In the Michigan Senate race John James is trying to unseat incumbent Debbie Stabenow.  The polls claim it is a long shot and that Mr. James does not really have a chance of winning, the RCP average has him down by 16.3.  I recently posted on some of the polls, which you can read about here and here.

Since those posts, I dug deeper into the most recent Mitchell Research poll where John James is within 9 points (8.8 to be exact).  A quick scan of the poll shows a 3.8% margin of error and 7.8% undecided.   Taking the poll at face value, John James could win, although it is still an uphill battle, however, I do not think it is as “uphill” as this and other polls suggest.  As a reminder, polls can be wrong…very wrong:

• 2016 Poll showing Michigan would go for Hillary
• Polls showed Pennsylvania going Democrat
• I have many more examples you can dig through on this site, or just go to www.realclearpolitics.com and look at any election and the polls before the election, chances are the polls were wrong

Accepting that the polls can be wrong, we can still look at the polls and see what is wrong with them and try to from an idea of what the mood of the electorate is.  Here is what I think is going on in the Michigan Senate race.

In 2016, Trump won Michigan by a little over 10,000 votes.  He received 47.6% vs Hillary getting 47.3%.  It was a close race!  Also note worthy is Gary Johnson got 3.6% (#NeverTrump) and Jill Stein got 1.1%.  These numbers matter, because that is an accurate assessment of who voters are in Michigan…it is not guessing, those are the certified results.

In the Mitchell Research poll one of the questions asked to respondents is “who did you vote for in 2016?”  Here is a comparison to the poll and what really happened in 2016:

You can see, people who voted for Trump were way under-represented and Stein was way over represented.  Looking at the cross-tabs of this poll you will see that 83% of the Trump voters support John James, 83.3% of the Stein voters support Stabenow, etc.  If we “re-weigh” the poll to match the 2016 election, we get the following:

This is a completely different picture than all the other polls suggest, a 5 point race (almost within the margin of error) and Stabenow is below 50%.