The Defense Department’s public casualty page showed 18 Americans killed in the Iran war on Wednesday, and 14 on Thursday, after four soldiers who died in the past week quietly disappeared from it.
That edit landed in the same week Trump’s job approval hit 34 percent in Pew Research Center polling, and the two facts are not a coincidence so much as the same instinct pointed at two different numbers.
The Four Names That Went Missing
The soldiers removed from the Defense Casualty Analysis System are 1st Lt. Tyler James Feehan, 25, Sgt. Angel S. Rampersad, 28, and Pvt. Isabella Gonzales, 19, all killed on July 17 when an Iranian missile struck Muwaffaq Salti Air Base in Jordan, along with Sgt. Michael Emmanuel Swinton, 30, who died on Sunday while dismantling a downed Iranian attack drone at Erbil Air Base in Iraq.
The New York Times reported the deletion first, citing three military officials who said the four were pulled because their deaths came after the ceasefire Trump declared in April. CNBC confirmed the revised toll on Friday. Acting Pentagon press secretary Joel Valdez blamed “temporary data disruptions” and called the story a non-story, a formulation that manages to deny the change and dismiss its significance in the same breath.
Here is the part that makes the denial hard to hold. Earlier on the same day the list was edited, Trump himself used the higher figure in public, telling reporters “It’s 18” before pivoting to a comparison with Vietnam that overstated that war’s American death toll by roughly a factor of three. The president was working from 18. By the next morning the government’s own database said 14. Nobody has explained who decided that a soldier killed by an Iranian missile in July is not a casualty of a war the Pentagon is still, at this moment, fighting.
Why the Ceasefire Framing Is the Whole Story
The reported rationale is the tell. If the four died “after the ceasefire,” they are not Operation Epic Fury casualties, and if they are not Epic Fury casualties, the official cost of the president’s war stops climbing at a politically convenient number. The problem is that the ceasefire in question is a fiction the administration has already abandoned. Trump declared the truce over on July 7. In the week that followed, U.S. forces struck roughly 140 targets. The Navy reimposed its blockade on Iranian ports on July 14, and Iran suspended its commitments under the June memorandum of understanding on July 18, one day after the missile strike in Jordan that killed three of the four soldiers now missing from the list.
So the administration is running a war it says is over, and it is using the “over” to keep the dead off the ledger. Rep. Adam Smith of Washington said this week that Congress has received no real casualty accounting from the Pentagon at all. We have been tracking this gap since the Pentagon’s injured-troop numbers stopped matching its own casualty system earlier this month, and the pattern has been consistent: the harder a number is to verify independently, the more favorable it gets.
Casualty counts are not a communications asset. They are the one federal record whose accuracy the families of the dead can check personally, and four families checked it this week and found their child’s name gone.
The Approval Number Is the One He Cannot Edit
Everything else on the president’s dashboard is administered by people who work for him. Polling is not. That is what makes this week’s numbers so difficult for a White House that has spent six months managing perception instead of prices.
Pew put Trump at 34 percent approval and 64 percent disapproval in a survey of 3,554 adults conducted July 6 to 12, with Democrats leading the generic congressional ballot 43 to 37 among registered voters. An Economist/YouGov poll conducted July 3 to 6 found 35 percent approval against 61 percent disapproval, a net of minus 26 that matches the lowest reading YouGov has recorded across both Trump terms and Biden’s. Emerson College, polling likely voters on July 19 and 20, had him at 39 percent. Forbes counted the through-line on Thursday: 42 percent of voters say they intend to use their midterm ballot against the president, against 22 percent who plan to vote for him.
The economy is doing the damage. CNBC’s survey put his economic approval at 38 percent against 60 disapproval, the worst mark of his political career on the issue he was hired to fix, with 61 percent describing themselves as pessimistic about the economy, the highest share since December 2023. Pew found voters naming the economy their top issue by a wide margin, with cost of living cited specifically by 15 percent.
Calling Affordability a Fake Word Is a Strategy, Not a Gaffe
At a rally in Marietta, Georgia on Wednesday, Trump told the crowd that “affordability” was a word “made up by the Democrats,” one he had never encountered before his first press conference of the term. Treat that as the rhetorical version of the casualty edit. The number is bad, so the category is illegitimate.
It is also, in its way, an admission. A president who believed grocery prices were falling would cite grocery prices. Instead he has attacked the vocabulary, which is what you do when the underlying data has stopped cooperating. He also promised oil prices would come tumbling down; crude has since crossed $100 a barrel as the Iran campaign squeezes supply, and American drivers are paying four dollars a gallon for a war the White House says ended in April.
The Tariff Round Is the Same Move in Legal Form
On Thursday the administration imposed new duties of 10 to 12.5 percent on 60 trading partners covering 99.4 percent of U.S. imports, effective at 12:01 a.m. Friday, the precise moment the previous temporary duties expired. CNN detailed the scope, which sweeps in Canada, Mexico, India and the United Kingdom.
The legal architecture is the interesting part. The Supreme Court struck down Trump’s original tariffs in February, with Chief Justice Roberts writing that the International Emergency Economic Powers Act does not authorize the president to set tariffs at all. Rather than accept that ruling as a limit, the administration opened a Section 301 investigation in March into whether those same 60 economies adequately police forced labor in their supply chains, and Thursday’s tariffs are the output. The policy did not change. The statute cited for it did. The Peterson Institute for International Economics judged the new theory unlikely to survive a court challenge, which suggests the point was never durability. It was continuity through the appeal.
Notice which goods got exemptions: fertilizer, some fuels, foods, autos, metals, pharmaceuticals. Those are the categories where a price spike shows up in a headline within a week. The administration knows exactly which numbers voters feel.
What Breaks First
There is a version of this presidency that survives a bad summer. Approval recovers, the war winds down, prices settle, and July 2026 reads as a trough. That version requires the underlying conditions to improve, and nothing this week suggests anyone in the West Wing is working on that problem. They are working on the readouts.
The trouble with governing by measurement control is that it has a floor. You can rewrite a casualty database, relabel a tariff statute, and declare a word illegitimate. You cannot make a grocery bill smaller by disputing the adjective attached to it, and you cannot un-notify four families. Sooner or later the last unmanaged number, the one taken by pollsters who do not report to the president, becomes the only one anybody believes. That is roughly where 34 percent comes from.
The four names were still absent from the Defense Casualty Analysis System as of Friday. Whether they come back, and whether anyone in the building is willing to say who removed them, is a more useful test of this administration than any poll released between now and November.