Over the course of 2018, Gallup polling found that President Trump was not very popular in Texas. Of the 30 states Trump won in 2016, nowhere was his approval rating lower than in the Lone Star State.

Trump’s job performance was viewed more negatively there than in Michigan, Pennsylvania or Wisconsin. He won those states by 78,000 votes, handing him an electoral college edge and the White House. If he’d lost Texas in 2016, though? That alone would have been enough for Hillary Clinton to prevail and head back to Washington.

New polling from Quinnipiac University offers a slightly better picture for Trump: His approval in Texas is now at 45 percent. That’s still not great, particularly given that half of Texans registered to vote view him with disapproval.

The eternal question — at least until next November — is how the near-universal mediocrity of Trump’s approval numbers will translate at the polls. To answer that, Quinnipiac was direct, evaluating if respondents to its poll would definitely or definitely not vote for Trump’s likely reelection bid.

Of those surveyed, 48 percent said that they would definitely not vote for Trump to have another four years. Thirty-five percent, barely over a third, said that they would definitely vote for him.

Democrats, expectedly, were overwhelmingly unified in opposing Trump’s reelection. Republicans were mostly supportive of voting for Trump again, though only about three-quarters said they would definitely support him. More than half of independents came down in opposition to a second Trump term.

If you’ve been paying attention to polling over the last two years (or, really, four years), the other splits above will look familiar. White Texans, like white Americans generally, are more favorably inclined toward Trump than black and Hispanic Texans. Whites without college degrees are more enthusiastic than those who’ve graduated college. Men view Trump more positively than women and plan to vote for him more heavily.

Notice one detail that’s not explicit above. Quinnipiac didn’t break out nonwhite men and women in the way that they did white men and women, but we can get a sense of how those groups feel by comparing the overall gender splits to the splits among whites. Only 30 percent of women plan to definitely vote for Trump — but among white women the figure is 46 percent. The implication? Nonwhite women overwhelmingly oppose his reelection.

The poll above was of registered voters in Texas. If we look at voter registration data in the state compiled by L2 Political, the data above seem particularly bleak. There are more Democrats who are active voters in Texas than there are Republicans at this point, according to L2′s analysis. (The state has nonpartisan registration.) That comports with past polling in the state and with its demographics: Texas is majority nonwhite.

If there are more Democrats than Republicans and Democrats oppose Trump’s election more strongly than Republicans support it, that’s game over, right? Barring some movement in the polls over the next (cough) 14 months, how can Trump win?

One answer is that “definitely won’t vote for” is not the same as “definitely will vote for his opponent.” One of the reasons that Trump is president right now is that a lot of voters disliked him in 2016 but also disliked Clinton. Some of those voters simply didn’t vote at all. (Those who did preferred Trump by 17 points.)

Or those voters might not vote simply because they are less frequent voters.

Turnout isn’t even across demographic groups. Among the demographic groups included in Quinnipiac’s poll, groups that are more hostile to Trump voted less heavily three years ago than those who more strongly support him. This isn’t specific to Texas or to 2016; lower-income and younger voters consistently vote less frequently. So do nonwhite voters.

We can see that effect in Texas. A caveat: The data below are among current, active Texas registrants, meaning that 2016 voters who may have died or moved aren’t included and new registrants since 2016 are included as non-voters.

Three-quarters of the Republicans currently on the voter rolls voted in 2016. Only a bit over half of Democrats did. Nearly 7 in 10 white women who are active voters cast a ballot in 2016, but, again, only about half of nonwhite women did.

Turnout among groups that historically vote more heavily Democratic is a central question for who wins the White House in 2020. If nonwhite women vote at the same rates as white women in the state, Trump’s in deep trouble. If, however, those who oppose Trump the most fervently either aren’t motivated to vote or aren’t able to — at times a function of state laws making it harder to vote — Trump can be expected to once again carry the state.

And, therefore, have a shot at four more years.