U.S. Census Bureau
Department of Commerce
4600 Silver Hill Road
Designation: ADDEMO/FRN
Washington, DC 20746
Via: Federal eRulemaking Portal – Regulations.gov
RE: Docket No. 260903-0005, FDMS Docket No. USBC-2026-0628, RIN 0607-AA75 – Decennial Census of the Population of Americans; Proposed Residence Criteria and Proposed Regulations for Demographic Questions (91 FR 57524, Sept. 10, 2026)[1]
Dear Secretary Lutnick:
The National Community Reinvestment Coalition (NCRC) appreciates the opportunity to comment on the Census Bureau’s proposed rule regarding the Decennial Census. NCRC opposes the proposed rule for four reasons. First, the proposed residence criteria would exclude people who live in the United States, including lawfully present residents, and introduce untested methods that threaten the accuracy of the census. Second, prohibiting race and ethnicity questions would undermine the demographic benchmarks used in civil rights enforcement, community reinvestment, and the federal statistical system. Third, these changes would directly impair NCRC’s existing research, public data products, and assistance to its members. Finally, the Census Bureau (Bureau) has not adequately assessed the consequences of these changes or provided sufficient opportunity for public review.
NCRC is a network of more than 700 community-based organizations dedicated to creating a nation that not only promises but delivers opportunities for all Americans to build wealth and attain a high quality of life. We work with community leaders and policymakers to advance solutions and build the will to solve America’s persistent racial and socio-economic wealth, income, and opportunity divides, and to make a just economy a national priority and a local reality.
I. The Proposed Residence Criteria Would Undermine Census Accuracy and the Reliability of Federal Data
The Decennial Census provides the population count on which a much broader federal statistical system depends. The proposed residence criteria would represent a major departure from longstanding census practice by excluding categories of people who live in the United States and by relying on administrative records and legal-status determinations that have not been tested for this purpose.[2]
The proposal would retain April 1 as the reference date but introduce a new three-month period, January 3 through April 1, for evaluating an individual’s usual residence, including consideration of tax and other administrative records. Proposed § 60.4(b) would also exclude foreign citizens who are not lawful permanent residents on April 1, including workers on employment visas, refugees and asylees who have not yet adjusted status, and individuals without lawful immigration status.
Determining legal status from administrative records poses a risk of classification errors.[3] The proposal contemplates relying on records from agencies including Treasury, Homeland Security, Department of State, Department of Defense, Health and Human Services, and the Postal Service, as well as state and tribal records, and third-party commercial data. It also seeks comment on the possibility of adding a legal-status question to the census questionnaire. Neither approach has been demonstrated to provide sufficiently complete and accurate information for determining whether millions of individuals should be included in the Decennial Census count.
Legal status is not necessarily static. Applications may remain pending for extended periods and individuals may move between immigration categories, and households frequently include people with different citizenship and immigration statuses. Linking records across multiple administrative systems therefore creates opportunities for missing, outdated, or incorrectly matched information. Those errors would not necessarily be confined to the population the proposal seeks to exclude; incomplete or incorrect record linkage could also affect citizens, lawful permanent residents, and other lawfully present residents. A legal-status question could create an additional source of error if it suppresses response among mixed-status households, a concern supported by the Bureau’s research on the citizenship question proposed for the 2020 Census.[4]
These changes create a substantial risk of increasing census undercounts. The census already has difficulty accurately enumerating some populations. The 2020 Post-Enumeration Survey estimated net undercounts of 4.99 percent for Hispanic residents, 3.30 percent for Black residents, 5.64 percent for American Indians and Alaska Natives living on reservations, and 1.48 percent for renters.[5] Adding legal-status classifications, administrative-record matching, and potentially a legal-status question to the enumeration process could compound existing coverage problems, particularly in households containing both citizens and noncitizens.
The proposed three-month period for determining where people live and reliance on tax records create additional risks of miscounting low- and moderate-income residents. Proposed § 60.2 would determine residence partly according to where an individual “lawfully spent the greatest number of days” during a January 3 to April 1 enumeration period and states that residence should be consistent with and evidenced by tax records. Yet tax and administrative records do not necessarily establish where a person actually resides. Lower-income households may not be required to file federal income tax returns; renters move more frequently; and people who are unhoused, doubled up, seasonally employed, or otherwise experiencing housing instability may not have a single administrative address that accurately reflects their usual residence. These are also populations that the census already has difficulty counting accurately. Reliance on records that incompletely represent them risks widening existing coverage gaps.
An inaccurate census count would have consequences far beyond congressional apportionment. The Decennial Census is not an isolated population count. It is a foundation for annual population estimates, the American Community Survey (ACS), geographic and demographic classifications, federal program administration, civil rights enforcement, and numerous other federal, state, local, and private-sector data systems. The Bureau’s research identified at least 353 federal assistance programs that used census-derived data to distribute more than $2.8 trillion in fiscal year 2021.[6]
Errors introduced into the Decennial Census population base can therefore propagate through other statistical products for an entire decade. Annual population estimates use the Decennial Census as their base, and those estimates in turn provide controls used to weight the ACS.[7] Census and ACS data are then used by federal agencies, state and local governments, lenders, researchers, and community organizations to measure population characteristics, identify community needs, allocate resources, and evaluate whether programs and institutions are serving communities equitably.
The proposal does not adequately explain how people excluded from the apportionment count would be treated in these other statistical products. It states that excluded residents would not be counted for apportionment and that the Bureau may maintain and publish separate statistics, but it does not clearly establish whether those residents would remain in the population estimates base, legislative redistricting data, small-area tabulations, or other datasets derived from the Decennial Census. Before making such a fundamental change, the Bureau should identify the population universe that would be used for each major census-derived statistical product and assess how exclusions or enumeration errors would affect those products.
This uncertainty has direct implications for community reinvestment and fair lending. Census and ACS data are embedded in the systems used to evaluate whether financial institutions are serving their communities. Regulators and community organizations rely on census-tract income and demographic characteristics in CRA, HMDA, fair lending, redlining, and bank merger analysis. Bank regulators, for example, classify census tracts as low-, moderate-, middle- or upper-income using tract and metropolitan-area income measures derived from federal statistical data. If populations are excluded or geographically misallocated, the resulting distortion does not remain confined to the census itself. It can affect the tract-level measures and benchmarks used to evaluate lending, investment, and access to financial services.
For a change in enumeration methods with consequences this broad, the appropriate standard cannot simply be whether the Bureau can assemble administrative records capable of assigning a legal status to individuals. The Bureau should demonstrate through rigorous testing and public analysis that the proposed methodology can accurately identify the affected population, avoid increasing differential undercounts, and preserve the integrity of the downstream statistical products that rely on the Decennial Census. Until that analysis has been completed and its implications fully assessed, the proposed residence criteria should not be adopted.
II. Eliminating Race and Ethnicity Questions Would Undermine Civil Rights Enforcement and Other Essential Uses of Census Data
The Decennial Census is the nation’s primary source of directly collected race and ethnicity counts at the census-block level. Proposed § 70.2(a) would bar any question about race or ethnicity on the short form or any questionnaire used for enumeration. The proposal treats these data as a burden on respondents which has disclosure risks and is of little use. In practice, they are the benchmark against which nearly every other federal estimate of race and ethnicity is built and checked.
The American Community Survey cannot replace the Decennial Census count.[8] The ACS is a sample survey, and its estimates are weighted to annual population estimates by age, sex, race, and Hispanic origin. Those population estimates start from the Decennial Census. If race were removed from the census, then the ACS would lose a direct source for validating its race and ethnicity estimates. ACS tract-level figures are already five-year averages with wide margins of error, and the ACS does not publish data at the block level, where redistricting and many local analyses happen. Cessation of an actual count would degrade the reliability of the ACS, with wide-ranging consequences for many statutorily mandated programs (listed below).
Administrative records are not a substitute for self-identification. Race and ethnicity fields in federal administrative records are often missing, outdated, or recorded by someone other than the person described. They have documented limitations in accurately identifying certain populations, including Alaska Native, American Indian, Hispanic, immigrant, and multiracial populations.[9] The proposal offers no evaluation of how well administrative records would perform in their place.
Eliminating the race and ethnicity questions would make it substantially more difficult to evaluate census coverage by these characteristics, particularly at detailed geographic levels. The 2020 Post-Enumeration Survey is how the public learned that the census undercounted Hispanic, Black, and reservation-based American Indian residents while overcounting non-Hispanic White residents.[10] Without race and ethnicity on the census itself, measuring these differential undercounts in 2030 would become far harder, and some errors could become more difficult to detect.
Federal agencies, states, and the private sector depend on these data to carry out their legal obligations. The table below summarizes major uses tied to Decennial Census race and ethnicity counts, directly or through the ACS and population estimates built on them.
| Agency or user | How census race and ethnicity data are used | Effect of the prohibition |
| FFIEC, CFPB, OCC, FDIC, Federal Reserve[11] | Public HMDA data report each tract’s minority population share; fair lending exams and redlining reviews compare lending in majority-minority tracts with peer lenders; merger reviews consider service to minority neighborhoods | Tract minority shares would rest on modeled estimates; a benchmark for redlining analysis would be lost or degraded |
| States, courts, and the Department of Justice[12] | Redistricting data under Public Law 94-171 include block-level race and ethnicity used to assess compliance with the Voting Rights Act and the Constitution | No directly enumerated block-level race data for 2030 legislative redistricting or litigation |
| HUD and fair housing organizations[13] | Fair housing investigations, segregation analysis, and the Indian Housing Block Grant formula, which uses American Indian and Alaska Native population counts | Weaker fair housing enforcement; funding formulas for tribal housing could lose accurate population enumeration |
| HHS, CDC, and state health agencies[14] | Population denominators by race and ethnicity for disease, mortality, and health rates | Accuracy and comparability of race-specific population denominators could be compromised |
| EEOC and employers[15] | The EEO Tabulation, built from the ACS, provides labor market availability benchmarks for workforce analysis | ACS-based employment availability benchmarks could be affected by changes in demographic population controls |
| Census Bureau[16] | Population estimates by race and Hispanic origin; ACS weighting; Post-Enumeration Survey coverage measurement | The statistical system loses a major directly enumerated demographic benchmark |
The loss would fall hard on community reinvestment. Congress passed the Home Mortgage Disclosure Act (HMDA, 1975) and the Community Reinvestment Act (CRA, 1977) in response to redlining, the practice of denying credit to neighborhoods because of their economic, racial and ethnic makeup.[17] Detecting redlining today still means comparing where lenders make loans with where low- and moderate-income individuals, families, or households and people of color live. That comparison requires accurate, tract-level counts of race and ethnicity. NCRC’s research on bank branch closures, for example, tracks closures in majority-minority census tracts separately from LMI tracts because the two do not always move together. Without a reliable count of who lives in each tract, regulators, lenders, and communities would lose the ability to see these accurately.
Eliminating race and ethnicity questions would not eliminate discrimination; it would make disparities more difficult to identify and measure. The proposal characterizes race and ethnicity as “immaterial personal characteristics” that should not “distort” the population count. But collecting self-reported race and ethnicity data is fundamentally different from using those classifications to determine eligibility for benefits or impose burdens. Federal statistical standards have long treated race and ethnicity categories as tools for consistent data collection and analysis. OMB’s own standards describe these categories as social and political constructs, used to measure conditions rather than define people biologically.[18]
Data on race and ethnicity make it possible for federal agencies, regulators, lenders, researchers, and the public to measure disparities and evaluate whether civil rights and fair lending laws are working as intended. Removing these data from the Decennial Census would not eliminate racial or ethnic disparities. It would instead weaken the statistical benchmark used to identify, quantify, and monitor them.
III. The Proposal Would Directly Harm NCRC’s Research, Data Products, and Services
NCRC’s research programs, public data products, and technical assistance depend on reliable census-tract counts of population, race, and ethnicity, as well as comparable data across Decennial Censuses. The proposed changes would directly impair these existing activities by degrading essential demographic benchmarks and undermining comparisons over time.
NCRC’s research on redlining, mortgage lending disparities, and bank branch access relies on these data. For example, NCRC’s Historic Redlining Indicator translates historic Home Owners’ Loan Corporation (HOLC) residential security grades into contemporary census tract geographies, enabling researchers to examine historical patterns of disinvestment alongside current neighborhood characteristics. NCRC’s HMDA Longitudinal Dataset and interactive mapping tools similarly rely on census tract-level demographic data to examine lending patterns across neighborhoods and over time. Changes that diminish the quality or comparability of census race and ethnicity data would directly impair these established research applications.
The proposal would also impair NCRC’s services to its network of more than 700 community-based organizations. NCRC provides lending analyses, maps, reports, and technical assistance that members use to evaluate financial institutions’ performance in their communities. Without reliable and comparable demographic data, NCRC could not provide these services with the same accuracy, detail, or historical context. Maintaining these services would require NCRC to redirect staff time to develop and validate alternative data sources and methodologies.
IV. The Bureau Has Not Adequately Evaluated the Proposal or Provided a Sufficient Opportunity for Public Comment
Fifty-four days is not enough time to comment on changes to 236 years of practice. The Bureau has historically set residence criteria for each census through a multi-year public process. For 2020, it requested public input in 2015, proposed criteria in June 2016, and finalized them in February 2018. The new proposal allows approximately 54 days. Executive Order 12866 calls for a comment period of not less than 60 days in most cases. This is a profound and unprecedented departure from longstanding Decennial Census practice, requiring greater study and deliberation. The Bureau should extend the comment period substantially and complete tribal consultation before taking further action.
Codifying these rules would make errors harder to fix. Residence criteria have traditionally been operational guidance, revisited each decade as circumstances change. Placing them in the Code of Federal Regulations would lock in an untested design and require a new rulemaking to correct problems that the Bureau discovers during testing or operations.
The regulatory analysis dismisses the rule’s effects as “indirect.” The proposal states the rule is not economically significant and that any lost information “would remain available or could be obtained in other ways.” That conclusion is not supported by any analysis of the federal programs, enforcement regimes, or private uses described above. A change that alters the population base for trillions of dollars in annual federal spending and removes the benchmark for federal race and ethnicity data, requires a full accounting of costs before it is adopted.
Proposed § 70.3(b) could restrict publication even further. It would bar publishing “any individual demographic data” if doing so requires introducing uncertainty into national or state population counts. The proposal does not explain which tabulations this would prevent, including tabulations of age and sex that remain on the form. Data users need to know what will be published before the Bureau finalizes a rule that limits it.
Conclusion
The proposed rule would undermine the accuracy of the Decennial Census and weaken the demographic data essential to community reinvestment, civil rights enforcement, and other public purposes. These changes would disproportionately affect low-income communities and communities of color. NCRC urges the Bureau to withdraw the proposal and retain a full population count, including self-reported race and ethnicity data, for the 2030 Census.
Thank you for considering our views on this important matter. If you have any questions, please contact Bruce C. Mitchell, Ph.D., Principal Researcher at BMitchell@ncrc.org, or myself at jvantol@ncrc.org.
Sincerely,
Jesse Van Tol
President and CEO
National Community Reinvestment Coalition
———-
[1]U.S. Census Bureau. “Decennial Census of the Population of Americans; Proposed Residence Criteria and Proposed Regulations for Demographic Questions.” 91 Fed Reg. 57524-57531, Sept. 10, 2026.
[2] U.S. Census Bureau. “Residence Criteria and Residence Situations for the 2020 Census of the United States.” 2020.
[3] Brown, J. David, Misty L. Heggeness, Suzanne M. Dorinski, Lawrence Warren, and Moises Yi. “Understanding the Quality of Alternative Citizenship Data Sources for the 2020 Census.” Center for Economic Studies Working Paper CES-18-38. U.S. Census Bureau, June 2019.
[4]Brown, J. David, and Misty L. Heggeness. “Citizenship Question Effects on Household Survey Response.” Center for Economic Studies Working Paper, CES-24-31. U.S. Census Bureau, June 2024.
[5] U.S. Census Bureau. “Census Bureau Releases Estimates of Undercount and Overcount in the 2020 Census.” News Release No. CB22-CN.02, March 10, 2022.
[6]Villa Ross, Ceci. “Uses of Decennial Census Programs Data in Federal Funds Distribution: Fiscal Year 2021.” U.S. Census Bureau, June 2023, p. 2.
[7]U.S. Census Bureau. ”Coverage Rates Definitions.”American Community Survey, Sample Size and Data Quality.
[8] U.S. Census Bureau. American Community Survey and Puerto Rico Community Survey Design and Methodology, Version 4.0. December 2024, Chapter 11, “Weighting and Estimation”.
[9] Ennis, Sharon R., Sonya R. Porter, James M. Noon, and Ellen Zapata. “When Race and Hispanic Origin Reporting Are Discrepant Across Administrative Records and Third Party Sources: Exploring Methods to Assign Responses.” CARRA Working Paper 2015-08. U.S. Census Bureau, Dec. 18, 2015.
[10]U.S. Census Bureau. “Census Bureau Releases Estimates of Undercount and Overcount in the 2020 Census.” News Release, March 10, 2022.
[11] Federal Financial Institutions Examination Council. “Online Data Dictionary.” FFIEC Census Data. Updated June 18, 2025.
[12] U.S. Department of Justice, Civil Rights Division. Guidance Under Section 2 of the Voting Rights Act, 52 U.S.C. § 10301, for Redistricting and Methods of Electing Government Bodies. Sept. 1, 2021.
[13]U.S. Department of Housing and Urban Development. “Indian Housing Block Grant Program (IHBG) Formula.” Office of Native American Programs.
[14]Centers for Disease Control and Prevention, “Population Denominator Data Sources.” United States Cancer Statistics.
[15]US Census Bureau, “Equal Employment Opportunity (EEO) Tabulation.” American Community Survey, Census Bureau.
[16]U.S. Census Bureau. American Community Survey and Puerto Rico Community Survey Design and Methodology, Version 4.0. December 2024. Chapter 11, “Weighting and Estimation.”
[17]Federal Reserve History. “Community Reinvestment Act of 1977.” Published May 8, 2023.
[18]Office of Management and Budget. “Statistical Policy Directive No. 15: Standards for Maintaining, Collecting, and Presenting Federal Data on Race and Ethnicity.” Revised March 28, 2024.
