Author: tmtm

There wasn’t much feedback on my suggestion that successor states have their UN membership data ‘reset’, but what little there was was positive, so I’m going to update Wikidata accordingly for each of those, which will adjust the time period we need to look at here as well.

For countries like Hungary or Albania where there is a change in entity for the state, but the office of head of government is currently modelled as a single item throughout, we’ll continue to treat it as one, but where there’s a clear distinction between the old and new entity, we’ll only look at the post-transition period.

After a lot of thought, and some experimentation, I’ve updated the scoring system. The new “max score” is now simply ten times the expected number of entries, and actual score is ten times the number we actually have minus two times the number of PositionHolderHistory warnings. This eliminates the edge cases (negative scores, or 0% even when there’s useful data), and better emphasises the value of data being present over there being an issue with it (on the basis that it’s generally easier to spot and fix problem data than missing data).

I’ve updated the spreadsheet, and will also backfill the new score across the previous posts here.

Wikidata only currently knows of the office of Prime Minister being Head of Government, not that before 1988 this was the President instead. So none of those earlier officeholders are ‘findable’ as head of government, and thus don’t currently score.

• Date of joining: 1945-10-24
• Expected count from Wikipedia (all): 25 + 12 earlier Presidents
• Number in Wikidata: 22
• PositionHolderHistory warnings: 7
• Score: 77/144 = 53% 206/370 = 56%

• Date of joining: 1980-09-16
• Expected count from Wikipedia (Milton Cato onwards): 4
• Number in Wikidata: 1
• PositionHolderHistory warnings: 0
• Score: 0/12 = 0% 10/40 = 25%

We’ve had a few cases before where our initial scoring approach has seemed shaky, but this is the worst yet. We would expect to score somewhere around 25% — though as the incumbent is expected to have slightly less data, it wouldn’t be terrible if it was slightly lower to reflect that — but we end up with 0% in the current system. (Should Wikidata not have any data at all we should expect a score of 0%, but our current approach gives us a negative score instead!)

So I suspect it’s time to rethink!

• Date of joining: 1973-09-18
• Expected count from Wikipedia (Lynden Pindling onwards): 6*
• Number in Wikidata: 8
• PositionHolderHistory warnings: 0
• Score: 28/28 = 100% 80/80 = 100%

This is the first case we’ve seen where Wikidata is more accurate than any of the Wikipedias, none of which include Cynthia A. Pratt being acting Prime Minister for a month in 2005.

• Date of joining: 1970-10-13
• Expected count from Wikipedia (all): 12
• Number in Wikidata: 9
• PositionHolderHistory warnings: 3
• Score: 29/44 = 66% 84/120 = 70%

As with a few other cases, we have a change in entity for the country midway through (from the Dominion of Fiji to the Republic of Fiji), but as the item for the Prime Minister currently spans both, we don’t need to worry about that yet. Someone may split these at a later date, however.

• Date of joining: 1976-12-15
• Expected count from Wikipedia (Tufuga Efi onwards): 7
• Number in Wikidata: 2
• PositionHolderHistory warnings: 0
• Score: 4/24 = 17% 20/70 = 29%

• Date of joining: 1979-09-18
• Expected count from Wikipedia (Allan Louisy onwards): 10
• Number in Wikidata: 4
• PositionHolderHistory warnings: 4
• Score: 32/100 = 32%

• Date of joining: 1978-12-18
• Expected count from Wikipedia (Patrick John onwards): 8
• Number in Wikidata: 8
• PositionHolderHistory warnings: 5
• Score: 70/80 = 88%

• Date of joining: 1978-09-19
• Expected count from Wikipedia (all): 18
• Number in Wikidata: 6
• PositionHolderHistory warnings: 8
• Score: 12/68 = 18% 44/180 = 24%

I think this is the first case we’ve encountered where Wikidata knows that someone relevant has been Prime Minister, but their “position held” statement is “bare”, with no information as to when, etc.