Open data
Everything public on this site is also published as plain data files, for software as much as for people. No account, no key, no interface to learn: files at fixed addresses, rebuilt with every update of the site.
The files
| File | Records | What’s in it |
|---|---|---|
events.json | 610 | Dated events, with their sources, actors, projects, laws, when they surfaced, and how each bears on each explanation (grades). |
sources.json | 1,228 | Citation records for the documents we rely on: headline, outlet, date and link. No quoted text. |
claims.json | 77 | Claims we make, each with the events that support or weaken it. |
explanations.json | 18 | The outcomes and explanations we test, with our published odds. |
predictions.json | 21 | Registered predictions: what counts, the window, what each explanation expects, and what was observed. |
projects.json | 58 | Resource projects we track: commodity, direction, capacity, status, owners. |
laws.json | 27 | Laws and other legal instruments, with their stages. |
actors.json | 379 | People, governments and companies named in events. |
series.json | 11 | Data series charted on the site, with their points and their source. These numbers belong to their sources: no licence of ours applies. (source’s licence) |
articles.json | 26 | Published articles: title, summary, date, series, link, translation, and the records each one cites. Article text is not included. (all rights reserved) |
Snapshot of the public records as of .
Licence
Our own records (events, summaries, grades, claims, odds, predictions, projects, laws, actors) are published under CC BY 4.0: reuse them freely, with credit to Carnet 49 and a link to the site. Two exceptions: the data series belong to their sources (named in the file), and our articles, titles and summaries included, stay all rights reserved (short quotes with credit are welcome). The files carry no quoted text from other publishers: for each source, only the headline, outlet, date and link.
How to read them
- Start with the index: it lists the files, the export date, the licence and the format version (currently 1).
- Every record carries the address of its page on the site (url, url_fr). Fields ending in _fr hold the French text; null means not translated yet.
- Records refer to each other by their short name (slug), or by code for explanations and predictions, so one file can be used on its own without rebuilding a database.
- Any website can read the files (CORS is open), read-only. Each data series is also a CSV file at /data/series/<name>.csv.
Why: we ask the outlets we cite for a route software may use (see here). We owed one of our own.