Last week a couple of things happened that threw me back into the world of Kansas Election Fraud. First, the story became big news again, drawing even some national attention. Second, a shady anonymous internet source (not really, but anonymous nonetheless) provided me with data for the 2014 Kansas Governor's election. This data was previously unavailable (Beth Clarkson's original analysis was on a Senate race) and considered highly suspect (due to a widely-held belief by progressives that Brownback couldn't have won the election).
I dove into the data once again. Per my prior diaries I found that Clarkson's suspicious correlation was easily explainable by exogenous precinct demographic variables, suggesting that precincts are not "stochastic" in creation. Another important piece of work on this was HudsonValleyMarklooking at democrat registration rates, and essentially finding that fewer democrats register in the larger precincts, which once again explains the variance prior attributed to Clarkson's correlation. But why do the largest of large precincts have differing demographic attributes?
Here's what happened when I dug into the Kansas Governor's race data:
-First, I easily validated Clarkson's correlation. The correlation that Clarkson suggests may be due to election fraud, certainly existed in this data.
-Second, I found that only two counties are driving the correlation, the only two counties which have a large number of 500+ voter precincts. (Sedgwick and Johnson) If we look at the other counties, the correlation actually runs in the opposite direction.
-Third, diving into those two counties, I found that they do not have arbitrarily precinct patterns. Specifically, the largest of 500+ voter precincts (the ones driving the correlation) are those in the conservative suburban rings, while the 500+ voter precincts in the urban core are smaller. This demonstrates, in yet another way, that Clarkson's correlation can easily be explained by precinct design factors (in this case patterns of suburbanization and growth creating larger suburban precincts).
I also observed, to agree with Clarkson, that the system is not designed for easy numeric analysis or auditing.
For more information, my most recent full analysis can be found here including precinct analysis maps and additional information. Other analyses on election fraud available through links on my blog.