Suspicious data pattern in recent Venezuelan election (statmodeling.stat.columbia.edu)
from fossilesque@mander.xyz to science@mander.xyz on 01 Aug 2024 02:08
https://mander.xyz/post/16130168

#science

threaded - newest

Don_Dickle@lemmy.world on 01 Aug 2024 02:16 next collapse

One of my cousins is stationed there and he said the elections were fucked from the start. He saw armed soldiers taking out ballots and checking them. The security is fucked…I am stationed there next month it will be neat on what I run into.

lvxferre@mander.xyz on 01 Aug 2024 02:43 next collapse

I’ve checked a few newspapers in Spanish (El Heraldo, El Deber*, Clarín), and the total number of votes seems rather consistent with what the article linked in the OP says. Based on that I think that the hypothesis in the P.P.S. (sloppy post-processing) is probably incorrect, as multiple independent sources wouldn’t be likely to retrieve the info from the same intermediary.

IMO the simplest explanation is the implicit main hypothesis (someone makes up the percentages, then apply them to the total # of votes, then round them down to the nearest integer). As in: fraud. Fairly common in the Americas in general, not just in Venezuela, just a bit too blatant for the traditional modus operandi (that involves stuff like: finding bullshit reasons to not count some urns, your grand-grand-grandma coming back from the grave to vote on your candidate, etc.).

*El Deber lists the wrong percentage for Urrutia, 42,2% instead of 44,2%. This looks like a typo, given that the total number of votes (4.445.978) is consistent with the other sources.

bamboo@lemmy.blahaj.zone on 01 Aug 2024 04:36 collapse

I might need an EIL5 for this. Is the fishy part that when you expand the vote percentages to 7 decimal places, both candidates are .19999 and then that gets rounded up to .2?

Skua@kbin.earth on 01 Aug 2024 06:34 collapse

0.1% of the vote with roughly 10,000,000 voters is 10,000 votes. So for each 0.1% result, there are 10,000 possible numbers of voters that would round to that.

For the sake of easy maths I'm going to briefly pretend that there were exactly 10,000,000 votes cast. To get 51.2% of the vote, I could get anywhere between 5,115,000 and 5,124,999 votes. What I got, though, was exactly 5,120,000 votes. The other two results look the same as well (although if the first two hit this requirement, the third necessarily must as well because it's the "other" category)

The seven decimal places thing is because there weren't exactly 10,000,000 votes cast, so there's not actually a number that produces exactly 51.2%. The author expanded the percentage out to show how close it is to 51.2% despite this - in fact, it is the closest possible whole number of votes

bamboo@lemmy.blahaj.zone on 01 Aug 2024 08:35 collapse

I see, so the fishy part is that the vote counts basically correspond exactly to the nearest 0.1% for the top two candidates, whereas that’s unlikely. And the odds of that happening for both candidates is a 1 in roughly (10,000^2)?

EDIT: rereading the article, I see the probability calculations at the bottom.