Dominion Election Machine Hypothesis Test, Setup Prior to 2022 Election

As many know, I wrote a paper about the 2020 election which tested whether there was a correlation between Dominion Voting Machines and the 2020 presidential election. (This paper can be seen here:

The paper concluded that there was a correlation between counties that had implemented Dominion and the change in election outcomes comparing 2008 vs. 2020. I made no attempt to assess the cause of this statistical correlation. The test outcome could have occurred randomly. If it was not random, the reasons behind the correlation could have been unrelated to the voting machines, or if they were related to the voting machines it could have been related to some hacking technique unknown to election and Dominion company officials.

At the conclusion of the paper I called for audits (not recounts) to assess what happened. The only place such an audit occurred was in Maricopa County, Arizona.

As we approach the 2022 election, a common question asked of me is what I would expect to happen in 2022 if there is actually a persisent correlation between Dominion and positive results for Democratic candidates.

From 10,000 ft. up, I would expect if the Dominion machines are somehow influencing elections that the Republican candidate for governor in Arizona, Kari Lake, to lose. Right now Kari Lake is winning in the Real Clear Politics average of polls. One of her main campaign issues is she wants to change election processes. It seems apparent that Lake would do everything in her power to remove Dominion from Arizona, so it conversely seems that if there is some control of the election emanating from Dominion machines, that those forces would defeat Kari Lake.

Compounding the Arizona situation, Lake’s direct opponent, Katie Hobbs, weirdly and unethically, as acting secretary of state, is overseeing her own election.

If Lake wins, this would be evidence that Dominion machines are not controlling elections. Even after winning, I suspect Lake will remove Dominion machines from Arizona. 

But if Lake loses, I expect the people of Arizona to be quite upset. And, if that were to happen, simply saying “Kari Lake lost, so there must be fraud” will not be enough. There needs to be more evidence that something went wrong. So, I have thought about this in advance and I wish to share what I plan to study after the elections close on Tuesday (I am typing this on Sunday 11/6).

I wish to test the difference between Real Clear Politics averages versus final results. There should be a national swing that happens, either for Democrats or for Republicans, but this national swing should be independent of which machines a state uses.

I took Real Clear Politics Averages over the weekend for the governor and the senate and they look like this:

State Election RCP Avg 11-5-22 Dem. Advantage
Maine Governor 6
Oklahoma Governor -1
Oregon Governor 0
Texas Governor -9.2
Kansas Governor 3
Florida Governor -11.5
Minnesota Governor 4.3
Pennsylvania Governor 10.7
Wisconsin Governor -0.2
Michigan Governor 4.4
New York Governor 6.2
Arizona Governor -1.8
Colorado Governor 11.7
Nevada Governor -1.8
New Mexico Governor 4
Connecticut Senate 11
New Hampshire Senate 0.7
North Carolina Senate -5.2
Washington Senate 3
Florida Senate -7.5
Ohio Senate -6.7
Pennsylvania Senate -0.1
Wisconsin Senate -2.8
Arizona Senate 1
Colorado Senate 5.3
Nevada Senate -2.4

For these locations, I know the percentage of the state that appears to use Dominion machines.

State Election RCP Avg 11-5-22 Dem. Advantage % Dominion
Maine Governor 6 0
Oklahoma Governor -1 0
Oregon Governor 0 0
Texas Governor -9.2 0
Kansas Governor 3 5
Florida Governor -11.5 8
Minnesota Governor 4.3 14
Pennsylvania Governor 10.7 19
Wisconsin Governor -0.2 31
Michigan Governor 4.4 54
New York Governor 6.2 55
Arizona Governor -1.8 60
Colorado Governor 11.7 92
Nevada Governor -1.8 98
New Mexico Governor 4 100
Connecticut Senate 11 0
New Hampshire Senate 0.7 0
North Carolina Senate -5.2 0
Washington Senate 3 1
Florida Senate -7.5 8
Ohio Senate -6.7 13
Pennsylvania Senate -0.1 19
Wisconsin Senate -2.8 31
Arizona Senate 1 60
Colorado Senate 5.3 92
Nevada Senate -2.4 98

I intend to complete the analysis after the election. What I mean by that is I will take the difference between the predicted vs the actual and run a regression that includes the Dominion percentage as a potential influencer. If a positive coefficient for Dominion emerges that is statistically significant, this would be further evidence of a connection between the machines and election outcomes.

Of course there is always true randomness. This means that whatever such a model produces only provides evidence but does not provide conclusive proof. Random events are just that, random.

I have attached the spreadsheet I intend to use for this test. By publishing this spreadsheet prior to the election I hope to avoid accusations of post hoc rationalization.

Right now this spreadsheet allows the user to simulate the election. However, after the election, the simulated election values will be replaced with actual values and the statistical test result will be produced. Please note, I have excluded Georgia, because of a complicated relationship with their incombent governor and Dominion machines, and Kansas and Maine because the polling data is too small. All other races were included that were available from Real Clear Politics when I wrote this.

I’d like to make one other point. I really hope the election results are such that our society can be comfortable with the outcome and we can move past the idea that the machines are somehow influencing results.