A few weeks ago, I wrote that new Commerce Department restrictions on the Census Bureau’s use of a group of data privacy methods called noise infusion could lead to less of the critical data, such as from the American Community Survey (ACS), that makes health policy research and analysis possible. But while the ban on noise infusion was unprecedented and unexpected, it also coincides with other recent legal and administrative efforts to significantly reshape the collection and publication of federal data on race, ethnicity, and citizenship status. This blog will dive deeper into the connection between data privacy methods, the potential citizenship question on the upcoming 2030 census, and Census Bureau race and ethnicity data.
The Census Bureau is required to balance protecting private information with publishing useful, accurate statistics. Noise infusion methods, as discussed below, have allowed the agency to publish highly detailed estimates without worry that data users could reidentify individuals or households. These methods are both more effective at protecting confidential information and are often the only way granular data like ACS microdata can be released at all, but critics claim that they distort the true estimates to the point of making data unusable, a high-stakes allegation for statistics that determine political representation and direct trillions of dollars in federal funding.
The 2020 census was the first to employ a method called “differential privacy” after internal testing at the Census Bureau showed that more than 52 million people in the 2010 census could be reidentified by linking census records to commercial datasets. (Differential privacy, widely considered the gold standard, adds a carefully calculated amount of noise to the underlying data to achieve a guaranteed level of privacy. The 2010, 2000, and 1990 censuses used “swapping,” switching records for nearby households with similar characteristics. Both methods have been prohibited by the recent noise infusion ban.) Importantly, differential privacy was not used in the state-level population data that were used to allocate congressional seats and Electoral College votes, but it was used to add noise to the more-detailed redistricting data used to draw individual voting districts (noise-free redistricting data has not been available since 1980 and included much less detail on race and ethnicity and geography, and no voting-age breakdown).
In March 2021, Alabama filed a lawsuit against the Department of Commerce and the Census Bureau to block the use of differential privacy in the redistricting data (16 other states filed an amicus brief in support). A district court denied the request in June, the state dropped the suit, and the data were released in August 2021. But then, in September 2025, America First Legal—a group co-founded by Stephen Miller, the current White House Deputy Chief of Staff for Policy and Homeland Security and Senior Policy Advisor during the first Trump Administration, and Gene Hamilton, who recently served as Deputy White House Counsel during the current Trump Administration and in positions at the Department of Justice and the Department of Homeland Security during the first—filed a similar lawsuit seeking to have the 2020 census results invalidated. The suit claimed that the census failed to produce the legally required “actual Enumeration” because differential privacy skewed population totals and improperly redistributed congressional seats away from states with Republican leadership in favor of states with Democratic leadership. A district court dismissed these claims because the statute of limitations on activity related to the 2020 census had already passed. (The 2020 census, despite funding and operational shortfalls and the pandemic, was accurate, as the Census Project explains. Though some groups, including renters, young children, and Black, Hispanic, and American Indian individuals living on reservations were undercounted, there was no statistically significant national undercount.) America First Legal refiled their suit in February 2026, another court dismissed it with prejudice last month, and the group last week asked the Supreme Court to hear an appeal.
It is easy to reject the technical allegations in these lawsuits because, to reiterate, 2020 census data were released five years ago and differential privacy was not used in the state-level apportionment data. But the methodological questions are really secondary to an argument that the “wrong” people are being counted in the first place. The America First Legal case goes on to argue that certain groups of noncitizens should not be included in the count because they do not fit a reading of pre-Fourteenth Amendment “persons” as only those people who make up the “body politic.” The Center for Renewing America, an organization founded by Russell Vought, the Director of the Office of Management and Budget, also released a strategy document last year on republishing “corrected” 2020 census results without differential privacy applied and without noncitizens included in the count, to “[signal] that political representation in the United States flows from lawful presence and civic belonging and not from numerical manipulation or bureaucratic algorithms.” These arguments are now carrying into disturbing new proposals for the upcoming 2030 census.
Republicans in Congress have now introduced legislation and written to the Commerce Secretary in support of adding a citizenship question to the 2030 census, and the Census Bureau is currently fielding a citizenship question in the 2026 Census Test (plans for the Census Test were scaled down and changed dramatically earlier this year, including reducing the number of test sites from 6 to 2, using the American Community Survey form instead of the decennial census form, and scrapping plans to improve the count of young children). In addition, the Administration is circulating a proposed rule that would not only exclude noncitizens from the count— lawful permanent residents and “approved” asylees and refugees would be counted while undocumented immigrants, international students and H1-B visa holders would not be—but also remove questions on race, ethnicity, and sexual orientation from the census short form altogether.
Fears about being counted in census data are two-fold. On the front end, noncitizens, individuals in mixed-status families, and others may worry that, as a Commerce Department employee reportedly put it, “if they answer the Census they will be rounded up.” On the back end, the recent noise infusion ban and legal challenges of methods like differential privacy raise serious concerns about how well the Census Bureau will be able to protect the confidentiality of survey responses. As the Center for Renewing America put it, “[e]ven if the citizenship question is added to the Census, it will be impossible to ascertain the status of individuals so long as differential privacy is used.” That clearly implies that one goal of prohibiting noise infusion is to help identify noncitizens. A 5-4 Supreme Court decision struck down a previous attempt to add a citizenship question in 2019 during the first Trump’ Administration, because the Commerce Department’s stated reason for adding it, to enforce the Voting Rights Act, did not “offer genuine justification.”
In addition, in April of this year, a 6-3 Supreme Court decision in Louisiana v. Callais struck down a key piece of the Voting Rights Act, sharply raising the bar for future plaintiffs to prove that district maps were drawn with intent to discriminate. Eliminating the collection of race and ethnicity information on the census form would raise this bar even further, cutting off a critical source of data and making it even more difficult to identify geographic and demographic inequities, not just for voting districts but in numerous areas like enforcement of civil rights, equal employment opportunities, and health and environmental disparities. The Census Bureau lists data privacy concerns about “sensitive” demographic information, lawsuits over the 2020 use of differential privacy, and other legal issues related to race and ethnicity, including Louisiana v. Callais, as the reasons they are “re-evaluating the proper role of these demographic questions.”
This “reevaluation” is also apparent in other administrative actions across the Census Bureau. The agency was supposed to adopt updated race and ethnicity standards for the 2027 American Community Survey (ACS) under revisions to Statistical Policy Directive 15 (SPD 15) issued in 2024, including combining the two separate race and ethnicity questions into a single question and adding a Middle Eastern or North African group. These changes, which had undergone more than a decade of testing and extensive comments and review processes, were designed to correct for rising nonresponse rates and an increase in selections of “Some Other Race,” particularly among Hispanic and Latino and Middle Eastern and North African individuals. The Census Bureau, however, announced in May of this year that they would delay implementing the revised standards for another year and instead use the earlier 1997 standards. Two weeks after the announcement, the Center for Renewing America published a report calling SPD 15 part of a “woke and weaponized agenda” related to DEI and racial gerrymandering.
The harmful implications of the noise infusion ban are already becoming clearer, as the Census Bureau recently began to publish implementation plans on how to comply with the Commerce Department order and announced an unspecified delay to the 2025 ACS data, and so far, it looks like predictions about there being “less privacy for our personal information, or less usable data, or both” are shaping up to be accurate. A few weeks ago, I wrote that we had more questions than answers about recent actions from the Census Bureau. But as the legal and administrative pieces begin lining up, we may not be left with enough data to even know what to keep asking.