Gold-standard cheap layer v1.1.0: datapackage.json (Frictionless, validated), croissant.json (MLCommons), CHECKSUMS.txt (SHA-256), DATASHEET.md (Gebru). Additive; data unchanged.
Browse files- DATASHEET.md +24 -0
DATASHEET.md
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Datasheet — Germany 2025 Electoral Divergence Electoral Divergence Dataset
|
| 2 |
+
|
| 3 |
+
Following *Datasheets for Datasets* (Gebru et al., 2021; arXiv:1803.09010). Part of the AFOS Analytics validated-cases collection, cross-referencing prediction market x opinion polls x press against the official result.
|
| 4 |
+
|
| 5 |
+
## 1. Motivation
|
| 6 |
+
Created by AFOS Analytics (independent, self-funded) to document, in an auditable form, the divergence between prediction markets, opinion polls and press narrative for this election (Federal (Bundestag), quem vence mais cadeiras). Use for research on market-vs-poll calibration, teaching, and replication of the AFOS divergence method.
|
| 7 |
+
|
| 8 |
+
## 2. Composition
|
| 9 |
+
Several observational units, each in its own table: a poll x party/candidate estimate (polls/), a daily market-implied probability (market-odds-timeseries), a divergence record (divergence), structural context (structural-context), and press items (news/). Instance counts and columns are enumerated per file in datapackage.json and DATA_DICTIONARY.md. Values are transcribed from public aggregates; nothing is imputed. No personal or sensitive data.
|
| 10 |
+
|
| 11 |
+
## 3. Collection process
|
| 12 |
+
Polls parsed deterministically from the public Wikipedia polling aggregation (CC BY-SA) by the shipped parse/build scripts; original pollsters credited. Market data from Polymarket on-chain via the AFOS server-side proxy. Structural context from the World Bank (WGI, WDI). Official result from the national electoral authority. Election dates: first round 2025-02-23; status completed.
|
| 13 |
+
|
| 14 |
+
## 4. Preprocessing / cleaning
|
| 15 |
+
Poll and market series are emitted in tidy long format (one row per poll x party/candidate, or date x party). Missing source values are left empty; nothing is smoothed except where a column is explicitly labelled as an average (e.g. a moving average). Raw market pulls are shipped as poly-raw JSON.
|
| 16 |
+
|
| 17 |
+
## 5. Uses
|
| 18 |
+
Supports the AFOS validated-case narrative and divergence analysis. Should NOT be read as a causal claim, nor as an official tally; the market-implied probability is not the same quantity as a vote share. Known limitation: poll aggregations may not carry sponsor/mode/question-wording/weighting per row (recoverable via each poll source).
|
| 19 |
+
|
| 20 |
+
## 6. Distribution
|
| 21 |
+
Hugging Face Hub (https://huggingface.co/datasets/AFOS-Analytics1/germany-2025-electoral-divergence); mirrored on GitHub; Harvard Dataverse DOI pending. License: data CC BY 4.0, code Apache 2.0. No redistribution-restricted source is included.
|
| 22 |
+
|
| 23 |
+
## 7. Maintenance
|
| 24 |
+
Maintained by AFOS Analytics (afos-analytics.com). Corrections and versioning via the repository; the shipped parser regenerates derived files from source. Country ISO3: DEU.
|