President Trump arrives to speak at the Ohio Republican Party State Dinner in Columbus, Ohio, in August. (Evan Vucci/AP)

For all of our focus on the fortunes of President Trump, the more urgent concern, with less than two months to go until Election Day, is the state of the midterm races. This isn't as neatly packaged as a metric, requiring exhaustive analysis and algorithmic activity, so we tend to default to consideration of the trend in the generic ballot question in polling. That's the one where poll respondents are asked whether they prefer the Democratic or Republican House candidate in their district. It may not tell us who will win every race, but it serves as a useful guide for how House elections will go.

On Wednesday morning, we got a new generic-ballot data point, courtesy of a poll from NPR and researchers at the Marist Institute for Public Opinion. In that survey, the Democrats have a 12-point advantage on the generic ballot. The Washington Post's average of recent generic-ballot results, including the NPR-Marist poll, has the Democrats up 10 points — enough to retake control of the House.

What's interesting in the NPR-Marist poll, though, is the trend that the pollsters have seen. NPR released a poll in July as well; since then, both approval numbers for Trump and the margin in the generic-ballot question have shifted against the Republicans.


(Philip Bump/The Washington Post)

These aren't huge changes, mind you. The change among registered voters on the approval question, for example, is within the margin of error. The particularly bad news comes when we look at how demographic groups have shifted.

On the approval question, many of the changes are subtle. A smallish drop in approval for Trump in the Midwest; a bigger one from people who live in small towns. Among rural and small-city residents, an increase in approval. Among white evangelical Protestants, a four-point increase in approval.


(Philip Bump/The Washington Post)

But again, the particularly interesting and important changes are on the generic ballot question.


(Philip Bump/The Washington Post)

First, the most dramatic number, that drop in the Midwest. That was the focus of NPR's article about the new poll.

"The Midwest is an area that is getting restless about what they hoped was going to occur and what they feel is not occurring,” Marist's Lee Miringoff said in that report.

There were some changes on the urban-rural split, too. Those changes, as in the approval changes, generally come with a larger margin of error because they generally rely on fewer respondents. That said, the shift among small-town residents since July moved them from preferring Republicans by 16 points to preferring Democrats by five.

In the next column are more important changes. Women are now seven points more supportive of Democratic candidates on net. Whites with college degrees shifted 15 points to the Democrats.

There's some overlap in these things. Republicans have been worried about white, college-educated women abandoning the party, with good reason. This is one poll, so we don't want to read too much into it, but it suggests that those women are continuing to move away from Republican candidates, and that there are shifts in a region that was seen as central to Trump's appeal in 2016.

All of that should make Trump nervous over the long term. He benefited from many Republicans voting for him despite their reservations in 2016, preferring him to Hillary Clinton. He may not be granted the same benefit of the doubt in a reelection bid against another candidate.

But there we go, bringing it back to Trump. More immediately, the NPR-Marist poll has bad news of a different sort. In two months, we will parse out how telling these numbers actually were — and quickly turn to an assessment of how those election results will affect the president.