Comment by tptacek
1 year ago
I can't see what this has to do with the analysis. Obviously, the counts themselves are subject to error, but it takes even more manipulation of vote counting to express counts that work out to round percentages, given error, not less.
Meanwhile: it is definitely not the case that the norm is to simply make up the vote counts, or to back them out from percentages.
You can certainly tell stories about how this publication isn't a smoking gun for the election, but they will all involve some deliberate manipulation of the numbers; for instance, you can try to rationalize a story where someone took the original vote counts, worked the percentages out, rounded them (so far so good), then discarded the original counts (uhhhhh) and synthesized new ones (triple yikes).
(I'm sorry to keep hedging this way, but I have to say again: cards on the table I think Maduro is awful, but my interest in this story is in the statistical anomaly, not the politics --- this for me is like that time HN got all wound up believing Google was dragnetting searches for pressure cookers after the Marathon bombing, which flunked plausibility due to a similar statistical argument; I like a pressure cooker, sure, but my real thing there was the base rate fallacy).
> You can certainly tell stories about how this publication isn't a smoking gun for the election, but they will all involve some deliberate manipulation of the numbers...
Not necessarily. The Occam's razor scenario if we saw this in a well organised country would be:
30 polling places report "We've counted <V> votes, each candidate has scored <X%>". The NEC works out the overall percent for each candidate appropriately by weighting and sums the votes. Then, stupidly, back-calculates the number of votes for each candidate from those two figures.
That'd be a concerning level of incompetence in an office dedicated to counting votes, but it isn't a huge problem (especially in a provisional number - it isn't exactly wrong as much as needlessly suspicious). Apart from the fact that nobody thought to report raw totals by candidate which should be their first instinct. But it isn't manipulation as much as incompetence. The % would be accurate and the numbers indicative.
Just so we're clear that at this point in the conversation I'm not all that wedded to any specific scenario, but:
Can you say more about why the CNE would back-calculate the raw vote counts for each candidate? Why would you even have that column in your spreadsheet? Keep in mind, you needed to keep the original raw counts around to compute percentages to begin with.
> Can you say more about why the CNE would back-calculate the raw vote counts for each candidate?
I can't see a reason why they would do that - but the world is large and maybe there is. It is clear someone has done something stupid here, so we've ruled out a highly competent electoral office.
So if I go with a creative scenario - some new grad gets the job of aggregating the provisional stats because how can they get that wrong. They have a template for the final vote, but only summary data provided by the polling stations. They have a teachable moment and realise that they can fill out the final template with what they have based on a few calculations. That gets reported officially without anyone senior asking where the vote totals came from.
Or it could be fraud. Also an easy option.