Theme 27 of 43
Opinion Polls, Public Perception & Political Behavior
In 2022, nine of ten pollsters underestimated Jair Bolsonaro on the eve of the first round — and in 2026 registered polls already cost R$ 206 million, paid mostly by banks and media groups.
Election polls compiled by Poder360 record voting intention, the pollster responsible, sample size, and margin of error for thousands of surveys conducted across the country, from municipal to presidential races. The National Health Survey and the continuous PNAD, both from IBGE, round out that picture with the population's perception of public services and well-being. And official TSE election results serve as a reality check: they let you verify whether the voting intention captured in polls actually held up at the ballot box. Cross-referencing this universe of public opinion with budget execution data, the pattern that emerges is of two spheres that barely talk to each other — what the population says it wants and what Congress decides to fund follow their own separate logics.
Pollsters missed Bolsonaro, not Lula
On the eve of the first round in 2022, the polls got Lula almost right; Jair Bolsonaro, not. Converted to valid votes, each pollster's last national survey, as compiled by Poder360, put Lula, in eight of ten cases, between 47% and 51% — he got 48.4% according to the Superior Electoral Court (TSE). For Bolsonaro, who got 43.2%, nine of ten pollsters came in below, with a median error of 4 points. Ipespe, Datafolha and Quaest missed by 6.9 to 8.1 points.
| Pollster | Date | Lula | Bolsonaro | Error on Bolsonaro (pp) |
|---|---|---|---|---|
| Ipespe | 09/30 | 48.9% | 35.1% | −8.1 |
| Datafolha | 10/01 | 50.5% | 35.8% | −7.4 |
| Quaest | 09/27 | 50.5% | 36.3% | −6.9 |
| Ideia Big Data | 09/28 | 48.9% | 38.5% | −4.7 |
| PoderData | 09/27 | 48.4% | 38.7% | −4.5 |
| MDA | 09/30 | 48.3% | 39.7% | −3.5 |
| Paraná Pesquisas | 09/29 | 47.2% | 40.1% | −3.1 |
| Futura | 09/28 | 43.6% | 40.5% | −2.7 |
| AtlasIntel | 09/28 | 50.8% | 41.1% | −2.1 |
| Brasmarket | 09/28 | 34.6% | 50.9% | +7.7 |
| Result (TSE) | 10/02 | 48.4% | 43.2% | — |
Valid votes here are each candidate's share of the sum of candidates listed in the poll, excluding blank, null and undecided — the same calculation pollsters publish. The error was not random: nearly all missed in the same direction, which points to something systematic — Bolsonaro voters less willing to respond, samples weighted on outdated figures in the absence of a recent Census, or last-minute vote migration from Ciro Gomes and Simone Tebet.
Lula reached the ballot box exactly where the polls said; Bolsonaro came in four points above the pollsters' median.
2026: more registered polls and more money
Every published election poll must first be registered with the Electoral Justice, with cost, client and methodology. For the 2026 general elections, the TSE registry lists 3,463 polls up to October 4, at a declared cost of R$ 206 million — 17% more polls and 44% more money, in current terms, than in 2022 (2,971 polls, R$ 142.8 million). Four in ten (1,438) are self-funded by the pollster itself.
| Pollster | Polls | Declared value (R$ million) |
|---|---|---|
| Veritá | 464 | 53.2 |
| Quaest | 214 | 39.9 |
| Datafolha | 104 | 18.2 |
| AtlasIntel | 180 | 13.1 |
| Real Time Big Data | 230 | 8.6 |
| Instituto Paraná Pesquisas | 109 | 7.3 |
Among clients, the press and the financial market pay the most: Grupo Globo declared R$ 16.5 million across 120 polls, Banco Genial R$ 12.5 million across 48 and Folha de S.Paulo R$ 8.8 million across 78. Parties barely appear: only 48 polls declared the party fund as the source of money, totaling R$ 1.9 million.
In 2026, banks and media groups pay for most of the polls voters see — only 1.6% of the money declared by clients came from the party fund.
Health leads earmarked spending — but not because of public opinion
Opinion polls repeat, year after year, that health sits at the top of Brazilians' declared priorities. The budget numbers seem, at first glance, to confirm that demand: the Health function absorbs more than half of all money moved through parliamentary earmarks, ahead of any other area. But the causal link doesn't work the way it appears — allocation follows the bargaining logic of Congress's thematic committees, which decide where the money goes regardless of whether any survey of popular priorities was ever consulted.
| Function | Amount (R$ bn) | % of total |
|---|---|---|
| Health | R$ 79.2 bn | 51.8% |
| Special charges | R$ 25.6 bn | 16.8% |
| Urban planning | R$ 11.6 bn | 7.6% |
| Agriculture | R$ 6.7 bn | 4.4% |
| Education | R$ 5.6 bn | 3.6% |
| Social assistance | R$ 3.9 bn | 2.6% |
More than half of all parliamentary earmarks go to health — not because opinion polls point to that priority, but because that's where the Health Committee wields the most control over the budget.
Four authors control the budget spigot — polls never enter the equation
Public opinion doesn't vote on earmarks one by one. What actually decides is a small group of committees and rapporteurships: together, four authors account for thirty billion reais in directed funds — a concentration of decision-making that no survey of public investment priorities has ever captured or directly influenced.
| Author | Earmarks | Amount (R$ bn) |
|---|---|---|
| HEALTH COMM. | 10 | R$ 9.6 bn |
| REGIONAL DEV. COMM. | 12 | R$ 8.6 bn |
| GENERAL RAPPORTEUR | 23 | R$ 8.6 bn |
| SOCIAL AFFAIRS COMM. | 8 | R$ 3.2 bn |
Four authors control R$ 30 billion in earmarks — a decision concentrated in few hands, shielded from any opinion poll or popular consultation.
Half the budget never leaves the page
Between 2018 and 2021, less than half of the amount authorized in earmarks was actually executed — the rest was booked as "carryover payables" (restos a pagar), an accounting category that pushes spending into future fiscal years, sometimes indefinitely. The execution rate improves from 2022 onward, but stays far from a hundred percent. While opinion polls measure the population's satisfaction with public-service delivery, a large share of the money meant to fund that delivery simply never gets spent in the year it was promised.
| Year | Execution rate |
|---|---|
| 2018-2021 | 44-49% |
| 2022-2024 | 62-70% |
Historically, roughly half the budget authorized in earmarks never gets executed — it becomes an accounting promise, not a delivered service.
Powerful cross-references
- Polls × Ballots: in 2022, nine of ten pollsters underestimated Bolsonaro; median error of 4 points.
- Polls × Money: 3,463 polls registered for 2026, R$ 206 million declared; Globo, Banco Genial and Folha lead the clients.
- Polls × Budget: public opinion doesn't influence earmark allocation.
- Earmarks × Concentration: four authors dominate R$ 30 billion in directed funds.
- Execution × Opinion: about 50% of the earmark budget becomes "carryover payables," regardless of what the population says it prioritizes.
Explanatory hypotheses
The disconnect between public opinion and budget allocation reveals a captured legislature: thematic committees and the general rapporteur decide where billions in resources go with no direct mechanism for consulting voters, and allocation by budget function reflects each committee's political weight, not surveys of social priority. The low execution rate, in turn, shows the structural incentive to commit funds without actually spending them — formally earmarking money creates immediate political capital for the sponsoring legislator, regardless of whether the promised public service ever materializes.
Policy implications
Active transparency about how earmark decisions are made could bring budget allocation closer to the priorities the population actually expresses in opinion polls. Tying earmarks to execution targets — not just commitment targets — could shrink the gap between the amount authorized and the public service actually delivered, turning the budget promise from a piece of political rhetoric into a measurable commitment.