With the midterm election less than three months away, there is considerable agreement among forecasters on the Democrats’ chances of regaining a House majority: 73 percent (538), 70 percent (The Economist), 67 percent (PredictIt), 67 percent (Good Judgment) and 64 percent (Predictwise). Now, a new GW Politics Poll provides a window into the origins of this Democratic advantage and the challenges facing Republicans.

This poll — which is sponsored by George Washington University — is unusual because it is tracking the views of the same group of registered voters over time. In total, 3,150 were interviewed online by the survey firm YouGov in the first poll in mid-May. In this most recent poll, just over 2,700 of these voters were interviewed a second time, most of them from July 23-31.

Five important findings stand out.

The fundamentals haven’t changed, and they’re not great for Republicans.

Two key fundamentals of midterm elections are presidential approval and the generic ballot. In this era of nationalized congressional elections, people’s feelings about the president have become an even more potent predictor of how they will vote.

In this poll, 45 percent approved of the job Trump was doing, slightly higher than his current polling average. This was virtually unchanged from the May survey. And as has been true in most polls, those who strongly disapprove of Trump (46 percent of respondents) outnumber those who strongly approve (28 percent). This may be one reason why framing the generic ballot explicitly in terms of whether candidates support or oppose Trump widens the Democratic lead.

The Democrats’ advantage in the generic ballot for U.S. House also remained stable: 44 to 39 percent in the May poll and 45 to 38 percent now. This seven-point edge is very similar to the polling average.

The stability of people’s preferences between the two interviews was striking: 92 percent of those who said in May that they planned to vote for the Republican candidate said the same thing in July. Similarly, 95 percent of those saying they would vote for the Democrat in May were planning to vote for the Democrat in July. Altogether, over three-quarters of all respondents (78 percent) were “consistent partisans” in these two interviews.

Among the 13 percent who did not have a preference as of May, most (58 percent) remain undecided. About a third shifted to a major-party candidate: 15 percent now say they favor the Republican and 18 percent now favor the Democrat.

Democratic incumbents remain more popular than Republican incumbents.

Just as in the May interview, respondents tended to have more favorable views of Democratic incumbents on several dimensions — even if their overall view of all incumbents was fairly cynical.

More respondents said that Republican incumbents cared about their party, career and donors than said this of Democrats. But more said that Democrats cared about the country, the local community and “people like you,” compared with the percent who said this of Republicans.

These numbers have changed only a little bit since May. Some of those changes reflect less well on Republican incumbents. For example, there was an eight-point increase in the percentage of respondents who said that Republican incumbents care about “their party’s agenda” (69 percent to 77 percent). There was also a six-point increase in the percent who said that Republican incumbents care about their own priorities, and a nine-point increase in the percent who said Republican incumbents care about their campaign donors.

On the other hand, this survey found slightly smaller Democratic advantages on how much incumbents care about their community and people like you, as well as on overall approval of the incumbent. But nevertheless, Democratic incumbents retain advantages on most dimensions.

The economy isn’t providing much lift for Republicans.

Hope springs eternal that positive economic news will help Republican candidates. There is reason to be doubtful, however. For one, seat swings are more strongly predicted by presidential approval than economic trends. In other words, to the extent that the economy matters, it is via its impact on presidential approval. The problem, though, is that Trump is much less popular than he should be given how people feel about the economy.

The GW politics poll makes this clear. Of course, people’s views of the economy are strongly colored by partisanship: 73 percent of Republicans said that the economy is getting better but only 11 percent of Democrats believed this. But among true independents — those who don’t lean toward either party and do appear to rely more on their economic perceptions — Trump is facing a challenge.

Among independents, only 36 percent believe that the economy is getting better. A third (33 percent) say that it’s “about the same” and 25 percent say that it is getting worse. Among the majority of independents who say “the same” or “worse,” Trump is not popular: 75 percent disapprove of him.

Major achievements of the Trump administration aren’t very popular.

If the economy isn’t an issue that can help Republican candidates that much, then what issue can? The challenge for Republicans is that few Trump accomplishments are that popular.

In this poll, we asked about several policies closely identified with the Trump administration. In no case did a majority of registered voters support any of these policies:

More people opposed than supported the repeal of the individual mandate (46 percent oppose vs. 44 percent support), the Muslim ban (48 percent vs. 42 percent), the tax law (48 percent vs. 40 percent), tariffs on steel and aluminum (49 percent vs. 39 percent), and especially separating the families of immigrants arriving at the border (58 percent vs. 34 percent).

Moreover, even if views of the individual mandate are mixed, other aspects of the Affordable Care Act are much more popular. The GW politics poll asked about two provisions that provide “protections for people with preexisting health conditions.” Overall, 84 percent supported the “provision that prohibits health insurance companies from charging sick people more” and 86 percent supported the “provision that prohibits health insurance companies from denying coverage because of a person’s medical history.” Given that the Trump administration no longer supports the ACA’s protection for people with preexisting conditions, it risks alienating even more voters. Indeed, Democrats have already seized on this in the midterm campaign.

Democrats continue to be more engaged.

In the first GW politics poll, similar fractions of Democrats and Republicans said they were likely to vote. But Democrats were more engaged than Republicans in other respects. That is still true:

For example, more Democrats than Republicans reported sharing political opinions on social media, signing a petition, talking to someone about how to vote, donating to candidates and going to meetings or rallies of various kinds. (A Pew Research Center survey shows a similar pattern.)

Overall, then, this poll echoes other indicators, such as fundraising, that show a Democratic advantage heading into the fall. It is no surprise, then, that election handicappers believe more and more House races are shifting in the Democrats’ direction.

The question now is whether this Democratic momentum will continue and how much they will benefit on Nov. 6. In the meantime, look for the next GW politics poll in October.