A new Commerce Department order forces federal statistical agencies to protect confidential information using only two older methods — data suppression or coarsening — sidelining modern privacy techniques like differential privacy that agencies have spent years refining. The directive, issued last month without public notice, applies to the Census Bureau and the Bureau of Economic Analysis and has drawn sharp criticism from organizations that rely on federal data.
The Council of Professional Associations on Federal Statistics, a coalition representing data users across academia, business, and government, says the order threatens both the utility of public data products and the integrity of federal science. Executive director Paul Schroeder warned that the rule could leave rural and local communities with far less usable information while offering no improvement in actual privacy.
What the Commerce Department Order Requires
Under the new directive, federal statistical agencies inside Commerce are limited to two approaches for safeguarding confidential information. Suppression means certain data simply are not released at detailed geographic levels. Coarsening means data are published only at broad aggregations — state totals instead of county figures, for example — even when the underlying records support finer detail.
Schroeder said the practical consequences are immediate for anyone who depends on granular federal data. A restaurateur planning to open in Montgomery County, Maryland, does not need the average disposable income for the entire state or for Western Maryland, he noted. That business owner wants neighborhood-level figures, the kind of detail the Census Bureau’s economic directorate has long produced.
“These data contain within it a lot of information that has policy implications, economic development, transportation infrastructure,” Schroeder said. Stripping that granularity, he argued, removes the very information that makes federal statistics useful for local decision-making.
No Federal Register Notice, No Comment Period
What has drawn equal alarm is how the order was issued. Schroeder said the directive arrived without advance notice, catching even staff inside the affected agencies off guard. Rules that reshape large swaths of public data typically move through the Federal Register with an open comment period, public discussion, and a written justification. None of that accompanied this order.
Schroeder tied the move to broader executive branch commitments on scientific integrity. The president’s executive order on restoring the gold standard of science calls for transparent and reproducible research, he said. Issuing a binding data policy without stated cause or public input, in his view, runs against that standard and risks eroding trust in federal science.
Differential Privacy Already in Use
The order does not adopt newer computational or artificial intelligence techniques that expand privacy protections, Schroeder said. Instead, it reverts to suppression and coarsening — methods that have existed for decades. The result, he argued, is a step backward at a moment when agencies have built more sophisticated tools.
Chief among those tools is differential privacy, a mathematical mechanism that injects a calibrated amount of noise into published statistics. The noise is large enough to mask any individual respondent’s answers but small enough that aggregate estimates remain accurate. Agencies can calculate precisely how much distortion they are introducing and balance it against the level of privacy they want to guarantee.
Differential privacy was used in the 2020 census and has since been extended to other Census Bureau data sets. Its adoption was not without controversy — researchers and state governments pressed the bureau over how much accuracy was being traded away — but the method ultimately allowed detailed tabulations to be published with formal privacy guarantees. Noise infusion, a related technique, has been in use for 20 to 30 years. Both address a real risk: outside data sets can be linked to federal records to re-identify individuals and their responses. Schroeder said the order leaves unclear what happens to data products that currently rely on these methods.
Rural Communities Face the Greatest Loss
Schroeder’s coalition argues the order could produce less privacy, less useful data, or both. The biggest worry, he said, is the loss of usable local data. National and state figures will remain available, but the geographies most vulnerable to re-identification — small towns, rural counties, and tight-knit communities — are precisely the ones most likely to be suppressed or coarsened away.
Federal law already requires confidentiality protections for census and economic data under Title 13 and related statutes. The question, Schroeder said, is whether reverting to blunt suppression leaves the public with information that is technically public but practically useless for the communities that need it most.
What Happens Next
Watch for three developments in the coming months. First, data users and professional associations are likely to press Commerce officials for a public explanation of the order’s rationale and for a formal comment process, possibly through petitions or congressional inquiries. Second, agencies inside Commerce will have to decide how to handle existing data products built on differential privacy — whether to reprocess them under the new rules, delay release, or seek waivers. Third, researchers, local governments, and businesses that depend on county- and neighborhood-level Census Bureau data will be watching the next major statistical releases for signs of suppressed detail.
The order also sits in tension with the Foundations for Evidence-Based Policymaking Act, which directs federal agencies to expand data sharing for evidence building. If the directive stands without revision, expect litigation or legislative pressure from coalitions arguing that blanket data suppression undermines both that law and the administration’s own scientific integrity commitments.
— Sofia Alvarez, government desk, AXO News