Thread by @shylockh

Evidence Suggesting Voter Fraud in Michigan Senate Race

5 highly suspicious late-night vote updates in Wayne, Genesee, Berrien and Muskegon counties contributed 18.5K net Democrat votes, almost the entire Dem margin of victory. They look implausible on multiple dimensions.

A 🧵

The updates boost Democrat votes at the expense of Republican votes, pushing the limits of what might be considered credible to a casual observer. However, they leave six properties that are consistent with fraud, and are collectively very hard to explain.

2/N

The suspicious properties are:
1. Very high Democrat vote share (89%+)
2. Enormous increases in Democrat vote share relative to past votes in county (26% to 52%↑)
3. Increases in Democrat share are huge outliers relative to all nationwide vote updates (above 99.6th pctile)

3/N

4. Implausibly low ratios of Republican to third party votes (between 1.7:1 and 2:8.1)
5. Reductions in Republican to third party votes that are also huge national outliers.
6. Implausible Republican reductions relative to *all* third parties across the political spectrum

4/N

These updates are not the largest in their counties, but are consistent with “leaning for the tape”, pushing the limits to add as many votes as possible during the final counts in each county, when the race was close overall. Let’s go through the properties in detail.

5/N

Property #1: Very high Democrat two-party vote share (89%+), at the limit of credibility.

Wayne: 92.0% (12,645 votes total)
Genesee #1: 93.1% (1,024 votes total)
Genesee #2: 93.3% (3,454 votes total)
Berrien: 92.6% (2,677 votes total)
Muskegon: 89.3% (3240 votes total)

6/N

Property #2: Much higher Democrat vote shares than all previous votes in the county, by 26% to 52%.

Wayne: 26% ↑
Genesee #1: 39% ↑
Genesee #2: 40% ↑
Berrien: 52% ↑
Muskegon: 41% ↑

These updates look nothing like the Democrat share for other votes in the county.

7/N

Visually, this is apparent in the colored vote updates being both high at an absolute level, and much higher than the rest of the updates in the county

8/N

Wayne and Genesee have other large updates earlier in the count that also look suspicious, but not quite as extreme (these are some of the large spikes in the overall graph). The anomalous updates are huge outliers even relative to those prior ones.
9/N

Property #3: The increases in Democrat vote share are huge outliers relative to all nationwide vote updates in 2024.

-They’re above the 99.7th percentile or more of the distribution (*FN 1 at end)
-They’re between 4.0 and 6.3 standard deviations above the mean

10/N

We’re computing changes in two-party vote share in updates for House, Senate and Presidential races nationwide in 2024 – that is, the difference between two-party support in that batch, and two-party support up to that point. We have 15,218 updates of 2000 votes or more.

11/N

In percentiles of largest swings to either party, we have:

Wayne: 99.6th percentile
Genesee #1: 99.8th percentile
Genesee #2: 99.8th percentile
Berrien: 99.9th percentile
Muskegon: 99.8th percentile

12/N

In terms of how large they are as outliers, we have:

Wayne: 4.0 standard deviations above the mean
Genesee #1: 5.2 s.d. above the mean
Genesee #2: 5.51 s.d. above the mean
Berrien: 6.3 s.d. above the mean
Muskegon: 4.8 s.d. above the mean

13/N

The graph below shows how extreme these increases in Democrat two-party share are in the five updates, relative to vote updates nationwide in 2024. The graph details are a bit complicated (see footnote 2), but it shows clearly in visual terms how unusual the updates are.

14/N

Mostly they show that if you’re a county with high Democrat to Republican support, you’re generally that way across most updates. You don’t often observe single batches of 2000+ votes that are extreme outliers in either direction relative to the rest of the count.

15/N

Indeed, Michigan makes up 21 out of the 100 most extreme instances (out of 15,218) of large shifts in two-party vote share in a single update. Wayne has 5, Muskegon has 3, Genesee has 3, and Berrien has 3.

16/N

Other anomalous updates are present in Clinton (1), Macomb (3), Oakland (2), and Washtenaw (1). Overall, Michigan is massively overrepresented in terms of suspicious vote updates across the board, with the 5 current ones being some of the most dubious of all.

17/N

Property #4: These batches have implausibly low ratios of Republican to Third Party votes.
Ratios in these batches are between 1.7 to 1 and 2.8 to 1. Previous batches of votes from the same counties had ratios between 7 to 1 and 19 to1.

18/N

Like with the two-party Democrat support, these updates have Republican to Third Party ratios that are not only implausibly low in absolute levels, but also are huge reductions relative to all other vote updates in the same county

19/N

Property #5. Reductions this large in Republican to Third Party votes are almost unheard of in nationwide vote data.
-They are greater than the 99.6th percentile of House, Senate and Presidential vote updates
-They are between 2.4 and 3.9 standard deviations below the mean

20/N

For the 5 updates in Wayne, Genesee #1 and #2, Berrien and Muskegon, we have percentiles of 99.6th, 99.8th, 99.8th, 99.9th, and 99.8th respectively.

In terms of standard deviations above the mean, they are 2.4, 3.5, 3.5, 3.9 and 3.0 standard deviations above respectively.

21/N

Like before, it is visually apparent that these reductions in support are huge outliers relative to how national votes work in general

22/N

Property #6: Implausible reductions in Republican support are evident relative to *every* individual third party vote total across the political spectrum: Green, Libertarian, Natural Law Party, US Taxpayer. Republican totals are extremely low under every benchmark.

23/N

This includes fringe parties you’ve never even heard of. For instance, the Genesee #1 update had Republican votes at only 7.6x votes for the Natural Law Party. By comparison, in previous votes in Genesee, Republicans had been 124x the votes of the Natural Law Party.

24/N

The same Genesee #1 update also had Republicans at 4.3x the US Taxpayers Party. In previous updates, they had been 52x as high.

Having such large reductions relative to *all* third parties is very unlikely, both intuitively, and in national vote data.

25/N

This helps distinguish these 5 updates from other batches in counties like Washtenaw and Kalamazoo. They also have unusually high two-party vote shares relative to prior updates, but the Republican to third party ratios are more in line with other county updates.

26/N

These 5 updates added 18,551 net votes for the Democrats, between 1:57am and 11:30am. This is 92% of the final margin of victory (20,217 votes)

Wayne: 10,135 net votes
Genesee #1: 856 net votes
Genesee #2: 2,900 net votes
Berrien: 2,207 net votes
Muskegon: 2,435 net votes

27/N

It is worth noting that slightly less extreme other updates in Wayne and Genesee added far more net Democrat votes – 95,326 and 95,807 votes respectively in two updates in Wayne, and 9,759 votes in Genesee.

28/N

If one believes these 5 updates are potentially fraudulent, it is far from obvious that they are the full extent of the fraud, or the largest in magnitude. They are simply the cases at the end where the updates are most extreme, and thus the easiest to identify.

29/N

Like in Wisconsin, the strength of the case comes from the number of ways that the updates are implausible.
x.com/shylockh/statu…
shylockholmes.substack.com/p/evidence-sug…

It is very difficult to have any confidence in the integrity of the Michigan Senate Race.

/END

If you liked this thread, please retweet the first post.

x.com/shylockh/statu…

A sharable version outside X is available here
shylockholmes.substack.com/p/evidence-sug…

Footnotes

FN1

For 4 of the 5 updates (except Genesee #1), the sample over which comparisons are made is all vote updates of at least 2000 votes total in House, Senate and Presidential races nationwide in 2024. For Genesee #1, which is smaller in size, the same is all updates of 1000 votes total. Limiting to similarly sized batches is important especially for the third party comparisons, otherwise the addition of one or two votes to a particular candidate will have an outsized effect on the ratios of support.

FN2
For two-party vote shares, I compute the natural log (Democrat votes in batch / Republican votes in batch). Then I take the difference between this, and the same value across all previous vote updates combined in the county in question. This “different in log two-party support ratio” is the number over which percentiles are computed, and the z-scores (number of standard deviations above the mean).

Using the log transformation is mostly a way of scaling the values to make the differences apparent, but the 5 updates are also outliers in other transformations – raw changes in support (not logs), two party percentage instead of ratios.