Press "Enter" to skip to content

Timeseries Precinct Data files for MI, PA, NC, GA. Need help with data analysis!

posted by MiChecker

Due to the efforts of several people on the .win, we can announce that Precinct Level Time Series data now collected from NYT. This data is similar to Edison data, but without percents and much more detailed. All analysis that is relevant to Edison can be done to the precinct data.

We just finished collecting this data as of a few days ago from queries to NYT. It is unclear to us how many others have it. I did see that Edward Solomon seemed to be using the PA precinct data, so it may be out there already.

It’s precinct election results from NYT, not the general election results people have been analyzing mostly.

The information in this post is a summary of the more relevant posts in the other thread. We want to raise awareness to fellow data analysts that this is a very important data set and can be used to pinpoint the broad issues down into the precincts and individual voting locations.

Actions you can take:

  1. Download the data.
  2. Analyze it like you did Edison state data.


The full downloaded dataset, with helper scripts.

Original download links, in case you want to download it yourself.



North Carolina:


Analysis scripts:

Converts JSON files to CSV. You can do some analysis with Excel, but for full data loads you’ll need a database.

Current Analysis


The GA precinct data is probably the best dataset. The total tabulations from the precinct json data matches up perfectly with the GA secretary of state official results, which means we actually have a very accurate period-to-period or timeseries change in votes in GA at the precinct level.


MI data from edison v.s. the precinct data does line up perfectly, off by 674 milliseconds with the count off by 30,273 out of a ~660k vote – this is for 2020-11-04T01:56:48.

First thing’s first, the screenshot below identifies all the precint level data in MI where a decline in votes occurred from one period to the next, which is discussed in the following paragraphs. Nothing conclusive, just documenting for now.

Here, Ingham county loses 96,828 votes at 2020-11-04T08:10:57:961Z (or 3:10 am est time on election day). Could be a data glitch, not sure. TODO Look to see if vote count ever comes back.

Kent County goes from 257,660 votes to 0 at 2020-11-04-T08:49:01.772Z, staying this way through 6 data updates until 2020-11-04T10:49:26.786Z when the vote sum pops back up to 281,775. Trump gains 9,913 and Biden gains 14,635 from the previous count before it went to 0. Equivalently, this took place from 3:49 am est to 5:49 est on election night.

Oakland county sees a decline of 6,649 votes at 2020-11-05T20:52:37.622Z or 3:52 pm on NOV5, two days after the election.

There is a big jump in Biden votes at 9:52 pm est seen in the graph below. Specifically, 2020-11-04T02:52:41.280Z for the corresponding UTC time.

Votes increase from 315,937 to 382,071. Of that 66,134 vote delta, 17,701 (26.7%) votes go to Trump and 47,614 (72%) go to Biden while the remaining 1,449 go to 3rd party. So about 4:1 ratio of Biden votes to Trump votes were counted/submitted within 4 minutes of the prior data update (which itself had a 0 vote delta/change).

So they pushed a vote data update at 2020-11-03 21:48:45 with 0 new votes and then 4 minutes later at 2020-11-04T02:52:41.280Z they pushed another data update but this time with 66,134 newly tabulated votes.

In other words, this county found or counted (really, USB injected) 66,134 new votes in a span of 4 minutes.

2020-11-04T02:48:45.922Z — 9:48 pm – no vote sum change from previous 2020-11-04T02:52:41.280Z — 9:52 pm – 66,134 vote increase This seems like a pretty big skew in terms of ratio, and that big jump essentially narrowed the gap sufficiently enough that Trump loses his lead shortly after.

Row Data


Involved Precincts

This is the list of precincts in Oakland county with precinct vote deltas from previous period that makes up the 66,134 vote increase discussed above. The heavy hitter seems to be isolated to precincts in Southfield.


Here is Southfield in relation to the outline of Oakland county.

Here is Oakland county with precinct level shapes. This map is showing 2016 presidential election.

PENNSYLVANIA: TODO – the screenshot at the bottom of the markdown file



Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.