The Trump administration has proposed a regulation that, if not reversed by the time the president takes office in 2029, would end the long-standing practice of collecting race and ethnicity data in the decennial census. This change could have significant implications for public health and resource allocation. The most widely reported aspect of the regulation is the decision not to count noncitizens without green cards, which could affect future elections and federal funding. However, the proposed change to race and ethnicity data collection is also considered important and could have large implications for public health. Race and ethnicity data are key in understanding population health outcomes, as differences in exposure to environmental stressors and underlying vulnerability can lead to stark health disparities. For example, Black children are 60 percent more likely than U.S. children overall to have asthma and four and a half times more likely to die from it. Hispanic women have a 40 percent higher incidence of cervical cancer and 20 percent higher death rate from it compared to non-Hispanic white women. Asian Americans have extremely elevated rates of certain diseases, including tuberculosis and hepatitis B. Government agencies use race and ethnicity data as a proxy or input into models determining the prevalence of these health conditions. The U.S. Centers for Disease Control and Prevention combine survey data with census race and ethnicity estimates to model the prevalence of chronic diseases at a fine-grained geographic level. This granular data allows decisionmakers to allocate resources and target outreach to the areas with the highest need. For example, Michigan has established a pilot program to place mobile health unit vans in at-risk areas, and private health systems also use this type of data frequently. The practice of collecting race and ethnicity data is as old as the census itself. The 1790 census differentiated between "Free White" individuals and all others, and race has appeared in every census since. The White House Office of Management and Budget promulgated the first government-wide standards in 1977, which have been used by every administration during the past five decades. The Biden administration made long-needed changes to the standards in 2024, including adding a "Middle Eastern or North African" category. That revision was undertaken with analytical rigor: The process took two years, involved 35 agencies, incorporated extensive new and existing research, and elicited over 20,000 comments. The Trump administration's proposed rule has no scientific justification or participation by federal agencies with experience in this area. The rationale provided by the rule does not withstand scrutiny. The administration first argued that lowering the number of questions would increase response rates, but this argument is based on two studies, one over 30 years old and the other not about the census. Additionally, the proposal would swap out the race question for one on citizenship, suggesting the concern over response rates is a subterfuge. The proposed rule also suggests that the collection of racial data may be unconstitutional under recent Supreme Court decisions in Students for Fair Admissions v. President and Fellows of Harvard College and Louisiana v. Callais. However, neither decision addresses the constitutional status of data collection. Callais concerned the role of race in drawing legislative districts under the Voting Rights Act, while Students for Fair Admissions involved racial preferences in college admissions. Courts have made clear that the collection of racial data does not raise the same constitutional concerns as racially based preferences. Finally, the rule argues that the collection of race and ethnicity data would continue through the Census Bureau’s yearly American Community Survey. However, the two are complements, not substitutes. The ACS has an annual sample size of only 3.5 million, whereas the decennial census has universal reach. That smaller sample means certain demographic groups may be over- or under-covered, biasing the total estimates. Researchers adjust the data using the demographic data derived from the decennial census, a process that would be impossible without the race and ethnicity question.