Just over seven months into his presidency, Donald Trump continues to face important challenges. These include the investigation into possible collusion between his campaign and Russia, a struggling legislative agenda and Trump’s own controversial statements, especially regarding the violence perpetrated by white supremacists in Charlottesville. Meanwhile, Trump’s approval numbers have fallen steadily — to an average of roughly 38 percent.

Now, a new poll sheds important light on a muchdiscussed group that is increasingly unfavorable to Trump: voters who supported Barack Obama in 2012 and Trump in 2016.

This poll, the 2017 Views of the Electorate Research (VOTER) Survey, was sponsored by the Democracy Fund and designed by the Voter Study Group, a group of about two dozen analysts and scholars from across the political spectrum. (I am the research director of this group, and this initial analysis was conducted by Robert Griffin.) The 2017 poll was conducted by the firm YouGov between July 13 and 24.

The VOTER survey is unique because it has been reinterviewing respondents who were originally interviewed in December 2011 and November 2012. The first VOTER survey, conducted in December 2016, reinterviewed 8,000 respondents who had been interviewed four years before. In July, 5,000 of these 8,000 respondents were interviewed again.

Thus, this survey can track how respondents have changed not only between 2012 and 2016, but in the first six months of Trump’s presidency. And this is where those Obama-Trump voters stand out.

In the 2016 VOTER survey, about 9 percent of people who said in November 2012 that they had voted for Obama reported voting for Trump when interviewed in December 2016 (see my report for more). This new July 2017 survey shows that they are particularly dissatisfied with Trump.

First, we asked respondents whether they had any regrets about how they had voted in 2016. Most did not, including both Trump and Hillary Clinton voters. But the group in which the largest fraction (16 percent) expressed regret was Obama-Trump voters. That figure is about twice as large as it was in December 2016.

The same is true when respondents were asked whether they approved or disapproved of Trump. Most Clinton voters strongly disapproved of Trump. Most Trump voters approved of Trump, but were divided between strong and less strong approval.

Obama-Trump voters were significantly more likely than Trump voters as a whole to say that they disapproved: 22 percent did so.

Obama-Trump voters were also more uncertain about supporting the Republican congressional candidate in their district in the 2018 midterm election:

While most Clinton voters currently plan to stick with their party in the next House election, 20 percent of Trump voters are not sure. That climbs to 44 percent among Obama-Trump voters.

Of course, none of this means that Obama-Trump voters will continue on this trajectory. They may rally to Republican candidates in 2018 and perhaps to Trump himself if his administration’s challenges begin to fade.

But in the meantime, it’s clear that a crucial group of Trump’s supporters from 2016 are now less supportive.