The city of Kalamazoo is heavily democrat, with minority areas particularly so. The most Republican areas are rural townships and some upscale lake areas. Among closer suburbs, Portage leans Republican and Oshtemo and Comstock are close to even.

Here is a map of the results in the state senate race by precinct.

Superficially, the results appear very similar to the other countywide elections.

We can measure candidates’ performance in precincts relative to the GOP average and their overall performance. Break down the precincts into four categories relative to the partisan average:

less than 95% (orange)
95-100% (yellow)
100-105% (light green)
more than 105% (green)

Here is O’Brien’s relative performance.

O’Brien overperfomed in Alamo, Brady, Charleston, Climax, Prairie Ronde, Texas, Wakeshma, and Portage

Here is McCann’s relative performance.

McCann overperfomed in the city of Kalamazoo and Kalamazoo Township.

Where did Wenke do best? His top precinct was the city of Galesburg (18.8%). Most of Wenke’s top precincts are in his old district (the eastern half of the county).

Wenke got 11.6% in the 63rd state house district.
Wenke got 8.4% in the 61st district.
Wenke got 7.6% in the 60th district.
Wenke got 9.1% in the 66th district.

Break his precincts in four categories:
less than 7% (orange)
7-9% (yellow)
9-11% (light green)
more than 11% (dark green)

The narrow result naturally raises the question of what effect Wenke had on the election. Short of asking all of Wenke’s voters, we can’t know for sure who they would have supported in a two-way race.

The conventional wisdom is that Wenke took votes from O’Brien. This is because Wenke used to be a Republican, he represented a mostly Republican area, and libertarians are usually closer to Republicans on the issues.

I examined the results precinct-by-precinct to see what could be learned. I calculated a GOP average by precinct based on the other five countywide races. Scaling this by the total votes O’Brien and McCann received, I projected how their votes should have been distributed. I compared this with Wenke’s performance in each precinct.

Consider Galesburg, Wenke’s top precinct (18.8%). O’Brien got 99.7% of her projected total. McCann got 85.9% of his projected total. This strongly implies that Wenke’s voters would have supported McCann.

There were 20 precincts where Wenke got more than 11%. O’Brien overperformed in 6 of them and underperformed in 12 of them (2 with less than 95% of her projected total.) McCann overperformed in 1 of them and underperformed in 19 of them (12 with less than 95% of his projected total.)

There were 20 precincts where Wenke got less than 7%. O’Brien overperformed in 7 of them and underperformed in 13 of them (9 with less than 95% of her projected total.) McCann overperformed in 18 of them (8 with more than 105% of his projected total.) and underperformed in 2 of them.

McCann underperformed in most precincts where Wenke got more than 11%, and overperformed in most precincts where Wenke got less than 7%. The same is not true for O’Brien. This strongly implies that Wenke took significantly more votes from McCann. Since O’Brien won by only 61 votes, Wenke cost McCann the election. Wenke appears to have done best with democrats in his former state house district.

Wenke absolute performance versus McCann relative performance

An analysis by Gongwer reached the same conclusion using poor reasoning. They correctly noted that McCann did worse than democrat state house candidates. But this does not prove that Wenke’s votes came from McCann, since he could just have been a weaker candidate (or O’Brien stronger). Using different races in different parts of the county is bad methodology, especially when two (60, 63) were uncompetitive and one (61) had a substantial Libertarian vote.

In the 61st state house district, the results were 48.4% Iden, 42.7% Fisher, 8.9% Stampfler. Notably, the correlation between Wenke and Stampfler in the 61st was only r=.032, implying that their supporters voted for them as individuals, not libertarians.