There’s been a lot of attention paid to Donald Trump’s appeal to a particular type of voter: white working class, no college degree, not that religious or socially conservative but anti-immigration. Let’s look at a few exit poll numbers to contemplate how a Trump coalition might be shaped very differently from Mitt Romney’s coalition, which drew together a respectable but insufficient 47% of the general electorate.

I did some simple algebra combining the share of each group in the electorate and the share won by each candidate, to consider what chunk of their voters fell in each group. For example, college graduates were 45% of the 2012 general electorate and Romney won 48% of them, so whereas non-college-grads were 53% and Romney won 47% of them – thus college-educated Romney voters were 23% of all voters, non-college-educated Romney voters were 25% of all voters, and accordingly college-educated voters made up 48% of the Romney vote. For purposes of this exercise I looked back at the Trump coalition in three states that were decisive (Indiana, Florida and South Carolina). While primary and general election coalitions are different animals, this is the data we have to work with so far, and it gives us a clue as to some of Trump’s challenges ahead, as well as how a candidate with Trump’s appeal to such groups could be an electoral force if that candidate wasn’t also as off-putting as Trump is to other core elements of the Romney coalition.

College graduates

As noted above, almost half of the Romney voters in 2012 (48%) had college degrees. Trump matched that figure in Florida,, but in both Indiana and South Carolina, just 42% of his voters had college degrees. The Romney-Obama race actually did not feature a big educational divide, at least around the college/non-college fissure. The GOP primary has featured a larger one. Trump’s ability to win educated voters is one of the major question marks for his fall campaign.

Women

49% of the Romney voters in 2012 were women, despite the partisan gender gap and reflecting the fact that Romney did better with white women than any candidate since Reagan in 1984. Trump’s primary coalition has been very male even by GOP primary standards – 41% of his voters in Indiana were women, 44% in Florida, 44% in South Carolina.

Religion

We don’t have applies-to-apples comparisons for all three states, but 52% of the Romney voters in 2012 were at least weekly churchgoers, compared to 45% of Trump’s voters in Indiana. 74% of Trump’s South Carolina voters said shared religious convictions were important to them. 60% of Romney’s voters said abortion should be illegal, a large contingent Trump will have a lot of trouble replacing if he continues to manifest signs of being a lifelong supporter of legal abortion.

71% of Romney’s voters were married.

Immigration

Exit polls often ask a binary yes/no question about whether illegal immigrants should be deported or allowed to stay in the country, a useful question for isolating the true hard-liners on the issue but a more fraught one for capturing the nuances of people who would be willing to entertain “amnesty” but only under fairly stringent conditions. And admittedly, attitudes may have hardened a good deal on the issue since 2012. Anyway, despite running mostly as an immigration hardliner in 2008 and 2012, 54% of Romney’s voters in 2012 were on the amnesty side of the amnesty/deportation question, compared to 45% of Trump’s voters in Indiana and Florida and just 36% in South Carolina.

Size of government

Do Trump’s voters agree with his pro-big-government stances? A whopping 84% of Romney’s 2012 voters told exit pollsters that government does too much, when offered the alternative of “it should do more.” We haven’t seen that question in a lot of primary polling, but it suggests another warning sign. Interestingly, 35% of Trump’s Florida voters said that we need to cut Social Security benefits – even though Trump is running on a platform of dissent from the GOP on that issue, he actually did better with those voters than with those who want the status quo.

Income

Lack of education may be a hallmark of Trump’s coalition, but lack of money is not as much one as you might have guessed. 67% of Romney’s voters made over $50,000 a year, compared to 74% of Trump’s voters in Indiana, 66% in Florida, and 71% in South Carolina. 32% of Romney’s voters made over $100,000 a year, matched by 32% of Trump’s South Carolina voters and exceeded by Trump in Indiana and Florida (36% each). (85% of Romney’s voters were not part of a union household; 62% worked full time).

Of course, we’ve covered previously Trump’s issues with self-identified “Very Conservative” voters, the heartland of the Ted Cruz primary vote. But demographically, Trump’s primary coalition does look less educated, less churchgoing and more male than the Romney 2012 coalition.  Obviously, the optimist’s theory (for Trump, and for down-ticket Republicans to the extent that Trump voters might also vote for actual Republicans for the House, Senate and Governorships) is that Trump’s coalition looks different because he is adding new groups.  But the danger sign is that just to pull even with what Romney did, he may still need more educated, religious, conservative voters and women than he has thus far been interested in appealing to.