SUITLAND, Md. — The Commerce Department, the parent agency of the Census Bureau and the Bureau of Economic Analysis, quietly removed language banning political interference from its scientific integrity policy, in an unannounced revision dated Aug. 19, NPR has learned. The change came the day after the Census Bureau issued a report on noncitizen voting with ties to a think tank founded by officials from President Trump’s first administration, which is the kind of coincidence that, in a different industry, would be called a conflict of interest, and in this one is apparently called a revision.
The removed language was specific. The 2025 version of the policy said it involved ensuring that scientific findings “are not suppressed, delayed, or altered for political purposes and are not subjected to inappropriate influence.” It also included bans on attempts to shape or interfere with data collection or statistical analysis without scientific justification. The revised policy, posted on the Commerce Department’s own website, adopts a similar definition of scientific integrity — but no longer says that protection from inappropriate influence is a “hallmark” of it. The hallmark is gone. The integrity, per the document, remains, standing alone, the way a house remains standing when the doors and windows are removed.
The timing is the detail the story keeps circling back to. A statistical agency, preparing for the 2030 census — the head count that will redistribute political representation and federal funding for the next decade — has its interference ban removed the day after its most controversial report of the year. The two events share a date the way a smoking gun and a fired bullet share a timeline. Former Census Bureau officials and data experts, per NPR, describe the change as the latest sign of an agency being weaponized, which is a word the agency’s own policy document, in its earlier draft, was designed to make unnecessary.
The scope is wider than the Census, which is the part worth underlining. The revised policy applies to every agency under Commerce, including the National Oceanic and Atmospheric Administration — the source of the weather forecast on the television — and the Bureau of Economic Analysis, the office that tracks gross domestic product. A single policy edit now governs the number that tells the country what the economy did last quarter and the number that tells you whether to carry an umbrella, and the edit’s central contribution was a deletion. The most consequential infrastructure update in American statistics this year is a missing sentence, which is the way a lot of institution history gets written.
Spokespeople for both the Commerce Department and the Census Bureau declined to respond to requests for comment, which is a fair response, in the context, the way silence is a fair response to a question about silence. The document, at least, was posted publicly, which is more than the change announced, which was nothing.
WHAT THE REVISION CHANGED
- Removed: the guarantee that findings "are not suppressed, delayed, or altered for political purposes"
- Removed: "protection from inappropriate influence" as a hallmark of scientific integrity
- Removed: bans on shaping or interfering with data collection without scientific justification
- Kept: a definition of scientific integrity that is, per the analysis, "similar"
- Timing: dated Aug. 19 — the day after the bureau's unusual noncitizen-voting report
The practical question — whether the numbers change — is the one nobody answers directly, because the answer is a process, not a fact. A statistic, once it is a statistic, is a thing that happens to a methodology, and methodologies, in the new policy environment, are documents rather than promises. The 2030 census is a decade away in the way that a deadline is a decade away: far enough to seem theoretical, close enough to be the entire point.
At press time, a data scientist who has worked with census methodology for twenty years said the revision “raises obvious questions,” and when asked what the questions were, said they were the kind of questions a scientific integrity policy, in its previous version, would have existed to prevent. The policy, per the previous version, still exists. It simply no longer contains the part about the questions.