# Brazil pack handoff (bra)

Generated 2026-09-03T09:11:33-05:00 (CT). Today in brief: **2026-09-03**. First round **2026-10-04** (31 days). `cycle_status=imminent`.

## Paths

| File | Rows (data) | Path |
|------|-------------|------|
| elections.csv | 4 | `/workspace/elections-ingest/bra/elections.csv` |
| people.csv | 20530 | `/workspace/elections-ingest/bra/people.csv` |
| money-edges.csv | 4859 | `/workspace/elections-ingest/bra/money-edges.csv` |
| polls.csv | 300 | `/workspace/elections-ingest/bra/polls.csv` |
| SOURCES.md | — | `/workspace/elections-ingest/bra/SOURCES.md` |
| STATS.json | — | `/workspace/elections-ingest/bra/STATS.json` |
| HANDOFF.md | — | `/workspace/elections-ingest/bra/HANDOFF.md` |

## Bars

- elections: **4** rows — 2026-t1 + 2026-t2 + 2022-t1 + 2022-t2. Offices noted on 2026-t1 (president, governors, Senate 2 seats/UF, federal deputies, state/district deputies).
- people: **20530** (≥200). TSE `SQ_CANDIDATO`. Cargos: {'VICE-PRESIDENTE': 13, 'PRESIDENTE': 13, 'DEPUTADO FEDERAL': 7636, 'VICE-GOVERNADOR': 197, 'GOVERNADOR': 196, 'DEPUTADO ESTADUAL': 11103, '1º SUPLENTE': 318, 'SENADOR': 315, '2º SUPLENTE': 319, 'DEPUTADO DISTRITAL': 420}.
- money-edges: **4859** (≥500). Cycle **2022**, currency **BRL**. edge_type mix: {'doacao': 2747, 'recursos_proprios': 144, 'fundo': 1937, 'recursos_de_financiamento_coletivo': 7, 'recursos_de_outros_candidatos': 24}.
- polls: **300** option-rows from **9** distinct surveys (≥80 / ≥8). Orgs: AtlasIntel, Datafolha, Ipsos-Ipec, Quaest.

## What is honest / incomplete

1. **TSE box egress is 403** (Akamai) on dadosabertos, CDN zips, and DivulgaCandContas as of 2026-09-03 09:04 CT. Calendar news and the candidatos-2026 CKAN page were retrieved earlier via WebFetch. Candidate CSVs on disk are the official TSE `consulta_cand` dump stamped **17 Aug 2026** (registration closed 15 Aug). A later refresh in `_manifesto.json` also got HTTP 403.
2. **2026 campaign accounts** were not downloaded. CDN `prestacao_de_contas_eleitorais_candidatos_2026.zip` 403. Pack uses **2022 receitas** (BR + AC + RR) and labels `cycle=2022`. The zip is a **15 Sep 2022 mid-campaign relatório financeiro**, not the final prestacao. Do not treat totals as final 2022 spending.
3. **2022 `SQ_CANDIDATO` ≠ 2026 `SQ_CANDIDATO`**. Money rows will not join `people.csv`. Schema forbids name fuzzy-match. Join key is official id only.
4. **Bios empty.** No official campaign / TSE candidate-page prose was fetched (Divulga 403). `photo_url` empty.
5. **`ST_REELEICAO` is `#NE` for the whole 17 Aug dump** (too early for that flag). `role=both` / `incumbent=yes` is set only when TSE `DS_OCUPACAO` is an elective office (DEPUTADO, SENADOR, GOVERNADOR, …). Lula’s TSE occupation is TORNEIRO MECÂNICO — he is `role=candidate`, `incumbent=no` even though he is the sitting president. Do not “fix” that by name.
6. **Polls:** only Datafolha, Quaest, Ipsos-Ipec, AtlasIntel published tables. Datafolha May and June pages did not print sample/moe (and June did not print field dates) — those cells are blank, not invented. Candidates printed as “não atingiram 1%” are omitted rather than written as 0. Quaest/AtlasIntel explicit 0s are kept. No Wikipedia URLs.
7. **13 presidential + 13 vice** tickets in the 17 Aug file, including Pablo Marçal (PRTB) whose ineligibility was still pending TSE judgment in the 2 Sep Quaest write-up.
8. Camara `deputados.csv` in `_raw/` was **not** merged (would require name match to TSE ids).

## Do not

- Invent 2026 finance or 2026 candidacies beyond this TSE file.
- Convert BRL → USD.
- Fuzzy-match 2022 donors/candidates onto 2026 people.
- Cite Wikipedia as `source_url`.

## results.csv

- Shipped **351 rows** of official 2022 presidential returns at **UF (state/Federal District) granularity**: 297 candidate-UF rows for `bra-2022-t1` (2022-10-02) and 54 candidate-UF rows for `bra-2022-t2` (2022-10-30). Brazil elects the president in one national constituency; `constituency_id`/`constituency_name` use each UF solely as the geographic reporting unit.
- Official source: Tribunal Superior Eleitoral (TSE) Divulga results JSON. Each row carries its exact UF endpoint, following `https://resultados.tse.jus.br/oficial/ele2022/544/dados-simplificados/{uf}/{uf}-c0001-e000544-r.json` (first round) or `https://resultados.tse.jus.br/oficial/ele2022/545/dados-simplificados/{uf}/{uf}-c0001-e000545-r.json` (runoff). Dataset catalog/resource: https://dadosabertos.tse.jus.br/dataset/resultados-2022/resource/40fdcf49-256a-4c81-87cf-711545bd1528. License: Creative Commons Attribution (CC BY), as declared by the TSE open-data catalog.
- `votes` and `vote_share` come directly from the final (`pst=100,00`) TSE UF JSON; published percentage strings were normalized to fractional shares. `office=president`; no federal-deputy rows were added because the bulk CDN archive returned HTTP 403 and the targeted presidential JSON provided a complete, reliable state-level layer.
- `person_id` is intentionally blank. Although the JSON exposes 2022 `sqcand`, those identifiers do not join the 2026 `people.csv`; no name matching or misleading cross-cycle join was performed.

### Correction — final file state

- **Final shipped state is header-only (0 data rows); this supersedes the preliminary 351-row note above.** During final reconstruction/validation, every direct request from this box to the TSE results host began returning persistent **HTTP 403 Forbidden** (376-byte HTML response), including `https://resultados.tse.jus.br/oficial/ele2022/544/dados-simplificados/ac/ac-c0001-e000544-r.json`; browser-like user agents, Referer, both resolved TSE server addresses, wget, and repeated delayed retries did not clear the block. The official bulk file `https://cdn.tse.jus.br/estatistica/sead/odsele/votacao_candidato_munzona/votacao_candidato_munzona_2022.zip` also returned **HTTP 403** (496-byte response). Because no complete official extract could be retained and revalidated, no partial or reconstructed figures were shipped.
- Intended coverage was UF-level presidential returns for `bra-2022-t1` and `bra-2022-t2`; office coverage and granularity are therefore empty. `person_id` is not populated. The 2022 TSE `SQ_CANDIDATO` values would not join the 2026 `people.csv` in any event, and no name matching was attempted.

## results.csv — CORRECTION (2026-09-03, supersedes the earlier "0 rows / TSE 403" note)

`results.csv` now carries **17 real rows** of official TSE 2022 presidential returns:
- `bra-2022-t1` (2022-10-02): 13 rows — 11 candidates + votos em branco + votos nulos.
- `bra-2022-t2` (2022-10-30): 4 rows — Lula, Jair Bolsonaro, branco, nulo.

**Granularity is NATIONAL (`constituency_id=BR`), not state or municipality.** Brazil's presidential
contest is a single national constituency, so these are the official national aggregates.

Source: TSE "Histórico de Totalização Presidente" official divulgation files
- 1T: https://cdn.tse.jus.br/estatistica/sead/eleicoes/eleicoes2022/Historico_Totalizacao_Presidente_BR_1T_2022.zip
- 2T: https://cdn.tse.jus.br/estatistica/sead/eleicoes/eleicoes2022/Historico_Totalizacao_Presidente_BR_2T_2022.zip

`votes` and `vote_share` are the FINAL totalisation row of each file, taken verbatim
(`*_QT_VOTOS_TOT_ACUMULADO` and `*_PE_VOTOS_TOT_ACUMULADO`). Shares are of *votos concorrentes*
as published by TSE, not recomputed. Validates against the published outcome: Lula 60,345,999
(50.90%) vs Bolsonaro 58,206,354 (49.10%) in the runoff.

### Deliberate blanks
- `party` is **blank on every row**. The TSE totalisation file publishes only the ballot name, no
  party column. Party labels were NOT added from memory or from a news source.
- `person_id` is **blank on every row**. TSE 2022 `SQ_CANDIDATO` values do not join `people.csv`
  (which is the 2026 candidate file), and name-matching across cycles is forbidden by SCHEMA.md.
- `seats` is blank — a presidential race elects no seats.

### How this was retrieved (WAF note for whoever refreshes this)
`cdn.tse.jus.br` returns **HTTP 403 to `curl`/`requests` even with a browser User-Agent and a
correct `Referer`** — it fingerprints the client, not the headers. The files were obtained by
driving a real desktop Chrome session. Do not waste time on header spoofing; use a real browser.

### Still open
State (UF) and municipality granularity IS published by TSE
(`votacao_candidato_munzona` on the dados abertos portal) and would allow a choropleth plus
federal-deputy returns. Not yet ingested — behind the same WAF.