I won’t be able to explain to you why (we can theorize, of course), but whiter counties were more predisposed to changing their votes from Democratic to something else. I put together the raw vote totals for Democrats for both the 2012 and 2016 elections and worked out the difference (Clinton vote minus Obama vote), and then used that as a percentage on Obama votes in 2012. In other words, how Clinton did compared to Obama by county in percentage terms. The counties are arbitrary borders for this sort of measurement but it does help us break down whether or not there was an effect. My dependent variable was the percentage I just explained and my independent variable was the percentage in every county in Michigan than is white, which ranged from 52.3 in Wayne County to 98.5 in Keweenaw County.
If you haven’t studied statistics, what this garbage in the table above means is that a) we can be confident our results are not arbitrary, and b) that the number on the bottom left, the “-0.20653” indicates that for every percentage point whiter a county is, Clinton lost 0.2% more of Obama’s share of the vote.
I didn’t control for anything because I spent a total of one afternoon putting this together, and I have known how to do this sort of work for something like 6 months, but found significance in my results. Here is this idea graphed (the charts are the same – one just has the county names).
This doesn’t explain too much – from this data I can’t tell you why counties that were happy to vote for Obama left Clinton for Trump or 3rd parties, but my results suggest that the whiter the county is, the more people changed their vote from Democrat to either Republican or 3rd party.
At some point I will do this for other states and look for data to back up a theoretical explanation. Don’t read too much into this. It just means there is something there. I cannot empirically tell you what it is. For better results it might be worth looking at party registration by county (if Michigan has that) and income, and how these counties have voted in other elections, and if anyone can come up with a measure of sexism, or white grievance, in a variable of numbers that would be just dandy.
Here is the chart for Pennsylvania. I found significance here too, and the result is more severe. For every 1-point increase in the white population, Clinton lost 0.7% in relation to Obama’s vote total in 2012.
Again, this doesn’t say why. Just that there does seem to be a relationship. As I say above, more data would be useful about each county, but we have enough here to begin theorizing. I would imagine Wisconsin looks similar.