# Chile (chl) — HANDOFF

## Settled next-election call
| field | value |
|---|---|
| next national contest | **2029 general — November 2029 (month precision only)** |
| `date_confidence` | **`constitutional_due`** |
| `cycle_status` | **`sitting`** |
| `election_id` | `chl-gen-2029` |

`election_date` is deliberately **month-precision (`2029-11`)**: Ley 18.700 sets the presidential and
parliamentary election in **November** of the year before the term ends, but the exact Sunday is not
fixed until SERVEL/legal convocation. A specific day would have been invented, so it was not.

`cycle_status=sitting` — confirmed by arithmetic. The runoff was **2025-12-14**, which is **263 days**
before 2026-09-03. That is well outside the **120-day** `just_held` window, so `just_held` is wrong and
`sitting` is right. José Antonio Kast Rist took office **2026-03-11**; the legislature elected
2025-11-16 is live.

## Contests filed (5)
| election_id | date | date_confidence | cycle_status | what |
|---|---|---|---|---|
| `chl-pres-2025-r1` | 2025-11-16 | confirmed_official | sitting | presidential 1st round |
| `chl-pres-2025-r2` | 2025-12-14 | confirmed_official | sitting | presidential runoff — **Kast won** |
| `chl-dip-2025` | 2025-11-16 | confirmed_official | sitting | Cámara de Diputados (155) |
| `chl-sen-2025` | 2025-11-16 | confirmed_official | sitting | Senado (partial renewal) |
| `chl-gen-2029` | 2029-11 | constitutional_due | distant | next general |

## people.csv — 213 rows
`person_id` scheme: **official ids from three different sources, not unified**:
- Diputados (155): official `DIPID` from the Cámara's own XML web service (`opendata.camara.cl`).
- Senadores (50): a stable slug derived from the official `senado.cl` roster page (senado.cl does not
  expose a numeric id on the public listing) — SAY SO: this is a site-slug id, not a numeric official id.
- 2025 presidential candidates (8): SERVEL's own `cod_candidato` from the preliminary-results XLSX.
These three id spaces **do not overlap** and must not be cross-joined by name.
`bio_text` is empty for all 213 rows — neither camara.cl, senado.cl, nor SERVEL publish a verbatim bio
field on these pages/files, so none was authored.

## money-edges.csv is POPULATED — 18,225 rows, CLP as filed
**No URL was invented.** SERVEL's financing pages were probed and the two *Reporte de Aportes* XLSX
files were confirmed live before use (exact URLs in SOURCES.md).

This is transaction-level donation data, not aggregates:
- 16,253 rows from the definitive report (`chl-dip-2025` — presidential 1st round + parliamentary donors)
- 1,972 rows from the runoff report (`chl-pres-2025-r2`)
- `edge_type` = `donation` for all rows (SERVEL's own "aportes" label)
- `amount` is **CLP exactly as filed** — no conversion, no scaling
- `as_filed_note` preserves the source report and the row's filed attributes verbatim, including the
  legal basis for anonymous small-donor rows (Ley 19.884, "aporte menor sin publicidad")

**`person_id` is blank on all 18,225 rows.** SERVEL's aportes register keys donations to the receiving
candidate's **name string** in the XLSX, not to `DIPID` or any other roster id, and SCHEMA.md bars
name fuzzy-matching between money and roster. Rather than guess a join, `person_id` is left blank and
the recipient's name is preserved as-filed in `committee_name`. **Gap**: `chl-sen-2025` (Senate) and the
1st-round-only presidential contributions were not found as separate downloadable files at fetch time —
this money layer covers the deputies race and the runoff only, not the full 2025 cycle.

## results.csv — 1,529 rows, verified against official figures
| contest | rows | grain |
|---|---|---|
| `chl-pres-2025-r1` | 170 | región |
| `chl-pres-2025-r2` | 68 | región |
| `chl-dip-2025` | 1,152 | distrito |
| `chl-sen-2025` | 139 | circunscripción senatorial |

**Runoff verified**: Kast **7,235,213 (58.25%)** vs Jara **5,186,244 (41.75%)** on 12,421,457 valid
votes. **1st round verified**: Jara 26.75%, Kast 23.95%, Parisi 19.80%, Kaiser 13.94%, Matthei 12.44%.

Grain/denominator conventions:
- `vote_share` denominator is SERVEL's own **`Total Sufragios Validamente Emitidos`** per constituency.
- SERVEL summary rows (`Total Suma Calculada`, `Total Sufragios Validamente Emitidos`) are excluded as
  candidate rows.
- `Votos Nulos` / `Votos en Blanco` are retained as bracketed `party` labels (`[VOTOS NULOS]`,
  `[VOTOS EN BLANCO]`) with blank `candidate` and blank `vote_share`.

⚠️ These are SERVEL's **preliminary** (*resultados preliminares*) publications — that is what SERVEL
exposes in machine-readable XLSX. Definitive results come from the Tribunal Calificador de Elecciones
(TRICEL) and are PDF-only. The `source` column says *preliminares* on every row.

## polls.csv — 79 rows, 5 published surveys, 4 organizations
| poll_id | rows | org | field dates |
|---|---|---|---|
| `criteria-2025-08-m2` | 12 | Criteria | 2025-08-12 to 08-14 |
| `criteria-2025-08-m4` | 18 | Criteria | 2025-08-26 to 08-28 |
| `panel-udd-2025-08-15` | 10 | Panel Ciudadano UDD | 2025-08-14 to 08-15 |
| `cadem-plazapublica-2025-08-10` | 17 | Cadem Plaza Pública | unprinted start; publish date 2025-08-10 |
| `atlasintel-bloomberg-2025-11-14` | 22 | AtlasIntel | 2025-11-10 to 11-14 |

Each survey contributes both a first-round topline and its published two/three-way runoff scenarios
(all counted as option-rows, one row per response option, per SCHEMA). All figures were transcribed
directly from each publisher's own PDF bulletin — no figure was read off a news article or Wikipedia.
Where a publisher did not print `sample_n`, `moe`, or an exact `field_start`, those cells were left
blank rather than estimated (see individual `notes` cells for exactly what each bulletin states).

**Excluded, with reasons — do not re-add without a real table:**
- **CEP** — survey #93 (Mar–Apr 2025) was downloaded and read in full. It contains presidential
  *approval/evaluation* figures ("evaluación de personajes") but **no** "if the election were held
  today, for whom would you vote" table in the published report, so no CEP row was filed.
- **Data Influye**, **Black&White** — no downloadable, citable topline document (PDF or data page) was
  located for either house within the time available. Not fabricated, simply not filed.
- 79 rows is short of the ">=40 option-rows" floor's spirit only in that it draws from 4 of the 7 named
  houses; it already clears the >=40 floor on its own.

## Gaps / next steps
- `people.csv` incumbents = 155 diputados + 50 senadores = 205, plus 8 SERVEL-listed 2025 presidential
  candidates = 213 total.
- Deputy period dates were ambiguous in the web service (multiple `FechaInicio` values); the current
  period was selected on inspection.
- Senate is only **partially** renewed each cycle; `chl-sen-2025`'s 139 rows are that 2025 half, not
  the whole 50-seat body.
- Money layer does not yet cover Senate or 1st-round-only presidential filings (see money-edges note
  above) — a downstream consumer should not assume `money-edges.csv` is complete for all four contests.