# Data Dictionary — AFOS Peru 2026 Electoral Divergence

All figures trace to a named pollster's published release (compiled from the Wikipedia aggregation, rowspan/colspan-aware parser) or to a public Polymarket market. Missing values are left **blank**, never imputed.

## `polls/peru-first-round-polls.csv` (long format)

One row per candidate per poll. 36 polls × up to 14 candidates = 327 rows, Jan→Apr 2026.

| Column | Type | Notes |
|--------|------|-------|
| `poll_date` | date | End of fieldwork (`YYYY-MM-DD`), derived from the fieldwork string. |
| `fieldwork` | string | Fieldwork window as published (e.g. "28 Feb–5 Mar 2026"). |
| `pollster` | string | Polling firm / publishing client (e.g. "Ipsos Perú/Perú 21"). |
| `sample` | integer | Sample size. |
| `candidate` | string | Candidate full name. |
| `party` | string | Party. |
| `percent` | number | First-round voting intention, %. |

**Candidates tracked:** Keiko Fujimori (Fuerza Popular), Roberto Sánchez (Juntos por el Perú), Rafael López Aliaga (Renovación Popular), Jorge Nieto, Ricardo Belmont (OBRAS), Carlos Álvarez, Alfonso López Chau (Ahora Nación), Marisol Pérez Tello, Carlos Espá, Fernando Olivera, José Luna (Podemos Perú), Yonhy Lescano, César Acuña (APP), Enrique Valderrama (APRA).

## `polls/peru-runoff-polls.csv`

| Column | Type | Notes |
|--------|------|-------|
| `poll_date` | date | End of fieldwork. |
| `fieldwork` | string | Fieldwork window. |
| `pollster` | string | Polling firm / client. |
| `sample` | integer | Sample size. |
| `fujimori_pct` / `sanchez_pct` | number | Runoff voting intention, %. |
| `lead_pp` | number | `fujimori_pct − sanchez_pct`, percentage points. |

## `polls/peru-polls.json`

Structured object: `{ description, source, election, counts, first_round[], runoff[] }`, where each poll carries `poll_date`, `fieldwork`, `pollster`, `sample`, and `results[]` (candidate/party/percent).

## `data/peru-market-odds-timeseries.csv`

Daily Polymarket win-probability per candidate, from the "Peru Presidential Election Winner" market.

| Column | Type | Notes |
|--------|------|-------|
| `date` | date | `YYYY-MM-DD`. |
| `candidate` | string | Canonical candidate name. |
| `polymarket_pct` | number | Implied win probability, % (daily close). |
| `volume_usd` | number | Cumulative market volume, USD. |

## `data/peru-divergence-timeseries.csv`

Each first-round poll result joined to the candidate's market odds on the poll date.

| Column | Type | Notes |
|--------|------|-------|
| `poll_date` | date | Poll fieldwork end. |
| `pollster` | string | Polling firm. |
| `candidate` | string | Candidate. |
| `poll_pct` | number | First-round vote intention, %. |
| `polymarket_pct` | number | Win probability on `polymarket_date`, %. |
| `polymarket_date` | date | Market date used: nearest available **on or before** `poll_date`. |
| `divergence_pp` | number | `polymarket_pct − poll_pct`, percentage points. |

> **Interpretation caveat:** a poll reports first-round *vote share*; the Polymarket contract prices *probability of winning the election*. These are different quantities — `divergence_pp` is the gap AFOS tracks editorially, **not** a like-for-like polling-error metric.