The Associated Press-NORC Center for Public Affairs Research released a survey on Sunday putting approval of the president's handling of immigration at 39 percent, down from 49 percent at the beginning of his second term [1].
Immigration is still his best issue. About 3 in 10 US adults, 32 percent, approve of how he is handling the economy, and about 3 in 10 approve of his handling of healthcare [1]. A ten-point fall on his strongest issue leaves it his strongest issue.
The finding that does not sit comfortably inside the administration's own framing is the one about legal immigration. About half of US adults say the president has "gone too far" when it comes to restricting legal immigration [1]. The enforcement program has been sold, consistently and by name, as a response to people in the country illegally. Half the country is now telling a pollster that the restrictions have reached the paperwork side of the system, the visas and the green cards and the applications that were never the stated target.
On deportations the question the pollster asked was about authority rather than desirability. Forty-seven percent say the president "has exceeded his authority when it comes to deporting immigrants living in the US illegally" [1]. That is a judgment about whether a process is lawful, not a judgment about whether the outcome is wanted, and the two can point in different directions in the same respondent.
The rest of the distribution is worth putting on the page, because a plurality finding means something different depending on what it is a plurality of. "In each case, about 3 in 10 say Trump's actions have been 'about right,' and closer to 2 in 10 say he hasn't gone far enough" [1]. About half, plus about three in ten, plus closer to two in ten, accounts for essentially the whole sample. The people saying he has gone too far are not a plurality assembled out of a large undecided block. They are a majority of the people who answered.
Two questions in the release move away from categories and toward specific tactics. Most Americans say he has gone too far on using traffic stops to find and arrest immigrants living in the US illegally, and on sending federal immigration agents into US cities [1].
The number that has not moved is the one that decides whether any of this becomes a political problem. About 8 in 10 members of the president's own party continue to approve of how he is handling immigration, a slight decline from about 9 in 10 [1]. Republicans who say he has "gone too far" on deporting immigrants are, in the release's phrasing, about 2 in 10 or fewer [1]. Republican approval on the economy is 69 percent, and Republican approval on healthcare is 71 percent, up from 59 percent last year [1].
That last figure deserves the concession it implies. Whatever is eroding in the national numbers, the president's coalition has gotten more satisfied with him on healthcare over the past year, not less. A poll that showed only decline would be a poll worth checking twice.
Here is what the write-up we fetched does not contain: a sample size, a margin of error, field dates, or a methodology statement. All four are absent. The version on the record for us is Al Jazeera's account of the release, credited to Al Jazeera Staff and AP, rather than the AP-NORC primary, which we have not fetched.
That absence sets a ceiling on how precisely any of these numbers can be repeated. Thirty-nine percent and 47 percent are printed as whole numbers, which means they can be quoted as whole numbers. "About half" and "about 3 in 10" are printed as approximations, which means they stay approximations. A ten-point drop from 49 to 39 has no confidence interval attached to it here, and nobody reading the write-up can say from the write-up whether that drop clears one.
This outlet spent yesterday on exactly that failure mode, tracking a poll sentence that shed three qualifiers in six and a half hours as it moved across platforms, until roughly one third of Democrats and Democratic-leaning independents describing themselves one way had become one in three Democrats identifying as something else. The compression happens on the way downstream. Printing the qualifiers the pollster printed is the whole of the defense against it.
Set beside the opinion numbers is a week of enforcement record this outlet has documented from primary data. ICE's own July 21 dataset and TRAC's count share a denominator of 65,765 people in detention, of whom 19,329 have a conviction and 26,189 have never been charged with anything. ICE arrests reached roughly 1,593 a day in July, the fastest pace in the agency's 23-year history. On July 30 the Seventh and Ninth Circuits each ruled 2 to 1 that immigrants arrested in the interior can seek bond hearings. California Secretary of State Shirley Weber says her office never gave the Department of Homeland Security the voter list DHS claims to have matched against a 250,000-noncitizen figure.
A poll records what people said. It does not record why they said it. Nothing in this release establishes that any respondent had the July 21 detention data, the arrest pace, the circuit rulings or the voter-roll dispute in mind, and no number here supports a causal claim in either direction. What can be said is narrower and still worth saying: the sharpest movement in the release is on legal immigration, which is the part of the system that the detention and deportation record does not describe.
One housekeeping distinction, because two polls landed in the same news cycle. The cost-of-living survey logged earlier today was CBS News, a different organization running a different instrument with a different question battery. Its toplines and these are not comparable, and stacking them into a single trend line would be an error rather than a synthesis.
The durable finding in the AP-NORC release is the floor rather than the fall. Approval on immigration is down ten points nationally and holding at roughly 8 in 10 inside the president's party. A national decline that leaves the coalition intact is a different political fact from a coalition coming apart, and this release, printed without a sample size or a margin of error, does not let anyone say which of the two is happening with any more precision than the word "about."
