.png)
It’s September, and football is back. Finally. It’s time for five months of wild plays, wilder hopes…and a deluge of dropped passes, missed kicks, and heartbreaking losses. Brutal stuff. And from the comfort of your couch, you might be tempted to think to yourself, “even I could have caught that ball… Only a fool would have called that play…”
But eventually, we begrudgingly simmer down. We remember football is a deeply complicated sport, played by experts who’ve trained their whole lives for this. They’re not immune to mistakes, and we remember that the pros make it look easy. Only a fool, while sitting on their comfortable couch, actually thinks they’d have been able to execute that play better. No matter how much football they’ve watched or played in high school.
Know what else is complicated? The job of data scientists, incorporating noise infusion, suppression, and coarsening into the U.S. Census datasets to strike a balance between granularity and privacy protection. I work with Census data nearly daily, and I wouldn’t pretend to fully understand how they produce their estimates. I leave that up to their expert career statisticians and data scientists.
“This order subverts processes developed over decades to foster transparency and public trust and creates a scenario in which there will either be less privacy for our personal information, or less usable data, or both.”
When the data becomes less trustworthy
However, a new Department of Commerce Administrative Order, 216-23, is specifically taking aim at noise infusion, a privacy protection technique the Census introduced in 2020 (read last week’s blog by Angela Maher unpacking this order). According to the statisticians at the census, noise infusion was critical for keeping folks unidentifiable, given that the “growth of commercial and public databases had made it possible to match records and re-identify individuals ‘at lightning speed.’”
Despite this, according to the Census’ new acting director George M. Cook, while noise infusion was “meant to protect confidentiality, it led to concerns that adding noise made the statistics less accurate and less trustworthy.” George M. Cook is a former BlackRock investor whose qualifications to run the Census Bureau, according to his bio, appear to be having “extensively used Census Bureau data, from which he developed a deep familiarity with and an appreciation for their work and products”
Statisticians across five reputable population research organizations have released a joint statement in response to these new changes. The experts warn, “This order subverts processes developed over decades to foster transparency and public trust and creates a scenario in which there will either be less privacy for our personal information, or less usable data, or both.”
But of course, this isn’t really a mistake. Less usable data is the point. This administrative order is already being used to justify ending the gathering of data on race/ethnicity, sexual orientation, and suppressing the counting of residents without legal status for the 2030 decennial census. You can read the official proposed rule change here, published 9/10/2026.
If you can’t identify the problem, you can’t solve it
This manipulation of census data, is part of an ongoing effort with the result of overrepresenting rural and largely white citizens and underrepresenting more diverse, younger urban residents. It also makes it nearly impossible to do things like, as the Census’ own website suggests, “evaluate government programs and policies to ensure that they fairly and equitably serve the needs of all racial groups and to monitor compliance with antidiscrimination laws, regulations, and policies.” In short, you can’t identify a problem if you make it illegal to look. And it almost goes without saying, you can’t allocate resources to address that problem. A feature, not a bug.
I’m tired of explaining how dangerous it is to censor and manipulate data to fabricate reality. Does anyone really believe fewer Americans are struggling to feed their families because the USDA stopped tracking food insecurity last year? This new administrative order is a potential disaster for our already flawed but critical tools for measuring and understanding our population.
This isn’t just a case of armchair quarterbacks thinking they’re experts just because they watch a lot of football.
Policy based on manipulated data is dangerous
Data collection is supposed to, in theory, strive to be apolitical. Deciding to count only what you want to show is inherently political. The Census is often the data underpinning data-driven and data-informed planning. Businesses, non-profits, health care systems, researchers, and governments of all size, rely on this information to improve and implement their services; these folks have no time for incompetent data. And most importantly, our communities rely on the correct allocation of vital resources.
We will all suffer (though certainly not equally) if laws and policies are developed on manipulated data.
Admittedly, my original metaphor isn’t really suited to what is currently happening. This isn’t just a case of armchair quarterbacks thinking they’re experts just because they watch a lot of football. This is more like the refs changing the rules right before a game starts, and then actively throwing the match for their preferred team. Sure, a game of football technically happened, but how meaningful were the outcomes of the game really? Public data isn’t a game.