CensusEasy
GENERAL

Census officials confirm entire datasets will disappear under the new privacy rule

By Brenda Smith·August 7, 2026·6 min read
Census officials confirm entire datasets will disappear under the new privacy rule

Census Bureau officials have now said it out loud: entire datasets the agency has historically published will be pulled from publication because of the Commerce Department's ban on noise-infusion privacy methods. The confirmation came on August 2 at a major statistical conference, in an exchange first reported by Talking Points Memo from notes taken by attendees, during the week the Joint Statistical Meetings convened in Boston.

When we covered the June order banning noise infusion, the open question was how the Bureau would implement it and what would be lost. This is the first answer from the Bureau itself, delivered by officials including Acting Director George M. Cook and Michael Lachanski, deputy director for data, science and policy, in front of a room of statisticians.

One official, unnamed in the attendees' notes, described who gets hit. "Generally speaking, small populations, small subgroups have historically been the hardest to protect," the official said. "They are also historically the ones who get hurt the most by disclosure avoidance. There is important work that needs to be done to meet those communities' concerns." In practice that means ethnic and language minorities, rural communities, and any group small enough that detailed data about it carries a re-identification risk that noise used to solve.

The question everyone in the room was asking

Steve MacFeely, chief statistician of the OECD, put it to the officials directly. "This appears to be a decision imposed on the Census Bureau," he said. "Usually disclosure control would be considered a technical decision that should be made by professionals. How worried are you that this may be perceived as political interference?" He later told TPM he found it "very curious" that officials described the change as a policy issue rather than a technical one.

Bureau officials denied the change had political objectives and said the Office of Management and Budget coordinated it. John Abowd, the Bureau's former chief scientist, told TPM that OMB involvement in policy isn't unusual by itself. What's unusual is the process: a change of this size normally comes with stakeholder engagement that takes years. Beth Jarosz of the Georgetown Massive Data Institute called it "really unprecedented" and said it moved the decision "out of that accountability framework" into "an administration political process."

The political reading has a specific basis. Groups tied to OMB Director Russell Vought and White House Deputy Chief of Staff Stephen Miller had campaigned against differential privacy, claiming it caused an overcount in the 2020 census that shifted congressional seats toward Democratic-leaning states. That claim doesn't hold up against how the 2020 census was built: the state population totals used for apportionment were published as counted, without any privacy noise added, so differential privacy couldn't have moved a single House seat. Abowd and other experts have made the same point.

Salil Vadhan, the Harvard computer scientist who advised the Bureau on differential privacy, told TPM the order takes the toolkit "back several decades." The replacement methods, suppressing data or coarsening it up to larger geographies, have a documented weakness: "They can often be reverse engineered and undone," he said, "and that was a lot of the motivation for the start of research on differential privacy."

The ZIP, census tract and neighborhood data CensusEasy publishes exists because noise infusion made small-area detail safe to release. Which datasets get pulled, and when, still hasn't been announced. The Bureau confirmed the loss is coming before it named what will be lost, all while the agency operates without a permanent director, something we covered in the staffing crisis across federal statistical agencies.

Sources

The conference exchange, attendee notes and all quotes are from Layla A. Jones's reporting for Talking Points Memo (August 2026). Background on the June order is in our piece on the noise infusion ban, which cites the Commerce Department order and the Population Association of America's joint statement. The Joint Statistical Meetings ran August 1 to 6 in Boston, per the American Statistical Association. The 2020 census apportionment counts were published without privacy noise, per Census Bureau documentation of the 2020 disclosure avoidance system.

Related stories
Uninsured rates rose in 427 counties and fell in 65GENERAL

Uninsured rates rose in 427 counties and fell in 65

New county health insurance estimates show deterioration outnumbering improvement six to one between 2023 and 2024, with children's coverage worsening in 185 counties against nine. The plateau after Obamacare's gains looks like it's tipping into a climb.

AUGUST 6, 2026 · 6 MIN
How to Find Age Demographics by ZIP CodeGENERAL

How to Find Age Demographics by ZIP Code

Median age swings from 74.9 to 18.2 between US ZIP codes. Here is where to look the figure up for any ZIP, why the under-18 and 65-plus shares tell you more, and what the extremes turn out to be.

AUGUST 6, 2026 · 6 MIN
Frequently asked

What did Census Bureau officials confirm about data loss?

Speaking at a major statistical conference on August 2, officials including Acting Director George M. Cook confirmed that entire datasets the Bureau has historically published will be pulled from publication as a result of the Commerce Department's ban on noise infusion privacy methods. Which datasets, and when, has not been announced.

Who is most affected by the loss of census data?

Small populations, according to the Bureau's own officials: ethnic and language minorities, rural communities, and any group small enough that detailed data carries a re-identification risk. Noise infusion made publishing that detail safe, and the replacement methods, suppression and coarsening, remove the detail instead.

Did differential privacy change the 2020 census apportionment?

No. Groups aligned with OMB Director Russell Vought and Stephen Miller claimed differential privacy caused an overcount that shifted congressional seats, but the state population totals used for apportionment were published as counted, with no privacy noise added. Former Census chief scientist John Abowd and other experts have said the claim is untrue.

Brenda Smith
Written by
Brenda Smith
Brenda Smith writes about demographic change, population trends, and the Census data that reveals how American cities and towns are transforming. She resides in suburban Atlanta.