Power loves a problem it can stop measuring. An overdose surge is easier to miss after the emergency room survey ends. Violence against transgender people gets harder to prove when they vanish from a victimization survey. A police agency hiring an officer has less evidence after the misconduct database shuts down.
Donald Trump’s administration has spent the past year and a half weakening each of those records. A new Federal Data Terminations Tracker now lists 375 verified entries documenting federal data terminations and alterations, including 38 terminated data products. The Guardian’s investigation of the tracker found the damage spread across more than 60 agencies, with health data hit hardest.
This is how a government chooses blindness. The underlying harm keeps happening when an agency stops counting it. The public record gets weaker, giving officials and corporations more room to deny the harm.
The careful count is still damning
Federal datasets change as surveys become obsolete, agencies consolidate reports, and budgets tighten. Any honest accounting has to separate ordinary maintenance from a political purge.
Data Index’s published methodology excludes routine catalog churn and temporary takedowns that were restored. It counts products instead of inflating the total by treating every file as a separate dataset. Each termination has to trace back to primary evidence such as an agency notice, a regulatory filing, or a record that the product is no longer available. The count also changes when courts restore data or new evidence appears.
The administration’s defense rests on administrative discretion. Agencies have always retired data collections, and some information can still be gathered by states, universities, or private groups. The tracker accepts both points and removes ordinary discontinuations from its total. After those exclusions, the documented pattern exceeds ordinary housekeeping.
Missing records protect the institutions being measured
The Substance Abuse and Mental Health Services Administration ended new collection for the Drug Abuse Warning Network on June 13, 2025. The agency’s own notice says the decision aligned its work with administration priorities. DAWN tracked drug related emergency room visits. Those reports can reveal an overdose wave before the death toll tells public health officials what they missed.
The Agriculture Department ended the Agricultural Labor Survey, calling it duplicative or unnecessary. Months earlier, the department’s own methodology said the survey produced official estimates of farm employment, hours, and wages used by the Labor Department, state and local governments, researchers, farm organizations, and employers. Farm wages did not become irrelevant between May and August. The government abandoned one of the tools that made them visible.
Trump also rescinded the order behind the National Law Enforcement Accountability Database. The Congressional Research Service found that 94 federal agencies had participated. The database held 4,790 qualifying misconduct incidents involving 4,011 officers. Agencies searched it 9,985 times in its first eight months, producing 25 matches between agencies. Deactivating that system made it easier for misconduct to follow an officer without following the officer’s file.
These choices benefit the institutions being measured. Employers face fewer official wage estimates, while agencies lose a shared warning about officers with serious records. Political appointees face fewer public numbers showing whether their policies are hurting people.
Erasing people from the question
The largest cluster of alterations follows Trump’s order imposing his definition of sex across the federal government. The order directed agency heads to change forms and communications and report those changes to the White House.
The Williams Institute identified about 360 federal collections that removed at least one measure of sexual orientation or gender identity during the administration’s first year. Of those, 338 dropped gender identity measures and 60 dropped sexual orientation measures. Another 23 removed those measures from questions about why a person was targeted for bias. The institute found that 83% of the changes went through requests labeled nonsubstantive, which avoided the fuller public notice and comment process.
Those changes had large consequences despite the nonsubstantive label. The National Crime Victimization Survey measures crimes that never reach police. Researchers found that removing gender identity from it makes transgender respondents impossible to identify. The federal government can still count assaults. Its choice makes violence targeting transgender people harder to see inside those numbers.
A group excluded from the question is easier to exclude from the answer. Officials can demand proof of discrimination after ordering agencies to stop collecting the proof. That is the political effect of bad measurement.
The record is part of the government
A dataset guides the people treating overdoses and protecting workers. It shows where the harm is, who is carrying it, and whether a response is working. That is public infrastructure, even when it lives in a spreadsheet instead of under pavement.
Trump’s administration is weakening that infrastructure in ways that protect power from scrutiny. The tracker explains the different kinds of loss and the evidence behind each one. Together they show a government narrowing what the country is permitted to know about itself.
The damage keeps happening after the measurement stops, but the evidence gets thinner. Every official who would rather deny the harm than fix it gets the benefit.