Interesting Question: Couldn’t other factors explain the patterns in the data?

The answer to this question is “Yes, it’s possible”. I’ve seen several claims that people can explain the patterns in the data that I find concerning with voter registration of party and/or % turnout by party and/or other demographic variables. I haven’t yet done much analysis into these factors, primarily because I haven’t had access to precinct level data. I’m working on that and may have an acceptable database of those statistics ready soon.

At the county level, for the data I’ve looked at, such factors only explain a portion of the trend. Further, the pattern in the cum sum % analysis is not what would be expected by demographic type factors. We should see a leveling out after a rise due to those factors, not the continuous increase those models frequently show. If you check the Wisconsin graph, broken down by voting machine type, the pattern I would expect to see exhibited for that type of issue is shown by the purple line as opposed to the red lines (increasing) or the green line (paper ballots only).

2014 Wisc. Gov Race

However, even if all these patterns (which are consistent across different states and elections while showing similarities by voting machine type) are explainable in such a manner. At this point, I consider the fact that our voting machines in Sedgwick County have never been audited or checked for accuracy after an election to be equally concerning. Without such an audit, we do not know whether the votes have been counted accurately. IMO, that’s a serious problem and my lawsuit at least illuminates that issue even if I never get access to the paper audit trail.

Interesting Question #1: Why consider a small linear trend in the largest precincts suggestive of vote rigging?

I received the following question from a reader:

I have poked around in the Kansas Senate and Wisconsin Governor election results, which to my eye look the most dramatic, and I see that there does appear to be a Republican bias in the larger precincts. Not much, just a few percentage points, but the effect is real. What I don’t understand is what suggests that this might be sinister, as opposed to being just one of those oddities that crop up from time to time.

And thank you for looking into this. I have family in the Wichita area, and I want their votes to count.

Regards, Bob

Bob, thanks for caring enough to poke around the statistics. I appreciate the concurrence regarding a Republican bias in the larger precincts. You ask good questions.

Why should this small trend disturb us? Why isn’t it just considered a statistical oddity?

It’s not a statistical oddity because it crops up all over the U.S. Therefore, it’s a pattern, not an oddity or ‘one-off’ situation that isn’t expected to be repeated. It cannot be dismissed as an oddity; it is a pattern that has some explanation. Vote rigging is one such explanation, but this analysis cannot establish cause. An audit could establish cause and that is why I am trying to get access to the paper records.

Why consider it ‘sinister’ or evidence of voting machine manipulation? The evidence for this is more complex and not necessarily statistical, such as the documented vulnerability of the machines to being hacked. The most convincing piece of statistical evidence against the various voting machines is that if you separate out the different types, only some types of machines demonstrate this trend. When I separated out the Wisconsin Gov. Election results for paper ballots only, no voting machines involved, the results not only do not show this trend, but form a beautiful example of what is expected under the null hypothesis of no trend.