Do all paths to a Trump victory run through Michigan?

 I think so, but there are reasons to believe that is wrong.

Sean Trende has an electoral simulator not unlike my own. Of course, I cannot get into his code, so I’m not sure what the underlying assumptions are about demographic change in 2012-16. But here is what you get if you assume that African-American and white turnout will return to 2004 levels, while the GOP loses one-third of the Latino votes that Romney got in 2012.

Oh, wait, that is the same as the 2012 map. For the GOP to win, even under the above assumptions, you need to up its share of national white vote from 60% to 64%. His model shows Virginia flipping at 62%, Florida and  Ohio at 63%, and then Pennsylvania, New Hampshire and Wisconsin for the win at 64%.

I think the difference is that I assume (based on past elections) that the white vote in the Badger State is far stickier than those flighty white people over in Michigan.

In the Trende model, a 3-point uniform swing with high white turnout will swing Wisconsin but not Michigan. That does not seem unreasonable, but it gets you the rather odd electoral map on the right.

Of course, if you assume that Trump will lose half the Latino and Asian-American vote, as opposed to a third of the former and none of the latter, then all the above assumptions won’t help. Under the Trende model, New Hampshire, Ohio, and Virginia flip, but the Democratic candidate still wins.

I do not really believe a scenario in which Virginia flips that early despite a crash in the GOP share of the Latino and Asian-American vote, but Trende is paid to do this stuff and I am not. He is also very smart. And the general result seems to hold. For a GOP candidate to win while alienating minorities, they need an electoral path right through the old Rust Belt, be it Wisconsin or Michigan or Pennsylvania. 

It is interesting to run the table in the other direction: at what swing among white voters do Democrats start expanding the 2008 map? (Moving the needle on Latinos and Asians alone flips only North Carolina.) That map is regained at 43% of white voters nationally, a three-point swing over 2012. (In 2008, exit polls showed Barack Obama winning a solid majority of white votes outside the old Confederacy.) At 46%, Georgia and Missouri flip. Arizona goes blue at 47%. Texas joins them at 49%, along with Montana and the Dakotas. Finally, at a ten-point uniform swing among white voters Mississippi and South Carolina vote Democratic ... even though that means the Republican is still getting almost 80% of white voters in those states.

The sad thing, though, is realizing how unlikely it seems that any Democrat could get as high a vote share as 20% in those states. I find it much much easier to imagine the entire Rust Belt swinging red than the deep south going blue, let alone Texas. 

Comments

[J.H.]

I believe Sean Trende gives his methodology behind the code here: http://www.realclearpolitics.com/articles/2015/08/26/rcp_demographic_interactive_map_methodology__127889.html

So it should be possible to find the sources he used to make assumptions in the code.

[J.H.]

Incidentally if you get a non-hispanic white vote of 63.6% for the Republicans and a white turnout of 65.1% then you get Pennsylvania, Virginia, Florida, Ohio and Colorado for the Republicans but Wisconsin still remaining for the Democrats.

[Noel Maurer]

Thank you for the link! I'm looking through it now. I think my model shows a slightly larger shift in the population of eligible voters, but I'm still checking. Mine also doesn't correct for the difference in CPS numbers versus others, which sounds important, but the underlying results are the same.

The big difference is the assumption of uniform shifts among ethnic groups nationally.

[Doug M.]

The uniform shift is a pretty deep simplifying assumption. Voter turnout rates vary widely by state, so why would they all shift the same amount?

In particular, there are a few states -- Minnesota, Wisconsin, NH, Colorado, Iowa -- that already have crazy high voter turnout rates. Minnesota was an eye-popping 76% in 2012, by far the highest in the country, with participation well above average in every group -- black, white, male, female, young, old. The other four were all above 70%. That's against a national average of 58.6% in that cycle.

So it will probably be quite difficult to raise turnout in these states, because turnout is already very high. That applies both to turnout generally and to individual groups.

(Reality check: national turnout was about four percentage points higher in 2008. Yet there were still only five states with turnout over 70%. This suggests that national changes in turnout don't have a lot of effect on high-turnout states one way or another.)

At the other extreme, the states with under-50% turnout were TX, OK, WV and Hawaii. Make of that what you will.

Additional wrinkle: turnout in swing states tends to be already well elevated over the national average. So, Ohio, 65%; Florida, 64%; Virginia, 67%. This suggests that significantly raising turnout in these states will also be difficult, since it's already unnaturally elevated.

And finally: states with Election Day Registration (EDR) tend to have much higher turnout rates than states without. The states that had EDR in 2012? Idaho, Illinois, Iowa, Maine, Minnesota, Montana, New Hampshire, Wisconsin, Wyoming, and DC. Since then, four more states have added it -- California, Colorado, Connecticut, and Illinois. (Hawaii would be another, but it won't take effect until the 2018 cycle.)

An alert reader will have noticed that of the five states with 2012 turnout over 70%, four have EDR. This is, as Uncle Karl said, no coincidence.

So if you wanted to boost turnout in OH or WI, you could try enacting EDR. But that's not likely to happen between now and November. If anything, I'd watch for GOP legislatures in swing states trying to get rid of it. This was actually done in Maine four years ago, targeting the 2012 election. (It got put back by voter referendum.)

Meanwhile, it's clear that EDR boosts turnout by several percentage points on average. The exact amount is hard to determine, because several of the states with EDR had historically high turnout rates anyway. A straight-up comparison suggests 8%, but that's probably a bit high. 5% seems quite plausible, though.

So if you adjust the list of states for EDR, then suddenly Wisconsin jumps to first place! Even without EDR, it's already rocking an astonishing 73%. This suggests that, absent passing EDR, Wisconsin turnout is unlikely to rise by more than a couple of points.

And Michigan? Historically, Michigan has been one of the higher turnout non-EDR states. (It doesn't have EDR and isn't likely to adopt it soon.) Turnout in 2008 was 69.7%, falling to 65.4% in 2012 -- a drop almost exactly consistent with the national change. Put another way, you can reasonably expect Michigan turnout to be around the national average plus 6-8% or so.

Doug M.