kuriwaki
commented
2 months ago I agree this is a "release as is and remaining issues are low priority" given that you've already checked precinct coverage.
Sounds like La Riva and Charles might be write-ins? So some versions of the count would lump them as write-ins, while other versions might list them as separate candidates.
Loop me in the email with Jim if he responds.
I was wondering if the 93 votes or so explained by write-ins (a write-in jpeg file got deleted during the Excel -> csv transfer per https://github.com/kuriwaki/cvr_harvard-mit_scripts/issues/123#issuecomment-2136374388). But from our spreadsheet it looks like both Biden and Trump are missing about 50 votes each, so this can't be it.
mreece13
Recommendation: We can use Harvard or MEDSL, not much different.
Harvard has 22264 votes for President for Amador while the official county tally is 22397, making us 133 votes short. These seem to be broadly scattered across candidates as well as precincts - i.e. no pattern for where they are missing. But the Harvard processed data is missing all votes for candidate La Riva - which the county office reports there were 41 in total. On the other hand we have 40 votes for Mark Charles, which the county office reports none of. So it appears Harvard miscounted La Riva's votes as Mark Charles', though this is inconsequentially low. I will let Jim know and he can fix it at a convenient time.
The Harvard data includes all precincts - i.e. the issue is not coming from one or a set of missing precincts.
It does not appear the issue is with type of vote either - i.e. though "Ballot Style" ("countgroup" in some CA county files) is the same for all observations in the raw data ("General Election"), when comparing the county's official tally of "Election Day Votes" and "Mail Votes", we have too many to be just "Mail Votes" or just "Election Day Votes".
The raw data also has presidential votes marked as "Overvotes" though the county reports at least 15. But that
I think the most likely source of the discrepancy comes from how the county handled "Undervotes" in their data release. By precinct, the difference in our presidential vote totals and the county's presidential vote totals is often equal or almost the same as the official undervote count. But I would need to look into this more to be sure. Probably not important given the small n of the county but let me know @mreece13 @kuriwaki