Buried in a New York Times article about Donald Trump's nascent (and likely doomed) efforts to convert Bernie Sanders voters to his cause is a quote from Sen. Jeff Flake (R-Ariz.).

"I think [Trump's] more likely to take Michigan than he is to take Arizona," Flake said, amplifying his colleague John McCain's recent comments about anger of Latino voters in the state at Trump's comments on immigration. Trump's play for Sanders voters depends on convincing the white men that have powered Sanders's success that Trump is the best route for them moving forward. Trump hopes to capitalize on the economic discontent and rejection of establishment politics that contributed to Sanders's wins in the Rust Belt, including his surprise victory in Michigan. Which seems to be what's driving Flake's thought process: heavily Latino Arizona is more likely to reject Trump and blue collar Michigan to embrace him.

But that's probably incorrect.

In the summer, shortly after Trump announced his candidacy, the Cook Political Report's David Wasserman projected the demographic composition of close 2012 states for November. Included in that list were Arizona — which Mitt Romney won by about nine points — and Michigan, which Barack Obama carried by about the same margin. In Arizona, Wasserman figures that the number of Latino voters will climb from 18 percent of the electorate in 2012 to 19.4 percent this year — as the number of non-college-educated whites drops from 39 percent to 35.1 percent. Both of those trends favor the Democrat in a general election match-up.

But the figures in Michigan aren't much different. The density of non-college-educated whites in the electorate there is expected to drop from 39.6 percent to 36.7 percent, as the small percentage of Latinos climbs. More importantly, the percentage of black voters in the electorate will remain about the same.

Why's that more important? Because black voters are much more heavily Democratic than Hispanic voters. In 2012, Hispanics nationally preferred Obama by 44 points, according to exit polls. Black voters preferred him by 87 points. It's likely that the absence of Obama on the ticket and the presence of Trump will shift those percentages to some degree, but it's hard to see them flipping. And since black voters will be nearly as much of the Michigan electorate as Hispanics are of Arizona's, that's a built-in bulwark for the Democratic nominee.

Trump's consistent argument is that he'll both draw more white voters to the polls and that those white voters will vote more heavily for him than they have for past Republicans. That's hard to evaluate, though we can largely dismiss the evidence he usually presents to that effect: Politico found that the big spike in primary voters isn't from new Republicans rushing to the polls to vote for Trump but were instead people who didn't usually vote in primaries but had voted in the general.

If Wasserman's estimates hold, the percentage of non-college-educated white voters in both states will be about the same, making Michigan no friendlier to Trump in that regard than Arizona.

For Michigan to go for Trump but Arizona not to, history would indeed have to flip inside out. In fact, since 1952, Arizona has always voted more Republican than Michigan — and often by a wide margin.

Even as white working class voters have shifted to the GOP, Michigan has stayed bluer than Arizona.

A new poll from Public Policy Polling released Tuesday shows Trump with a small lead in Flake's home state. The subtext to his quote, of course, was that this is a weird cycle in which demographic splits could be disrupted. That's probably true.

But given the demographics in each, the level of disruption it would take to flip Arizona for Democrats and Michigan for Trump is probably outside the capabilities of even Trump.