Vintage and restatements
"The most recent published factor" resolves to a different year for different countries, and a published year can be revised afterwards. A filing has to be able to say which version it used.
The bars are the distribution; the line is what a reporter has to decide. Insisting on a 2025 factor covers 43% of countries. Accepting 2024 or later covers 94%, and the two years below that add the last six points. Filtering the table to a single latest year would silently drop more than half the world, and tightening the cut-off by one year costs fifty points of coverage.
The countries furthest behind
Grid size is no protection: Ukraine's most recent published factor is 2022, on a 111 TWh grid.
Has the factor been restated?
A disclosure is filed against the factor published at the time. When the publisher revises that year afterwards, the filing does not become wrong; it becomes a filing against a superseded factor, which you have to be able to demonstrate. This warehouse keeps every version of every factor from 2015 on.
Nothing restated yet, and on a warehouse built from scratch that is the honest answer rather than a broken query: a snapshot can only record a revision it was present for. The first run stores version 1 of every factor, and a row becomes restated the first time a later run finds a different number.
That version history is the one part of this table a rebuild cannot reproduce, so every build carries it forward from the previous data release instead of recomputing it. The Restatements page does the same for OWID's CO₂ estimates.