Data Table of Vote Switches Against Trump from NYT Edison Data

posted 158 days ago by Beegie 158 days ago by Beegie +20 / -0

I'm making a new post incase my comment is missed in the sticked post found here: https://thedonald.win/p/11Q8O2wesk/happening-calling-every-pede-to-/

Here is the permlink to the comment in that thread if people want to kindly give some upvotes: https://thedonald.win/p/11Q8O2wesk/x/c/1Biviad4I4

I've made a data table, aggregated by state, summing up the total vote switch amounts between candidates.

EDIT for TL;DR -

Pennsylvania is looking real dodgy here. It looks like Trump may have lost at-least 212k votes. But a bunch of other usual suspects are popping up:

Michigan 20k votes
Georgia 17k votes
Virginia 12k votes

Plus some other "swing states" as they were reported by the fake polls.

This is using the same dataset from OP.

https://imgur.com/a/GBOZVj1

The csv of the full dataset:

https://pastebin.com/eZ4KuhJk

I'll be working on getting the code up into github. Need to make a private account. Unfortunately I work in a very left leaning industry as a scientist. I don't feel like getting fired because of some emotional folks.

EDIT: There are some data quality issues to potentially be aware of

During a timestamp, it's possible that a "switch" along with additional counts came in at the same time. This means for example that:

Trump lost more votes than reported, but gained some during that period. Or similarly, Biden gained some votes from a switch along with some additional counted votes.

The time granularity on the edison data stream isn't great, and I don't have full trust in the quality of that dataset unfortunately.

I'd use this more as a hint of where to look. Pennsylvania keeps popping up though. Smells like fish.

EDIT 2:

Here is the raw data highlighting the specific timestamps of the switch events:

https://pastebin.com/TCXZu9FA

Hopefully we can get some sleuths on the case.

EDIT 3:

I added one more additional preprocessing step that the original post this is based on. I removed all events where the vote count drops between time-steps. So we'll report slightly different numbers. I figured I'd try to remove as much 'noise' as possible so we can focus on the events where: