Nearly 400 federal data collections have been deleted or altered since January 2025, according to a new public tracker built by a group of data policy experts, researchers and developers. One of the casualties is a federal database of law enforcement officer misconduct records.
The tracker, published at dataindex.us, counts 28 datasets removed outright and another 338 modified. Together they span more than 60 federal agencies and cover the environment, labor, law enforcement, education and public health.
What the Tracker Found
Federal statistical collections are the plumbing of American public life. They are not glamorous, but almost everything downstream depends on them. School funding formulas, hospital staffing models, insurance rates, environmental permits and criminal justice research all rest on numbers the government gathers and publishes on a regular schedule.
That regularity is what makes the data valuable. A single year of measurements tells you very little. Twenty consecutive years of the same measurement, collected the same way, tells you whether something is getting better or worse. When a collection stops or its definitions change midstream, the comparison breaks.
Health data absorbed the largest share of the changes, making up roughly 40 percent of everything on the tracker. That is the category with the widest downstream use, from state health departments to university epidemiology programs to the private insurers who price risk.
The Datasets That Vanished
Three removals in particular have drawn attention from researchers.
The first is a database of federal law enforcement officer misconduct records, a centralized accounting of disciplinary findings against federal officers. Police accountability researchers had used it to study patterns across agencies rather than one department at a time.
The second is a long-running survey measuring violence and abuse experienced by children and teenagers. It was one of the few national instruments that asked young people directly rather than relying on cases that had already reached police or child welfare agencies.
The third is a set of real-time pollution readings collected by air quality monitors at United States embassies around the world. In a number of countries those American embassy monitors were the most credible independent air quality figures available to the local public.
Separately, several federal health surveys stopped collecting information on gender identity, following an executive order directing agencies to recognize sex through a strictly biological framework. That change does not delete a dataset outright, but it does end a series and prevents future comparison with the years that came before.
The Scramble to Archive
Researchers who depend on federal numbers describe the past 18 months as a race against the clock. Volunteer archiving efforts sprang up to copy datasets onto independent servers before they came down, often working from rumor and short notice.
The tracker itself grew out of that work. Rather than simply saving files, its authors set out to document precisely what changed and when, so that anyone using an old citation can tell whether the underlying data still exists in the form it did.
Two Very Different Readings
The administration has framed the changes as bringing federal data collection in line with its own priorities and executive orders. Agencies do retain broad discretion over what they collect and publish, and every administration adjusts what the government measures. Supporters argue that trimming collections outside an agency core mission is ordinary management, not concealment, and that a federal government cannot be obligated to measure everything forever.
Critics see a different problem. Their argument is not that any single dataset is irreplaceable, but that the record itself is. Once a decade of consistent measurements goes offline, no future administration can go back and collect it. The years are simply gone, and the gap stays in the record permanently.
What This Means for Americans
Most people will never open a federal dataset. But the numbers reach them anyway. They shape how much a school district receives, whether a neighborhood qualifies for an air quality review, how a hospital plans its staffing, and what a researcher can prove about whether a policy worked. When the measurements stop, those decisions get made on older information, or on estimates, or on nothing at all. The effect is not immediate and it is not dramatic. It shows up years later, in arguments that can no longer be settled with evidence.
Stay informed on the stories that matter most. Follow Palmedia News on Facebook and bookmark palmedianews.com for breaking news and analysis.