kuriwaki / cvr_harvard-mit_scripts

6 stars 1 forks source link

[CA] Inyo #323

Closed aconevska closed 2 months ago

aconevska commented 2 months ago

Recommendation: Use Harvard, though cannot recover missing votes because redacted in raw data for the three smallest precincts. We know which precincts but no way to recover at voter level.

Both MEDSL and Harvard have issues with Inyo. Harvard is short 104 votes for president (60 for Biden, 43 for Trump, and 1 for Jorgensen), and MEDSL has 217 excess votes for these presidential candidates.

For Harvard, the problem is due to a set of redacted votes in the raw data, of which there are 107. These come from three tiny precincts (the smallest in the county): 114 (3), 115 (56), and 118 (48). Note that these redacted votes amount to the total reported ballots cast in each precinct. And they account for the votes we are missing for each candidate:
Precinct 114 Trump votes = 2, Precinct 115 Trump votes = 24, Precinct 118 Trump votes = 17 = 43 Precinct 114 votes for Biden = 0, Precinct 115 votes for Biden = 31, Precinct 118 votes for Biden = 29 = 60 Precinct 118 votes for Jorgensen = 1.

I would have to look into MEDSL's data further to see how they over counted. They overcounted by 217, and 110 if we assume that includes the redacted votes somehow (MEDSL extra votes are also more than twice the redacted vote total, so probably not coming from there).

This is a rural county but an extremely small one. But still, if we're worried about rural/conservative representation for CA, I can look more closely at where MEDSL is overcounting.

mreece13 commented 2 months ago

Related to redaction, it should be fixed in the next build.

mreece13 commented 2 months ago

MEDSL now matches Harvard at least, we are still both a bit short from the truth because of redaction.