AFOS Daily · Daily Synthesis
May 19, 2026
Prediction Markets × Polls × News
Synthesis generated from auditable data. Every claim cites its source.
AtlasIntel published and the three signals that shook everything: Flávio drops six points in the national poll, Polymarket opens historic gap of 17.35pp Lula over Flávio (from +11.95pp yesterday), and the inflation tail above 6.50% collapses to 9.00% (drop of 5.05pp in 26 hours).
Three movements define the day. AtlasIntel published its first national poll after the audio between Flávio Bolsonaro and Daniel Vorcaro was released: Flávio drops six points in the runoff, falling to 41.8% versus 48.9% for Lula. Polymarket reacted materially — Lula rose to 45.50% (↑2.00pp in 26 hours) and Flávio fell to 28.15% (↓3.40pp), opening a gap of +17.35pp, the largest recorded in the May series. The inflation tail above 6.50% fell to 9.00% (↓5.05pp in 26h), and Flávio's team filed a complaint with the TSE requesting suspension of poll release.
1. Prediction market
In the Polymarket presidential race, Lula rose to 45.50% (↑2.00pp in 26h, vs 43.50% at May 18 close) and Flávio Bolsonaro fell to 28.15% (↓3.40pp, cycle low since inception). The gap between the two, which had narrowed to +11.95pp on May 18, reopened to +17.35pp Lula — a ↑5.40pp movement in 26 hours, the largest daily movement of the month. Accumulated volume in the presidential market: USD 78.85M, with the Lula contract at USD 5.54M and Flávio at USD 5.70M.
The third way reorganized. Renan Santos surged to 12.55% (↑2.85pp in 26h) and consolidated the top of the third way in the presidential race, above Romeu Zema. In the 3rd place market, Renan took the lead with 35.50% (↑4.00pp), surpassing Zema, who collapsed to 28.50% (↓8.00pp in 26h). Caiado, in the same market, fell to 13.50% (↓4.00pp). The migration is asymmetric: Flávio yields and the market prices Renan as the primary recipient, not Zema or Caiado.
In the 2nd place market, Flávio remains at 62.50% (↑0.50pp) — the PL electoral base holds in the runoff scenario despite the decline in the presidential race. Lula fell to 14.50% (↓1.50pp) and Renan rose to 9.50% (↑0.60pp in 26h, ↑1.80pp in 48h). Haddad recovered to 4.25% (↑1.70pp in 26h), gaining relevance as an internal PT hedge.
The contract on STF minister impeachment before 2027 registered a slight recovery to 6.05% (↑0.75pp in 24h vs the cycle low of 5.30% on May 18), but remains below the 8% pre-launch baseline. In the Senate, PL remains stable at 78.50% (clear dominance), while Republicanos collapsed to 4.20% (↓6.80pp vs the May 17 peak of 11.00%) — returning all recent gains.
In the 2026 inflation bands, the high tail suffered the day's most material movement. The 7.00%+ band collapsed to 3.30% (↓5.35pp in 26h vs 8.65% on May 18), and the combined bands above 6.50% fell to 9.00% (↓5.05pp in 26h, ↓6.85pp vs the May 17 peak of 15.85%). Central band 5.00-5.49% remains at 28.00% (stable), and the 4.50-4.99% band rose to 21.75% (↑3.25pp). Technical reading: the market withdrew weight from the extreme fiscal scenario without moving the central band.
2. What the polling institutes recorded
AtlasIntel published its first national poll after the Flávio-Vorcaro audio, with sample n=5,000. In the first round, Lula registered 47.0% and Flávio 34.3% — a gap of +12.7pp. In the runoff, Lula 48.9% × Flávio 41.8%, gap of +7.1pp. Relative to the same institute's previous poll, Flávio dropped 6 points in the runoff (was 47.8%). Rejection of Flávio rose from 49.8% to 52.0%, and the survey also found that 47.4% of voters fear a Flávio election while 40.5% fear Lula reelected. Gazeta do Povo reported that AtlasIntel included mention of the audio in the questionnaire for this round — the company denies that this impacted the result.
Quaest published in parallel another snapshot: 32% of voters point to Lula's reelection as the best outcome, 24% indicate a Bolsonaro family return, and 22% prefer an outsider victory. The Datafolha of May 16 (1st round Lula 38% × Flávio 35%, gap +3pp; 2nd round 45×45) remains the last reference point for the São Paulo anchor poll — record available on the TSE disclosure portal under protocol BR-074892026. The gap between AtlasIntel and Datafolha in the first round is 9.7pp — the largest divergence between two national institutes in this cycle.
The next expected publication is Datafolha on May 22, with sample n=2,004 and explicit scenario including Michelle Bolsonaro as a replacement hypothesis on the PL ticket — direct test of the right's reorganization that Polymarket is already discussing (Michelle presidential 2.25%, 2nd place 1.55%, replacement narrative with low traction in the market even with public discussion).
📅 Research calendar — next 7 days
Polls registered with the TSE with release scheduled between May 20 and May 26. Inclusion in the table does not mean publication is confirmed — institutes may delay or cancel release. Filter applied: sample ≥ 1,000. Each protocol linked to the TSE public query.
| Date | Institute | Sample | Scope | TSE Protocol | Conf. |
|---|---|---|---|---|---|
| May 20 | VETOR ARROW 🔥 | 9,000 | state | BR049702026 | 0.7 |
| May 20 | Inst. Vox Brasil | 2,100 | state | BR024162026 | 0.7 |
| May 20 | Real Time Big Data | 1,600 | state | BR017442026 | 0.8 |
| May 20 | A J Tech | 1,600 | state | BR092092026 | 0.6 |
| May 21 | VETOR ARROW 🔥 | 9,000 | state | BR093012026 | 0.7 |
| May 21 | 100 Cidades | 2,000 | state | BR065292026 | 0.7 |
| May 21 | Real Time Big Data | 1,600 | state | BR071142026 | 0.8 |
| May 22 | Datafolha | 2,004 | national | BR074892026 | 0.9 |
| May 22 | Gerp | 2,000 | state | BR079712026 | 0.7 |
| May 22 | PNH | 1,500 | state | BR038502026 | 0.6 |
| May 23 | Data Census | 2,000 | state | BR023422026 | 0.7 |
| May 23 | Econometrica | 1,607 | state | BR082622026 | 0.6 |
| May 24 | Inst. Vox Brasil | 2,100 | state | BR079612026 | 0.7 |
Source: public record from TSE disclosure portal and AFOS Analytics API. 🔥 highlights samples ≥ 3,000. Status "registered ≠ published" — confirmation of effective release requires verification from two primary sources before citing numbers.
3. What the press covered
Flávio Bolsonaro's team filed a request with the TSE to suspend the release of the AtlasIntel survey — institutional escalation via TSE as an attempt to block the information circuit. In parallel, columnist José Casado wrote in VEJA that Flávio had discussed his withdrawal from pre-candidacy with his father and was attempting to contain damage within the PL — treatment as a signed report, not as consolidated fact, given that the withdrawal itself was not formally announced.
In government news, Folha de S.Paulo reported that Lula brought Jorge Messias to an event with electoral tone before formally communicating to allies about the resubmission of the nomination to the STF. O Globo reported in parallel that the president's insistence on Messias's nomination irritates Senate president Davi Alcolumbre. The observational reading: the government treats the resubmission as a political priority even with signs of friction with the Senate, and the Polymarket for STF impeachment remained low (6.05%) throughout the day — the market does not price this tension as an institutional threat.
Still in government news, a cabinet minister acknowledged to Folha that the tax on the blouses "was a mistake" and confirmed that the Flávio-Master link will formally enter the reelection campaign. For the third way, O Globo reported that the replacement of Aldo Rebelo by Joaquim Barbosa in the Democracia Cristã pre-candidacy provoked internal split and threat of judicial action — a sign that the center space remains fragmented even with the new name.
4. Day's divergences
Market × national survey: Polymarket reacted in the direction of AtlasIntel but with different magnitude. AtlasIntel registered a gap of +12.7pp in the 1º turno; the market prices a gap of +17.35pp in implicit probability. The difference of 4.65pp between the two indicates that Polymarket continues more aggressive than the published survey — something of the Vorcaro impact remains priced into the market and was not fully captured by the institute.
Survey × survey: AtlasIntel 1T of May 19 (+12.7pp Lula) vs Datafolha 1T of May 16 (+3pp Lula) = divergence between institutes of 9.7pp. Difference attributable to the time window (Datafolha in field before the audio's impact reached the sample), to methodology (AtlasIntel includes mention of the audio in the questionnaire, fact confirmed by the company itself) and to sample design. The Datafolha of May 22 will be the next data point capable of calibrating this divergence.
Market × institutional event: Flávio's team filed with the TSE requesting suspension of survey release, but Polymarket continued moving in the direction unfavorable to Flávio throughout the day. Movement contrary to the narrative of institutional blocking as an effective tool for electoral protection — market prices that the information circuit has already circulated.
In summary
-
Published AtlasIntel confirmed part of what the market had already priced, but the residual difference of 4.65pp between Polymarket and survey suggests that the Vorcaro impact remains partially outside the survey's picture. The next Datafolha (May 22, with Michelle scenario) is the next data point of divergence.
-
The reorganization of the third way was asymmetric: Renan Santos absorbed the market migration (presidential 12.55%, 3L top 35.50%); Zema gave up the 3L top for the first time (28.50%, ↓8.00pp); Caiado also lost position (3L 13.50%, ↓4.00pp). Market did not treat all of the non-Bolsonarist right the same way.
-
The extreme fiscal tail (inflation ≥6.50%) collapsed 5.05pp in 26 hours to 9.00%, an atypical movement for a day with electoral survey as central event. Technical reading: the inflation market appears to react to the government's institutional signaling (admission of error, legislative recalibration) in parallel to the political movement — not as a direct reflection of the survey.
Sources Consulted
Articles with direct link to the news:
- O Globo — Troca de Aldo Rebelo por Joaquim Barbosa provoca racha no Democracia Cristã
- O Globo — Insistência de Lula em reenviar indicação de Messias ao STF irrita Alcolumbre
- Polymarket — Brazil Presidential Election
- Polymarket — STF Justice impeachment before 2027
- Polymarket — Brazil Senate election most seats
- Polymarket — Brazil Annual Inflation 2026
- TSE — Portal de divulgação de pesquisas
Secondary articles (Google News redirect URL — click resolves to the article):
- BBC — Pesquisa AtlasIntel: Lula aparece 7 pontos à frente de Flávio após áudio a Vorcaro
- CNN Brasil — Atlas: Lula tem 47% das intenções de voto no 1º turno; Flávio, 34,3%
- CNN Brasil — Atlas: Flávio cai seis pontos e tem 41,8% contra 48,9% de Lula no 2º turno
- CNN Brasil — Atlas: 47,4% têm medo de eleição de Flávio; 40,5% temem Lula reeleito
- Folha de S.Paulo — Lula leva Messias para ato com tom eleitoral antes do reenvio ao STF
- Folha de S.Paulo — Flávio Bolsonaro pede que TSE suspenda divulgação de pesquisa
- VEJA — Flávio Bolsonaro discutiu renúncia com o pai e tenta conter danos no PL (José Casado)
- GZH — AtlasIntel: Lula tem 48,9% e Flávio Bolsonaro 41,8% no segundo turno
- Gazeta do Povo — AtlasIntel inclui áudio atribuído a Flávio em pesquisa; empresa nega impacto
- G1 — Quaest: 32% dizem que reeleição de Lula é o melhor resultado para o Brasil
- Valor Econômico — A nova pesquisa Datafolha para presidente que inclui Michelle Bolsonaro
Sources cited in this text: AtlasIntel via BBC, Estadão, Folha de S.Paulo, CNN Brasil, Poder360, JOTA, Valor Econômico, Gazeta do Povo, GZH, Money Times; Datafolha May 16 via TSE; Quaest May 19 via G1; O Globo; Polymarket; TSE disclosure portal.
Method: this synthesis is generated automatically from the auditable data of the AFOS Analytics platform, under versioned code rules in git. All claims can be verified on the platform or in the linked sources.
Integration: to see the live data and detailed candidate analysis, access the full Dashboard. To understand the method in depth, read The Method.
Method: this synthesis is generated automatically from auditable data on the AFOS Analytics platform, under code-versioned rules in git. All claims can be verified on the platform or in the linked sources. Understand the automated governance.
Integration: for live data and detailed candidate analyses, access the full dashboard. To understand the method in depth, read The Method.
Glossary: Brazilian political terms used in the syntheses (TSE, STF, BolsoMaster, lideranças envelhecidas, etc.), definitions in 3 languages. See the full glossary →