at the University of Campinas’ public opinion center, but even then, some questions and microdata were excluded. The remaining institutes kept their data private.
For 2026, there is a chance to do things differently. Of the 31 institutes that registered national polls, only 22 covered at least 20 states in any of their 191 surveys. While methodology details improved, four institutes mentioned controlling for the first-round vote in the runoff, though only two formally reported it to the TSE. Without microdata, it is impossible to determine whether differences stem from last-minute vote changes, sample composition, or weighting adjustments.
Greater transparency is urgently needed. Voluntary disclosure has not been sufficient. The second-round estimates rely on whether the sample includes all respondents, likely voters, or only the highly engaged. This definition appears only in voluntary reports, not in TSE filings. Nexus/BTG and Datafolha used turnout probability models. Nexus excluded the 21 percent of respondents least likely to vote and highlighted this result. Datafolha presented the total sample but listed high-propensity voters as supplementary information. Quaest adjusted for “attendance patterns” without revealing the base calculation. AtlasIntel claimed to weight by “past electoral behavior” but did not specify if it used the 2022 or 2026 first-round vote.
Read Also: Truck Crashes Demand Specialized Claims Handling, Experts Warn
Congress could mandate transparency through legislative reform. Resolutions like TSE Resolution 23.600/2019 or the 1997 Election Law could be updated to require public deposits of data within six months of the second round, similar to the CESOP repository. Media groups and sponsors play a key role in ensuring transparency.
Proposed Solutions for Poll Transparency
One approach to enhance transparency involves normative or legislative reforms. This would ensure open access to polling data for scrutiny and analysis.
Another avenue lies with poll financiers. Entities like Globo, Banco BTG Pactual, Banco Genial, 100% Cidades, and Empresa Folha could include contractual clauses requiring public deposition of polling data. Such measures would bolster the credibility of the polls they commission and disseminate, supporting greater accountability in the polling process.
