Comments from Richard Besser, MD, Opposing the Proposal to Undermine the Accuracy and Integrity of the Census

The following comments were submitted by Richard Besser, MD, Robert Wood Johnson Foundation (RWJF) President and CEO, in response to the U.S. Census Bureau’s proposed major changes to the decennial census, which would severely undermine the accuracy and integrity of the decennial census count.

On behalf of the Robert Wood Johnson Foundation (hereinafter “RWJF” or “the Foundation”), I am submitting comments to oppose the Department of Commerce and the Census Bureau’s proposed rule to change the residence criteria and regulations for demographic questions for the decennial census (hereinafter “the proposed rule”), published in the Federal Register on September 10, 2026.¹

As written, the proposed rule would:

• Redefine “usual residence” to exclude undocumented residents and temporary noncitizen visa holders from the congressional apportionment base.
• Ban questions about race, ethnicity, and sexual orientation from the decennial census short-form questionnaire.

RWJF is a leading national philanthropy dedicated to taking bold leaps to transform health in our lifetime. Through funding, convening, advocacy, and evidence-building, we work alongside communities, practitioners, researchers, nonprofit organizations, and public institutions to achieve health equity faster and pave the way together to a future where health is no longer a privilege, but a right. While the decennial census serves several fundamental purposes, including congressional apportionment and representation, RWJF's comments focus on the proposed rule's implications for the quality, completeness, and usability of the census data and the resulting consequences and harms for public health, healthcare, research, social services, and community wellbeing.

Over the last decade, RWJF has supported efforts to prepare for and ensure a complete, fair, and accurate census count, with a particular focus on communities that have historically experienced undercounts (e.g., people living in poverty, rural communities, young children, and communities of color). In partnership with our grantees, the Foundation has also led efforts to advance policies that guide race and ethnicity data collection, analysis, and reporting, with the aim of improving the health and wellbeing of every individual living in this country.

The proposed rule would have direct and harmful impacts on the communities across the country represented or served by our grantees and partners, as well as have broader impacts on public data, representation, and resource allocation that benefit every community. Accurate and complete data derived from the decennial census form the statistical backbone of the nation’s data infrastructure, supporting fair congressional apportionment and enabling governments, healthcare systems, researchers, community organizations, and others to understand population needs, plan effectively, allocate resources, evaluate outcomes, and support the health and wellbeing of all communities in the country.

Overview of the Proposed Rule’s Impact on Complete and Accurate Data

Data Impacts of Redefining “Usual Residence” to Exclude Undocumented Residents and Temporary Noncitizen Visa Holders

The proposed rule does not specify how the Census Bureau (hereinafter “the Bureau”) would determine whether a person meets the revised “usual residence” criteria, including whether it would add a question about citizenship or immigration status to the decennial census. Because such a question is one foreseeable means of implementing the proposed criteria—and the Administration has previously sought to add a citizenship question—these comments address the serious data-quality and participation concerns that would arise if the Bureau pursued that approach.

Since its inception in 1790, the U.S. census has included every person living in the country, regardless of citizenship or legal immigration status.² While the proposed rule does not explain how the Administration proposes to ascertain “usual residence,” we presume that it could include adding a citizenship or immigration question given proposals during the first Trump administration (2017–2021). Thus, our comments address this potential since no other explanation of how to determine this fact has been provided. Research has demonstrated that the addition of a citizenship question would have a significant chilling effect on individuals’ willingness to fully participate, would likely worsen the existing undercounts of certain populations, and would put the accuracy and completeness of the census count at significant risk.

Regardless of how implemented, the proposed changes will likely worsen the undercount of Hispanic, Asian American, and Native Hawaiian and Pacific Islander (NHPI) respondents.³  Asian American and NHPI communities are among the nation’s fastest-growing populations over the last decade. Although these communities were overcounted nationally, most counties actually experienced undercounts in both 2010 and 2020.⁴  Significant overcounts in some places—particularly states on the coasts, in the Upper Midwest, and in the Southwest—concealed undercounts in states in the Mountain West and the South.

Using data from the 2019 Census Test (led by the Census Bureau) and federal administrative data, research has found that asking a citizenship question would disproportionately increase survey nonresponse among Latino⁵  individuals compared to Whites and decrease self-response rates in households with at least one household member who is likely to be undocumented.⁶ Importantly, as a result of increased nonresponse, more federal resources will be required to support these increased in-person efforts needed to obtain survey responses.

Undercounts such as the ones listed above have drastic implications for federal funding for major child and family programs across multiple states. For example, even a modest undercount of Hispanics in a decennial census would likely result in 37 states experiencing reduced federal funding across Medicaid and the Children’s Health Insurance Program (CHIP), Title IV-E Foster Care, Title IV-E Adoption Assistance, and the Child Care and Development Block Grant.⁷  Together, these five programs account for close to half of all federal funding to states. Texas, Florida, Arizona, Pennsylvania, and Colorado would likely lose the most due to undercounting.⁸

The chilling effect of a citizenship question will also likely exacerbate the existing undercount of young children.⁹ This group is more likely than other age groups to be in complex living situations, such as multigenerational households and mixed-status households, where older household members who are being asked to complete the survey might be an undocumented immigrant while younger residents are U.S.-born citizens.¹⁰ Children ages zero to four were the largest undercount of any age group in the 2020 Census.¹¹ The undercount of young children has significant consequences for essential programs that support children and families, like Medicaid and CHIP, Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), and Head Start, among many other vital programs.

Finally, in addition to the impact on response rates to the census survey as a whole, research has shown that the addition of a citizenship question could influence responses to individual questions appearing later in the survey. Using an experimental design of over 9,000 study participants, researchers have found that asking about citizenship status significantly increased item nonresponse (i.e., skipped questions), further diluting the quality and accuracy of census data overall.¹²

Although the impact of a citizenship question has been examined in prior research, the impact of a legal residency question has never been assessed. Yet given the sensitivity of the question, the chilling effect is likely to be similar to that of a citizenship question. The quality of information gathered based on asking respondents about whether they are legal residents should also be questioned. Immigration law is complicated and, for many, immigration status involves frequent changes over time as a result of renewals, policy shifts, or adjustments to permanent residency, which could likely lead to confusion in how to accurately respond to the question.¹³

The proposed rule will affect the overall quality of census data and will have far-reaching consequences. The proposed rule would deepen existing undercounts—especially among Latino households, mixed-status families, and households with young children—and introduce additional bias into the census. The resulting data would be less complete, less accurate, and less reliable for allocating resources, measuring population needs, and supporting decisions at every level of government.

Data Impacts of Banning Questions About Race, Ethnicity, and Sexual Orientation

Since its inception in 1790, the U.S. census has always collected data on race and ethnicity.¹⁴ The decennial census uniquely provides comprehensive race, ethnicity, and other demographic data at the block level—the smallest geographic unit defined by the Census Bureau. These data enable researchers and organizations to map population patterns, construct and weight nationally representative surveys, and conduct more effective in-depth research.¹⁵  Government agencies also use census race and ethnicity data to assess programs and policies and to promote fairness, equity, and compliance with anti-discrimination laws and policies.¹⁶

Even though the decennial short-form questionnaire does not currently ask about sexual orientation, adding an explicit ban stops future administrations from introducing these identity-based questions down the line.

Banning the collection of data on race, ethnicity, and sexual orientation from the decennial census would have serious consequences not only for health equity research, but also for the programs and resources communities receive. Decennial counts by race and ethnicity inform grantmaking and funding formulas, helping governments and organizations identify needs and direct resources where they are most needed. Without accurate, disaggregated data, researchers and decisionmakers would have a less complete picture of health and wellbeing, and communities that have historically experienced disadvantages and barriers to health could become even less visible to policymakers. The result would be both weaker research and less effective programs, leaving critical needs unmet.

Inadequacy of Alternative Data Sources

The Census Bureau is currently administering the American Community Survey (ACS) in lieu of the census short form in its 2026 Operational Test, an indicator that it is being considered as a potential alternative to the decennial census. This raises multiple concerns.

The ACS is a continuous, year-round survey that relies on a probability sample. In other words, the quality of the estimates based on these data directly depends on information garnered from the decennial census. The ACS actually relies on the census to correct for groups being over- or under-counted in its own sample. And if race and ethnicity are removed from the decennial census, the ACS and other national datasets lose the benchmark they use to correct for those same biases, so any undercount of a racial or ethnic group would carry through uncorrected into virtually every other federal data source. The effects would also extend to the many surveys that use census-based weights to represent the U.S. population (e.g., state-level population surveys, polling research, and market research), undermining data that researchers and decisionmakers rely on.

The decennial census administration involves rigorous effort to count every individual through surveys and in-person enumeration. To do so, the Census Bureau goes to great lengths to identify every potential housing unit around the country, an intensive effort that is not replicated for the ACS or any other federal survey. In a recent study focused on a rural West Virginia county with rugged topography, the Health Affairs Institute at West Virginia University found that in every field-validated block group (i.e., census block groups or local geographic areas that have been physically checked by on-the-ground observation), verified occupied addresses exceeded the Census Bureau’s ACS estimate.¹⁷ The coverage gap ran in one direction, with the newly identified units clustered in areas hidden by rough terrain and forest. Research like this demonstrates the value of these intensive efforts that are unique only to the census.

In addition, the census consists of only nine questions and takes 10 minutes to complete.¹⁸  In contrast, the ACS consists of a much wider range of questions (nearly 80 questions covering more than 40 topics) and takes about four times longer (40 minutes).¹⁹ The brevity of the census is intentional. Survey length and time required to complete a survey both increase the likelihood that individuals will skip questions or choose not to response entirely.²⁰ Even more, these factors erode data quality because they produce bias in the data—respondents who are more busy or less interested are less likely to fully participate, reducing the generalizability of the data.

Sample-based estimates from surveys like the ACS will not provide sufficient precision for smaller geographic areas or population groups. ACS estimates are often unstable or unavailable for small and geographically dispersed communities. The survey produces annual single-year estimates only for counties with populations of 65,000 or more. In contrast, the decennial census counts every person in every area, so its accuracy does not depend on population size. Many counties, most school districts, and most census tracts have fewer than 65,000 people and the ACS sample alone cannot produce trustworthy annual estimates at that geographic unit. Instead, it relies on pooling five years of data, and even then, the result carries a larger margin of error than would an actual count. For example, cities across the country use data that rely on the census population counts to identify what types of housing their communities need in the present and the future. Because the average municipality is fewer than 10,000 people, samples of the population are insufficient to understand neighborhood housing and infrastructure needs.

While disaggregating data based on race and ethnicity is critical to have meaningful and actionable information, doing so leads to even smaller population sizes, making ACS estimates especially unreliable in those cases. Weaknesses of sample-based estimates are particularly salient for Native communities, which are comparatively small, internally diverse, frequently rural or geographically dispersed, and already subject to substantial undercounting and racial misclassification. Once ACS data are divided by tribe, geography, age, disability, gender, or another policy-relevant characteristic, the sample may become too small to support a reliable estimate.

Existing administrative data (data collected by organizations and government agencies to support program operations) cannot reliably substitute for decennial census data in determining citizenship or legal residency status, race, and ethnicity. Research conducted by the Census Bureau has demonstrated that linking decennial census data to administrative records is most effective for White, higher-income households.²¹ Accurately linking data across multiple sources requires having consistent personally identifiable information (PII) across administrative data sources, which is less likely for individuals living in poorer communities and people of color for a number of reasons. For example, they are more likely to access services at facilities that are under-resourced and unable to prioritize data quality.²² Sources like IRS data might disproportionately exclude individuals whose employment relies on the informal cash economy and who earn less than the threshold required to file federal income taxes. Administrative data for people of color are more likely to include missing data or errors because data capture systems used to convert paper or digital documents into computer-readable data are typically designed around Western name standards and lack the flexibility to accommodate cultural differences in naming structures.²³

Further, many federal data sources are not fully or consistently inclusive of direct, self-reported race and ethnicity data, which is the gold standard methodological approach. Vital statistics data, for instance, include birth and death rates (based on birth and death certificates). Race and ethnicity at birth is likely reported by the mother. In contrast, these data at death are frequently reported by an observer (e.g., physician and mortician) and have a high likelihood of differing from an individual’s self-reported race and ethnicity, particularly for Native Americans and Alaska Natives, Asian Americans, and Native Hawaiians and Pacific Islanders.²⁴ In addition, the disaggregated race and ethnicity subgroup options that were anticipated to be collected on the 2030 Census do not exist in many federal administrative records. These data are collected by some states but not by all.²⁵ Researchers will not be able to rely on administrative data for race and ethnicity as accurate or comprehensive.

Examples of How Census Data Impact the Nation’s Ability to Advance Health and Wellbeing

Below, RWJF provides numerous examples grounded in the perspective of our grantees of how census data are used to inform program and service provision, policy decisions, and disease surveillance and response at the local, state, and national levels. Many of these examples also highlight the critical role of race and ethnicity data in making census data actionable.

Census data inform local, state, and federal policy decisions about where critical resources are needed and, as such, accurate and complete race, ethnicity, and sexual orientation data are of paramount importance. These data cannot be substituted for by any other federal survey or administrative data source.

How Census Data Support the Nation’s Statistical and Data Infrastructure

• The State Health Access Data Assistance Center (SHADAC) at the University of Minnesota administers the Minnesota Health Access Survey, which also uses the census as the basis for weighting their results. These data have long been used to monitor and address disparities in the state and are part of the governor’s strategy for monitoring unfair treatment for the One Minnesota Plan.
• Census data are also the backbone for SHADAC’s sampling and weighting the broad range of federal data sources used to create the State Health Compare tool. The tool is regularly used by media to highlight racial and ethnic inequities to inform policy action. The tool was recently used to demonstrate the disproportionate impact of upcoming Medicaid cuts on Black workers in Minnesota.²⁶
  

• The Center for State Health Policy at Rutgers University provides impartial information to state policy decisionmakers about health and healthcare services. It uses census population data as the denominator for understanding healthcare utilization, population needs, and the distribution of healthcare resources across the state of New Jersey. Its New Jersey Population Health Cohort (NJHealth) Study illustrates why these data matter: researchers rely on the ACS to weight the study data, and the ACS, in turn, relies on decennial census data on race and ethnicity to correct for sampling biases.²⁷ Without this benchmark, the study’s findings would be less representative of New Jersey’s population and therefore less useful to decisionmakers.
• The full decennial census count is indispensable for understanding how long Americans live (life expectancy calculations). Even in America's biggest, most data-rich cities, the decennial census reveals population changes missed by other data gathering methods. For instance, the NYC Department of Health and Mental Hygiene revised its prior estimates of the city's life expectancy when the 2020 Census data arrived.²⁸ Citywide life expectancy estimates changed because population estimates produced between decennial censuses were not fully capturing population changes in New York City. This tells us that the decennial census full count is critical for understanding how long people live in the nation's biggest city. It is even more essential for understanding health in rural areas of the country, where population underestimates or overestimates will have a proportionally larger effect on our understanding of how long people live. 
• States rely on census data to accurately weight (adjust) state-level survey results to match population demographic distributions and inform policy decisions. The California Health Benefits Review Program used California Health Interview Survey (CHIS) data in its analysis or background for 15 bills. CHIS data on food insecurity, mental health, and hate acts were also cited in resolutions, committee hearings, and other legislative efforts.

How Census Data Inform Community Planning, Public Investment, and Local Services

• The Conservation Law Foundation uses census data to show how certain communities are overburdened by environmental problems, which can then be used in litigation. For example, in its recent case against a waste transfer station in New Bedford, Mass., census data helped to demonstrate that the 24 square mile city is already overburdened with two Superfund sites, one closed landfill, two inactive dumps, one former waste site, eleven brownfield sites, and the highest risk for childhood lead poisoning in any Massachusetts city.^{29, 30, 3}1
• Sonoma County used A Portrait of Sonoma County 2021, which combines Census Bureau education and income data with health indicators, to assess community needs and inform its American Rescue Plan investment strategy. The findings helped orient the review committee for approximately $40 million in funding for 27 community programs addressing needs including housing, childcare, educational disparities, and mental health; race and ethnicity data helped identify populations disproportionately affected by the pandemic.³²
• The Los Angeles County Office of Education’s Community Schools Initiative transforms schools into community hubs that coordinate academic, mental health, and family supports to address barriers to learning. Participating districts were selected using data derived from census indicators in A Portrait of Los Angeles County 2017 alongside county public health and educational system data, directing resources toward communities with greater educational and social needs.
• The Childhood Lead Poisoning Prevention Program (CLPPP), part of Louisville Metro Department of Public Health and Wellness, used census-tract data on housing along with locally collected data on blood lead levels to identify which Louisville neighborhoods have elevated risk of lead contamination.³³ As a result of these data, CLPPP and a local collaborative, Lead Free Louisville, have been able to work with the community to increase testing, raise awareness about the dangers of lead and how to reduce exposure, and strengthen local enforcement of housing codes to support healthier and safer housing.
• The Michigan Grand Rapids Invest Health team used census data on unemployment to designate 17 census tracts as "Neighborhoods of Focus." That designation helped direct about $50M in low-income housing tax credits, or 30 percent of Michigan's available credits at that time, plus other public and private investments, to those neighborhoods. This funding resulted in the preservation of 10 units of affordable housing and the addition of more than 50 units of affordable housing.³⁴
• Using census data, state-level program providers discovered that in central Louisiana, many more White young people were disconnected compared to their Black peers, contrary to most parts of the country. Local funders used that information to scale up their response to reach thousands of high schoolers, matching the scale of the disconnection challenge with culturally appropriate efforts. This is one example of why accurate local data are important for identifying needs rather than assuming racial patterns remain constant everywhere.
• The city of Rancho Cucamonga, Calif., used census-tract data on children in poverty, high school completion, and racial/ethnic diversity to identify higher poverty in its southwest neighborhoods. These findings were then used to write a community engagement policy, which led to tailored outreach, including translation services and community specialists, for Spanish- and Mandarin-speaking residents and other groups that had been left out of past engagement, helping them enroll more survey participants to address specific needs.³⁵
• The Evanston Health Department and the Cradle to Career coalition used census-tract data on children in poverty to find that 1 in 4 children in tract 8092, the majority-Black Fifth Ward, live in poverty, compared with near-zero child poverty elsewhere in the city.³⁶ Local groups used these findings to guide community planning, and Evanston Community Fridges expanded into the Fifth Ward with the goal to better serve a neighborhood that has experienced long-term divestment and unequal access to resources.
• Head Start programs and state early care and education agencies disaggregate eligibility data based on the census by race, ethnicity, and geographic area to decide where to add slots and hire bilingual staff. One analysis found that White children eligible for Head Start live in neighborhoods with an average of 60 eligible children per Head Start center, compared to 88 for Black and 96 for Hispanic eligible children, a gap equivalent to more than one “missing” center in underserved communities.³⁷ Without these data, funding and staffing decisions are at risk of being made without regard to communities’ needs.

How Census Data Enable Public Health and Healthcare Decisionmaking

• Hospitals use census data when conducting community health needs assessments to compare population information to convenience sample survey results, a process which culminates in the provision of community benefit activities that go beyond clinical care to serve all residents.³⁸ Of importance are data variables related to race, ethnicity, education and poverty levels, and health insurance status. As noted above, ACS data have limitations when conducting county/hospital service area level analyses due to sample sizes.
• New York hospital systems—including Northwell Health, New York-Presbyterian, Mount Sinai, and NYC Health+ Hospitals—have used Measure of America’s DATA2GO.NYC for work including community health needs assessments. The tool brings Census Bureau socioeconomic and demographic data together with health outcomes and behavior data, helping users understand the conditions surrounding patients’ lives and differences between neighborhoods. These data can assist the development of population-specific and culturally aware healthcare strategies. Data that are disaggregated by race and ethnicity can inform policy to ultimately reduce differences in health outcomes and can also help federal, state, and local governments tailor programs and more effectively target limited resources.
• Public health agencies use decennial census population counts as denominators to calculate and monitor population-based rates of disease, illness, and injury (cases per 100,000 population) so they can pinpoint the origin, spread, and effectiveness of strategies to reduce illness. The proposed rule will distort accurate counting of disease rates by omitting selective parts of a state's, county's, city's, and neighborhood's population. This is not only a threat to the population excluded, but to the entire population as diseases spread across communities without regard for citizenship or legal residency status.
• During the COVID-19 pandemic, census data were fundamental to understanding disease prevalence, transmission, and vaccination across the nation. Importantly, these data also enabled public health departments to report COVID-19 rates across racial and ethnic groups, which allowed state and local agencies to target their resources accordingly.
  • Disaggregated data enabled local public health departments to determine whether they needed to partner with specific trusted local institutions. Census data are necessary for weighting to produce accurate measures. For example, data weighted using census data showed differences in COVID-19 vaccination uptake and confidence across Black communities, a vestige of lingering mistrust of the medical system rooted in unethical experiments conducted on Black Americans.³⁹ With this knowledge, public health departments conducted outreach and disseminated media using culturally appropriate, trusted messengers and ethnic media outlets.
  • During the early months of the pandemic, lawmakers urged the federal government to produce adequate in-language translations for public health guidance.⁴⁰ Census data on racial and ethnic groups better enabled local organizations to know which neighborhoods and counties to distribute those materials and in what languages public service announcements should be prevalent in counties around the country.
  • Census race data played an essential role in discovering that NHPI populations had higher COVID-19 death rates than any other group in 6 states and the highest COVID-19 case rate in 14 states.⁴¹ The NHPI Data Policy Lab’s findings provided data for community partners that were integral in their ability to successfully secure approximately $2 million in COVID-19 resources directed toward NHPI communities.

How Census Data Support Tribal Nations and Native Communities

• Race and ethnicity data within the census are important for Native communities, particularly given their unique political and legal relationship rooted in tribal sovereignty and historical treaties. Those data help agencies identify places where Native communities may have distinct jurisdictional relationships, transportation barriers, cultural responsibilities, subsistence practices, or relationships to threatened lands and waters. Tribal governments and state and local emergency-management agencies need block-level race data to plan evacuations, emergency communications, cooling and clean-air centers, disaster shelters, and recovery services for communities facing wildfire, flooding, extreme heat, storms, and environmental contamination.

• Tribal health departments, Tribal Epidemiology Centers, Indian Health Service facilities, urban Indian health organizations, and state and local health departments use race, ethnicity, age, household, and geographic data to calculate health disparities, identify service areas, locate clinics and outreach programs, and pursue funding. These decisions require accurate denominators. An organization cannot determine whether diabetes, maternal mortality, suicide, cancer, infectious disease, or insurance gaps disproportionately affect Native people without knowing how many Native people live in the relevant area.
• Urban Indian organizations, local governments, universities, and community-based organizations use detailed race and geographic data from the census to identify Native people living away from reservations and other federally recognized tribal lands. These data inform decisions about cultural programming, behavioral health services, elder support, youth programming, transportation, and institutional partnerships.
• Tribal environmental programs, public health departments, and environmental justice organizations use race and ethnicity data alongside information about water access, housing, transportation, pollution, and proximity to hazardous facilities to identify unequal environmental exposures and prioritize remediation and infrastructure investments. Removing race and ethnicity would make it much harder to determine whether contamination, inadequate water systems, energy insecurity, or climate hazards disproportionately affect Native communities. Broad regional statistics can also conceal localized harms affecting a particular tribe or Native community.
• School systems, colleges, state education agencies, and community organizations use race and ethnicity data from the census to plan Native student services, family outreach, language access, recruitment, retention programs, and partnerships with tribes and Native organizations. Detailed race and ethnicity data are especially important for Indigenous migrants from Mexico and Central America, who may be recorded only as Hispanic or Latino even when they speak an Indigenous language and have distinct educational and interpretation needs. Removing race and ethnicity data would deepen this invisibility, while an ACS sample may include too few members of a particular community to produce usable local estimates.

Continued Implementation of Statistical Policy Directive No. 15 (SPD-15)

In contrast to the current proposal to eliminate race and ethnicity from the census, we urge the administration to continue implementing the revised 2024 SPD-15 minimum standards, including using a single combined question for race and ethnicity; adding the Middle Eastern or North African (MENA) response option; and collecting more detailed race and ethnicity categories by default.

The significant amount of research conducted by the Census Bureau over the last 20 years demonstrated that individuals of Hispanic or Latino descent found the two-question format to be confusing, leading to the overreporting of “Other” as a racial category. These findings were closely studied to inform the Office of Management and Budget’s decision in 2024 to combine the two questions. While additional research and education needs to be conducted to ensure that this change leads to reliable and valid data, the revision was strongly supported by many communities.

Currently, the only way to capture data on MENA individuals from the ACS is through place of birth, resigning researchers to studying first-generation immigrants. Without an opportunity to identify second, third, and later generations of MENA immigrants through a MENA identifier, the ability to understand the heterogeneity in the community and to understand the historical and intergenerational impacts of policies on the wellbeing of these communities is lost. 

Lastly, the collection of detailed racial or ethnic subgroup affiliation beyond the six main categories is critical for meaningfully understanding how individuals in this nation are faring and how to target the government’s resources. The current practice of aggregating distinct populations into a single group masks significant heterogeneity in health and wellbeing within those groups. Without nuanced data collection and disaggregation beyond broad racial categories, the research from which we draw policies and strive to influence better health outcomes is dramatically misinformed.

Conclusion

The proposed rule would have far-reaching consequences for the accuracy and completeness of the decennial census, the equitable allocation of federal resources, political representation, public health surveillance, healthcare and social-service planning, and the ability of governments and community organizations to understand and respond to population needs. Although framed as changes to residence criteria and demographic questions, the proposal would exclude people who live in the United States from the apportionment base, discourage participation among mixed-status and immigrant households, and eliminate race and ethnicity data that are essential to identifying disparities and directing programs and resources.

The Department of Commerce and the Census Bureau have not provided any explanation of how it would determine one’s “usual residence” or a sufficient evidence-based justification for these sweeping changes or addressed the documented risks of greater nonresponse, inaccurate responses, deeper undercounts, and cascading bias across the ACS and other data systems that depend on decennial census benchmarks. As described in these comments, neither the ACS nor administrative records can substitute for a complete count or reliable race and ethnicity data, particularly for small, rural, geographically dispersed, and historically undercounted communities. Rather than adopting this proposal, the agencies should preserve the census as a complete and accurate count of everyone living in the country and continue implementing the revised 2024 SPD-15 standards.

For the reasons stated above, RWJF strongly opposes the proposed rule and urges the Department of Commerce and the Census Bureau to withdraw the proposal in its entirety. We have included numerous citations to support research, including direct links to research. We direct the Department of Commerce and the Census Bureau to each of the materials we have cited and made available through active links, and we request that the full text of each of the studies and articles cited, along with the full text of our comment, be considered part of the formal administrative record for purposes of the Administrative Procedure Act. If the Department of Commerce and the Census Bureau are not planning to consider these materials part of the record as we have requested here, we ask that you notify us and provide an opportunity to submit copies of the studies and articles into the record.

1  Decennial Census of the Population of Americans; Proposed Residence Criteria and Proposed Regulations for Demographic Questions, 91 Fed. Reg. 57524 (proposed Sept. 10, 2026) (to be codified at 15 C.F.R. pts. 60, 70).

2  Drishti Pillai and Samantha Artiga, Proposed Changes to Demographic Data Collection in the Census Could Impact Efforts to Address Health and Health Care Disparities, KFF (Sept. 25, 2026), https://www.kff.org/racial-equity-and-health-policy/proposed-changes-to-demographic-data-collection-in-the-census-could-impact-efforts-to-address-health-and-health-care-disparities/.

3  Shadie Khubba, Krista Heim, and Jinhee Hong, National Census Coverage Estimates for People in the United States by  Demographic Characteristics, United States Census Bureau (March 2022), https://www2.census.gov/programs-surveys/decennial/coverage-measurement/pes/national-census-coverage-estimates-by-demographic-characteristics.pdf.

4  Asian Americans Advancing Justice, The Quality of the Decennial Census for Asian American and Native Hawaiian and Pacific Islander Communities: An Expanded Approach, AAJC and Demographic Analytics Advisors (March 2023), https://www.advancingjustice-aajc.org/publication/quality-decennial-census-asian-american-and-native-hawaiian-and-pacific-islander.

5  This comment uses the terms “Hispanic” and “Hispanic or Latino” when those terms appear in the cited source or when referring broadly to people of Spanish-speaking origin or ancestry. It uses “Latino” when referring specifically to people of Latin American origin or when that term appears in the cited source.

6  J. David Brown and Misty L. Heggeness, Citizenship question effects on household survey response, Journal of Policy Analysis and Management (March 20, 2025), https://onlinelibrary.wiley.com/doi/10.1002/pam.70004.

7  National Research Center on Hispanic Children and Families, States to lose federal funding for child and family programs in potential undercount of Hispanics in 2020 Census (Aug. 14, 2019), https://www.hispanicresearchcenter.org/hisp-family-facts/hispanic-family-facts-states-to-lose-federal-funding-for-child-and-family-programs-in-potential-undercount-of-hispanics-in-2020-census/.

8  David Murphey et al., Undercounting Hispanics in the 2020 Census will result in a loss in federal funding to many states for child and family assistance programs, Child Trends (Aug. 14, 2019), https://www.childtrends.org/publications/undercounting-hispanics-in-the-2020-census-will-result-in-a-loss-in-federal-funding-to-many-states-for-child-and-family-assistance-programs.

9  Eric Jensen, Census Bureau Expands Focus on Improving Data for Young Children, United States Census Bureau (March 10, 2022), https://www.census.gov/library/stories/2022/03/despite-efforts-census-undercount-of-young-children-persists.html.

10 William H. Frey, These are the groups most likely to stop participating in the US census, Brookings (April 6, 2018), https://www.brookings.edu/articles/these-are-the-groups-most-likely-to-stop-participating-in-the-us-census/.

11 Michelle Dallafior, Why We Must Count All Kids in 2030, First Focus on Children (July 5, 2026), https://firstfocus.org/resource/why-we-must-count-all-kids-in-2030/.

12 Matthew A. Baum et al., Sensitive Questions, Spillover Effects, and Asking about Citizenship on the US Census, The Journal of Politics (April 4, 2022), https://www.journals.uchicago.edu/doi/abs/10.1086/716288.

13 Forum, Explainer: Temporary Immigration Statuses (Jan. 16, 2026), https://forumtogether.org/article/explainer-temporary-immigration-statuses/.

14 Beverly M. Pratt, Lindsay Hixson, and Nicholas A. Jones, Measuring Race and Ethnicity Across the Decades, 1790-2010, United States Census Bureau, https://www.census.gov/data-tools/demo/race/MREAD_1790_2010.html.

15 National Research Council, Modernizing the U.S. Census, Chapter D: Research Uses of Census Data, The National Academies Press (1995), https://www.nationalacademies.org/read/4805/chapter/16. 

16 United States Census Bureau, Why We Ask Questions About Race (Last visited Oct. 6, 2026), https://www.census.gov/acs/www/about/why-we-ask-each-question/race/.

17 Shelly Davidov, WVU Health Affairs awarded RWJF grant to strengthen census data and community resources in rural West Virginia, West Virginia University (Sept 1, 2025), https://hsc.wvu.edu/research-and-graduate-education/News/Story?headline=wvu-health-affairs-awarded-grant-for-mcdowell-county-census-study.

18 United States Census Bureau, Census Household Survey (Feb. 19, 2025), https://www.census.gov/programs-surveys/chs.html.

19 Cynthia Robins et al., 2016 American Community Survey Respondent Burden Testing, Westat (Oct 26, 2016), https://www.census.gov/library/working-papers/2016/acs/2016_Westat_01.html.

20 Larry Hugick and Jonathan Best, Questionnaire Length in Encyclopedia of Survey Research Methods (p. 660), Sage Publications (2008), https://methods.sagepub.com/ency/edvol/encyclopedia-of-survey-research-methods/chpt/questionnaire-length#_. 

21 J. David Brown et al., Real-Time 2020 Administrative Record Census Simulation, United States Census Bureau (May 5, 2023), https://www2.census.gov/programs-surveys/decennial/2020/program-management/evaluate-docs/EAE-2020-admin-records-experiment.pdf.

22 Maura Dwyer, Quality Data Plays Key Role in Defining and Addressing Health Inequities, Pew (July 8, 2022), https://www.pew.org/en/research-and-analysis/articles/2022/07/08/quality-data-plays-key-role-in-defining-and-addressing-health-inequities.

23 National Academies of Sciences, Engineering, and Medicine, Toward a 21^st Century National Data Infrastructure: Enhancing Survey Programs by Using Multiple Data Sources, National Academies Press (2023), https://www.nationalacademies.org/read/26804/chapter/2. 

24 Elizabeth Arias et al., The Validity of Race and Hispanic Origin Reporting on Death Certificates in the United States, National Center for Health Statistics (Oct. 2008), https://www.cdc.gov/nchs/data/series/sr_02/sr02_148.pdf.

25 Asian Americans Advancing Justice, Assessing Availability and Quality of Administrative Records for Asian Americans and Native Hawaiians and Pacific Islanders: Recommendations for the Future (Last visited Oct. 6, 2026), https://www.advancingjustice-aajc.org/sites/default/files/2025-02/AdRex%20Final%20Report.pdf.

26 Izzy Canizares, Medicaid disruptions are affecting healthcare providers and will disproportionately impact Black Minnesotans, Minnesota Spokesman-Recorder (July 22, 2026), https://spokesman-recorder.com/2026/07/22/medicaid-disruptions-are-affecting-healthcare-providers-and-will-disproportionately-impact-black-minnesotans/. 

27 Joel C. Cantor et al., Health implications of established and emerging stressors: design of the prospective New Jersey Population Health Cohort (NJHealth) Study, BMJ Open (Feb. 2025), https://bmjopen.bmj.com/content/15/3/e088489.abstract.

28 Wenhui Li et al., Summary of Vital Statistics 2022 (pg. 113), Bureau of Vital Statistics and New York City Department of Health and Mental Hygiene (2022), https://www.nyc.gov/assets/doh/downloads/pdf/vs/2022sum.pdf.

29 Mike Naughton, Judge Hears Arguments in New Bedford Trash Facility Proposal, Conservation Law Foundation (Sept 3, 2026), https://www.clf.org/newsroom/judge-hears-arguments-in-new-bedford-trash-facility-proposal/.

30 Conservation Law Foundation, New England at Risk: Hazardous Sites and Climate Change (Oct. 2021), https://www.clf.org/wp-content/uploads/2021/10/Climate-Hazard-Mapping-external-methodology-document-FINAL.pdf.

31 Adilson Gonzalez Morales, Taking the Stage: The Community Behind the Victory, Conservation Law Foundation (Last visited Oct. 14, 2026), https://www.clf.org/blog/feature/taking-the-stage-the-community-behind-the-victory/.

32 County of Sonoma, Recovery Plan Performance Report (June 30, 2025), https://home.treasury.gov/system/files/136/Sonoma-County-California-2025-Recovery-Plan-SLT-0757.pdf.

33 City Health Dashboard, Integrating data to identify most at-risk communities for lead poisoning in Louisville, KY, NYU Langone Health (May 25, 2019),  https://www.cityhealthdashboard.com/impact/integrating-data-identify-most-risk-communities-lead-poisoning-louisville-ky.

34 City Health Dashboard, Accelerating investments to advance health equity in Grand Rapids, MI, NYU Langone Health (July 15, 2021), https://www.cityhealthdashboard.com/impact/accelerating-investments-advance-health-equity-grand-rapids-mi.

35 City Health Dashboard, Ensuring all residents have a voice in Rancho Cucamonga, CA, NYU Langone Health (July 25, 2019), https://www.cityhealthdashboard.com/impact/resident-voice-rancho-cucamonga.

36 Annie Robinson, The Story Behind the Map: How Evanston, IL Uses Data to Drive Community Health Action, Part 2, City Health Dashboard at NYU Langone Health (Sept. 14, 2026),  https://www.cityhealthdashboard.com/impact/data-drive-community-action.

37 Erin Hardy and Rebecca Huber, Unequal neighborhood availability of Head Start: Exploring patterns in the data, Diversitydatakids.org (Jan. 15, 2020), https://www.diversitydatakids.org/research-library/data-visualization/unequal-neighborhood-availability-head-start-exploring-patterns.

38 Kevin Barnett, Best Practices for Community Health Needs Assessment and Implementation Strategy Development: A Review of Scientific Methods, Current Practices, and Future Potential, Public Health Institute (Feb. 2012), https://www.phi.org/wp-content/uploads/migration/uploads/application/files/dz9vh55o3bb2x56lcrzyel83fwfu3mvu24oqqvn5z6qaeiw2u4.pdf.

39 Cary Funk and Alec Tyson, Intent to Get a COVID-19 Vaccine Rises to 60% as Confidence in Research and Development Process Increases, Pew Research Center (Dec. 3, 2020),  https://www.pewresearch.org/science/2020/12/03/intent-to-get-a-covid-19-vaccine-rises-to-60-as-confidence-in-research-and-development-process-increases/.

40 Cary Funk and Alec Tyson, Intent to Get a COVID-19 Vaccine Rises to 60% as Confidence in Research and Development Process Increases, Pew Research Center (Dec. 3, 2020),  https://www.pewresearch.org/science/2020/12/03/intent-to-get-a-covid-19-vaccine-rises-to-60-as-confidence-in-research-and-development-process-increases/.

41 NHPI Data Policy Lab, NHPI COVID-19 Dashboard, UCLA Center for Health Policy Research (Last visited Oct. 6, 2026), https://healthpolicy.ucla.edu/our-work/nhpi-data-policy-lab/nhpi-data-policy-lab-data-dashboards/nhpi-covid-19-dashboard.

About the Robert Wood Johnson Foundation

RWJF is a leading national philanthropy dedicated to taking bold leaps to transform health in our lifetime. Through funding, convening, advocacy, and evidence-building, we work side-by-side with communities, practitioners, and institutions to get to health equity faster and pave the way together to a future where health is no longer a privilege, but a right.