Noise infusion banned from statistical products published by Census Bureau
Points and comments are a snapshot, not live.
U.S. Department of Commerce bans noise infusion from Census Bureau statistical products, replacing differential privacy with coarsening and suppression.
The Department of Commerce ordered the Census Bureau and Bureau of Economic Analysis to stop using noise infusion in statistical releases, prioritizing coarsening first, then suppression as a last resort. Noise addition, particularly through differential privacy, became standard after the 2010 Census's swapping method proved vulnerable to record reconstruction attacks. The author argues banning noise will force a severe privacy-utility tradeoff: coarsening and suppression are blunt instruments that either destroy data utility for small populations or remain vulnerable to privacy attacks. Competing disclosure avoidance methods (cell key, swapping, sampling, imputation) also involve randomness, so the ban may have broader collateral impact on statistical products beyond census releases.
What commenters are saying
Commenters split between those defending census privacy protections and those questioning why such sensitive data should be collected at all. Defenders note the census needs respondents to trust it with citizenship status, income, disability data, and other sensitive information; publishing complete records would undermine participation and data quality. Critics argue the constitutional requirement is merely a headcount by state, and demographic data should come through voluntary surveys or existing administrative sources like tax returns. One commenter notes detailed records already face a 72-year publication delay for privacy. Another points out Census data has historically enabled gerrymandering and discrimination, complicating privacy versus transparency tradeoffs.