Closed zdj-garai closed 1 day ago
Can you report more about what's wrong here?
So for an example, take the presidential contest. Biden won 78,154 votes—and this is recorded in the precinct and Harvard tallies, while only 44,425 votes are recorded in the MEDSL dataset. Thus, he is missing 43% of his votes. Trump won 82,260 votes, but only 70,839 are recorded in MEDSL data; he's missing 14% of his votes. The Libertarian and Green candidates are missing 21% and 30% of their votes, respectively.
This dynamic trickles down to other contests, too. In State House district 66, MEDSL data includes only 83% of the Republican candidate's votes, and only 58% of the Democratic candidate's votes.
Harvard is correct, MEDSL is incorrect. Should figure out what is going on.
I think we can add McHenry in https://github.com/kuriwaki/cvr_harvard-mit_scripts/blob/main/code/04_build-release/metadata/counties_harv.csv for now. The Harvard counts match exactly, it'd be our largest Illinois county by far (though we'd still only have <4% of the state) and it will make Illinois slightly less Republican.
That sounds good to me, still haven't looked into this problem on the MEDSL end.
Use the Harvard/validated data; the MIT data undershoots the reported, official vote counts.