II. The Census Is a System: Failure in One Area Can Worsen Failure Elsewhere

Most of this guide examines individual components of the 2030 Census design: questionnaire content, language access, administrative records, field operations, coverage measurement, and the release of public data. Looking at each issue separately is necessary, but it can also understate the stakes. The most serious outcomes may arise not from one isolated decision, but from several changes interacting with and reinforcing one another.

This section traces some of those possible chains of consequences. It is not intended to predict that every step will occur or that the most serious outcome is inevitable. Instead, it shows how a policy or operational change in one area could create new pressures elsewhere in the census, particularly for historically undercounted communities. Understanding those connections can help advocates identify which decisions require the strongest safeguards, where early intervention may prevent later problems, and which combinations of changes deserve particular attention.

The operational components of the census do not function independently. Decisions about questionnaire content can affect whether people participate. Participation affects how much field follow-up Census must conduct and how often it relies on proxies or administrative records. Those collection methods, in turn, affect the accuracy of the final data and the ability of Census to measure who was missed. A change that appears confined to one part of the census can therefore create consequences throughout the system.

A. Loss of Race and Ethnicity Content Could Weaken Both the Data and the Count

The questions Census asks are not separate from the operation used to count the population. They help determine whether people see value in responding, whether community organizations are willing to encourage participation, and whether the resulting data meet the needs that justify that outreach. A decision to remove or significantly weaken the race and ethnicity questions would therefore affect more than the content of the final dataset. It could also make the population harder to count.

Race and ethnicity data serve several foundational purposes. States use decennial census data to draw legislative districts, and race and ethnicity tabulations are essential to analyzing and enforcing protections under the Voting Rights Act. Federal, state, tribal, and local governments also use these data to enforce civil rights laws, evaluate whether programs serve communities equitably, and understand differences in access to housing, education, health care, employment, and other public services. Census data also inform funding and program administration, both directly and by providing the population measures, geographic information, and benchmarks used to design and evaluate government programs.

The level of detail collected matters as much as whether race and ethnicity questions appear at all. Broad categories can conceal substantial differences among communities that have distinct histories, experiences, and needs. Data limited to categories such as Asian, Black, Hispanic or Latino, or white cannot show whether particular groups within those populations experience different outcomes or are being reached by public programs. The loss of detailed data would be especially consequential for smaller communities, which are often already difficult to identify in federal statistics.

The 2024 revisions to Statistical Policy Directive No. 15 were intended to improve this information. The updated standards call for a combined race and ethnicity question, establish Middle Eastern or North African as a new minimum category, and generally require agencies to collect more detailed information beyond the broad minimum categories. If implemented in the 2030 Census, these changes could produce the first comprehensive national and local census data on the MENA population and improve information about many detailed racial and ethnic groups. A rollback could prevent those improvements from taking effect and discard the results of years of research, testing, and public engagement.

These questions also affect the operation of the census itself. People and organizations are more likely to invest in census participation when they believe the resulting data will recognize their communities and support representation, civil rights enforcement, public services, and advocacy. For many racial justice organizations, Get Out the Count work requires substantial staff time, funding, and community trust. If Census removes race and ethnicity questions or reduces them to categories that no longer produce useful information, some organizations may conclude that encouraging participation no longer provides sufficient value to their communities.

This is therefore not simply a dispute over questionnaire content. The usefulness and perceived legitimacy of the census can affect whether trusted organizations participate, whether households respond directly, and how Census ultimately obtains information about communities that have historically been missed. Census itself recognizes that it depends on community partners to reach people who may not respond to federal communications alone.

A loss of partner participation could produce consequences throughout the count. Communities that already face barriers to response could receive less outreach from organizations they know and trust. Self-response could decline, leaving Census to obtain more information through field interviews, neighbors, landlords, facility administrators, or existing government records. Those methods may produce a plausible population total in some cases, but they are less likely than a direct household response to provide complete and accurate information about race, ethnicity, household relationships, and other characteristics.

The resulting chain of events could look like this:

The Administration removes the race and ethnicity questions or significantly reduces the information they collect.
Racial justice organizations conclude that the census will no longer produce data their communities need.
Some organizations reduce or withdraw their Get Out the Count work.
Self-response declines in communities that already face barriers to participation.
Census relies more heavily on field interviews, proxies, imputation, or administrative records.
Both the population count and the remaining characteristics data become less accurate.

Each step is contingent. A change in questionnaire content would not automatically cause every organization to withdraw or every affected household to go uncounted. But the sequence is credible because community engagement is part of the infrastructure Census relies on to conduct an accurate count. Census cannot assume that organizations will continue investing in participation regardless of whether the census produces meaningful information about the people they serve.

Current commitment

The 2024 version of SPD 15 establishes seven minimum race and ethnicity categories, including MENA, and generally calls for the collection of additional detail.

Open design question

Census has not publicly finalized how the 2024 standards will be implemented in the 2030 Census or confirmed what race and ethnicity content will appear on the questionnaire.

Credible risk

Removing the questions or limiting them to less useful categories would weaken the resulting data and could reduce the willingness or capacity of racial justice organizations to support census participation.

Worst-case scenario

Race and ethnicity questions are removed or stripped of meaningful detail. Major racial justice organizations withdraw from Get Out the Count work, self-response falls in historically undercounted communities, and Census increasingly relies on records and third-party responses that provide less accurate information about those same communities. The census produces both a less accurate population count and far less useful data for redistricting, civil rights enforcement, funding, and program administration.

Where to look: Census Bureau, 2020 Census Frequently Asked Questions About Race and Ethnicity; Census Bureau, Updates to Race/Ethnicity Standards for Our Nation; Census Bureau, Research to Improve Data on Race and Ethnicity; and 2030 Census Stakeholder Engagement Strategy. Within the 2030 Census Operational Plan, see section 3.2.6, “Communications, Partnerships, and Engagement,” and section 3.2.7, “Content and Materials Design.”

B. English-Only Implementation Could Push More Households Out of Self-Response

Language access is sometimes treated as a communications issue: a question of whether Census produces translated brochures or makes its website easier to navigate. For the decennial census, however, language services are part of the basic infrastructure used to collect accurate responses. They determine whether millions of people can understand the questionnaire, identify everyone who should be included, and provide information directly rather than through an interpreter, proxy, or government record.

The 2030 Census Operational Plan assumes that substantial multilingual support will continue. It describes translated questionnaires and respondent materials, online and telephone response in multiple languages, bilingual mailings in selected areas, and translated communications, webpages, videos, audio, and scripts. Census plans to use data on households with limited English proficiency to determine which languages receive support and what level of assistance is needed.

The 2020 experience demonstrates that these services are not marginal. Census collected approximately 2.86 million self-responses in languages other than English, including about 1.8 million online responses and nearly 800,000 responses using the Spanish side of a bilingual paper questionnaire. Enumerators completed approximately 1.9 million additional field interviews in Spanish. About 13 million households were initially sent bilingual English-Spanish materials, while online and telephone response were available in 12 non-English languages and print and video guides were provided in 59.

Those plans are now uncertain. Executive Order 14224 designated English as the official language of the United States and revoked an earlier executive order directing federal agencies to provide meaningful access for people with limited English proficiency. The order did not itself require agencies to discontinue translated documents or services. Instead, it gave agency heads discretion to decide which language services were necessary to fulfill their missions.

The Department of Commerce subsequently issued DAO 201-46, which designates English as the official language for department programs and activities and appears to allow a narrower range of exceptions. Census has not publicly explained how it will reconcile that policy with the multilingual census design described in the Operational Plan. It has not confirmed which translated questionnaires, telephone services, field instruments, mailings, or outreach materials will remain available for 2030.

Restrictions in any one of these areas could affect the others. A household that cannot complete the online questionnaire in a familiar language may try to call Census for assistance. If telephone help is also unavailable, the household may wait for a paper form or assistance from a community organization. If the official paper questionnaire and partner materials are available only in English, the household may not respond at all or may submit an incomplete response.

Community partners are particularly important to this work. Organizations serving immigrant and language-minority communities often explain why the census matters, help households determine who should be included, and assure people that official materials are legitimate. They also help residents distinguish Census communications from scams and misinformation. Those organizations can supplement the Bureau’s work, but they cannot fully replace an official language program.

If Census stops producing reliable translated materials, community organizations may attempt to create their own. That could provide some assistance, but unofficial translations would not necessarily use the same terminology, reflect the latest questionnaire, or receive the quality testing applied to official products. Organizations may also lack the resources to translate online instruments, operate multilingual telephone lines, or provide trained interpreters for field interviews. The result could be uneven access depending on where a household lives and which community organizations happen to serve it.

The consequences would extend beyond whether a household responds. People completing an unfamiliar English-language questionnaire may misunderstand residence rules, omit household members, or provide less accurate information about relationships and personal characteristics. The risks may be especially high for complex or multigenerational households, households with young children, and people who divide their time among multiple residences.

A reduction in direct self-response would also interact with Census’s planned expansion of in-office enumeration. When a household does not respond, Census may eventually rely on an enumerator, a neighbor or landlord, or administrative records. Those sources may establish that a housing unit is occupied and provide a plausible population count. They are less likely to provide the same quality of information as a knowledgeable member of the household responding directly, particularly for detailed race and ethnicity, household relationships, or people whose government records contain an outdated address.

The possible chain of consequences is therefore:

Census restricts translated forms, online instruments, telephone assistance, field support, or official outreach materials.
Households with limited English proficiency have greater difficulty understanding and completing the census.
Community partners have fewer reliable materials and tools to support participation.
Direct self-response falls and incomplete responses increase.
Census relies more heavily on field interviews, proxies, imputation, or administrative records.
Household rosters and characteristics data become less complete or accurate for the affected communities.

Each step is contingent. Not every reduction in language support would produce the same effect, and many households may find other ways to respond. The consequences would depend on which languages and response modes are affected, whether trained bilingual field staff remain available, and how much support trusted organizations can provide. But the risk is credible because the 2020 Census received millions of responses through non-English instruments and interviews. Removing those options would not merely change how Census communicates with the public; it would remove established pathways through which people were counted directly.

Current commitment

The Operational Plan describes a multilingual 2030 Census that includes translated materials and online, telephone, and field support in languages other than English.

Open design question

Census has not explained publicly how the English-language executive order and Commerce DAO will affect the Language Program or identified which non-English services will remain available.

Credible risk

Restrictions on official language services would reduce direct response, make it harder for community partners to support households, and increase reliance on collection methods that provide less complete information.

Worst-case scenario

Census questionnaires, telephone assistance, online tools, field support, and official outreach materials are available only in English. Community organizations must rely on unofficial translations with uneven resources and quality control. Self-response falls among households with limited English proficiency, and Census increasingly counts those households through proxies or administrative records that omit residents, link them to outdated addresses, or provide incomplete information about household composition and characteristics.

Where to look: 2030 Census Operational Plan, section 2.2, “Hard to Count,” pp. 13–15; section 3.2.1, “Self-Response,” pp. 29–30; section 3.2.6, “Communications, Partnerships, and Engagement,” pp. 40–41; section 3.2.8, “Language Program,” pp. 42–43; and section 3.2.11, “Census Questionnaire Assistance,” pp. 45–46. See also the 2020 Census Language Program Operational Assessment, Executive Order 14224, and Commerce DAO 201-46, “Designating English as the Official Language for Department Programs and Activities.”

C. In-Office Enumeration Could Magnify Access Failures

In-office enumeration is intended to help Census resolve households that do not respond after outreach and field contact. In some cases, administrative records may provide better information than a neighbor’s estimate or the assumption that an unresolved home is vacant. But the method creates a fundamental concern: many of the households least likely to respond directly to a Census invitation are also among those least likely to be represented accurately in government and other outside records.

The barriers that reduce self-response often affect administrative data quality as well. A household that has difficulty responding because translated services are unavailable may also include people with limited contact with government programs. A family that is reluctant to interact with Census may have similar concerns about other public agencies. People who move frequently may receive census materials at the wrong address while also remaining linked to an outdated address in tax, public benefits, or school records. People living in doubled-up, informal, or multigenerational households may not appear in records as a single, clearly defined household at all.

Administrative records may therefore be weakest precisely where Census most needs an alternative to direct response. They may omit people, place them at an old address, or identify some members of a household but not others. Even when the records produce a plausible population count, they may contain incomplete or inaccurate information about household relationships, detailed race and ethnicity, housing tenure, or other characteristics that are best reported by a knowledgeable member of the household.

This creates the risk of a reinforcing cycle. Barriers like reduced language access services negatively impact the likelihood of self-response. Census then turns to administrative records to resolve the resulting nonresponse. If those records reflect the same underlying barriers, the substitute method may reproduce the original omission rather than correct it.

The models used to decide when in-office enumeration is appropriate could intensify this problem. Census plans to consider factors such as a household’s predicted likelihood of responding and the apparent sufficiency of available administrative data to determine whether in-office enumeration is an appropriate method to resolve the household. Those tools could help direct additional resources toward communities that need them. But they could also lead Census to conclude that further outreach is unlikely to succeed and that the available records are “good enough,” even when those records perform less well for the population being counted.

Aggregate quality measures may not reveal the resulting disparities. An administrative data source can appear highly complete nationwide while omitting a substantial share of a small population or performing poorly for highly mobile households. Census will therefore need to evaluate not only whether a source usually contains a person and an address, but also whether its accuracy differs by geography, population, household type, and characteristic.

The possible chain of consequences is:

Language barriers, distrust, mobility, housing instability, or complex living arrangements reduce direct self-response.
Census identifies those households as candidates for in-office enumeration.
The administrative records used to resolve them contain many of the same omissions, outdated addresses, or misunderstandings of household structure.
Census accepts a plausible but incomplete or inaccurate record as a completed enumeration.
Historically undercounted groups are missed, misplaced, or described inaccurately at higher rates.
Because the case has been marked complete, fewer opportunities remain to correct the error through direct contact.

Each step is contingent. Administrative records vary substantially in quality, and they may improve the count for many households. The concern is not that outside data are always inaccurate, but that Census may rely on them most heavily for the people and households for whom they are least reliable. A design intended to solve nonresponse could therefore deepen the disparities that produced the nonresponse in the first place.

Current commitment

Census plans to use administrative and supplemental data to enumerate some nonresponding households after at least one in-person visit and to develop models and decision rules governing when those data are sufficient.

Open design question

Census has not publicly released the thresholds it will use to determine that records are sufficient, how it will measure differential data quality, or when additional field contact will be required.

Credible risk

Households facing the greatest barriers to direct response may be disproportionately counted through records that omit residents, use outdated addresses, or misrepresent household composition.

Worst-case scenario

Census uses predicted response difficulty as a reason to reduce further field effort in historically undercounted communities. Administrative records are accepted as complete even when they reproduce the same language, mobility, housing, and trust-related gaps that prevented self-response. The households most in need of direct follow-up are instead the most likely to be counted through less reliable substitutes.

Where to look: 2030 Census Operational Plan, section 3.2.3, “In-Office Enumeration,” pp. 32–34; section 3.2.12, “Person Characteristic Frame Management,” pp. 46–48; and sections 3.2.13–3.2.14 on quality assurance and response processing, pp. 48–52.

D. Weak Post-Census Measurement Could Conceal the Problem

Counting the population accurately is only part of the census challenge. Census must also be able to measure who was missed, counted more than once, or counted in the wrong place. Those post-census evaluations are essential because they provide the official evidence used to assess disparities in coverage, understand whether new methods worked, and improve future censuses.

For 2030, Census is redesigning its traditional Post-Enumeration Survey as a broader Coverage Estimation operation. The operation will remain institutionally separate from the census it is evaluating and will continue to produce national- and state-level estimates using dual-system estimation. It will also include some independent field interviews. At the same time, however, administrative and supplemental data will play a much larger role than in prior censuses. Census plans to use those data to create the national list of housing units and as the primary source for constructing independent rosters of people. It will then compare those records with census results before fieldwork and concentrate interviews on households where the two lists differ.

This approach could improve efficiency and make it possible to produce useful new estimates below the state level. Focusing interviews on apparent discrepancies may also allow Census to investigate many likely errors more thoroughly. But the design creates a central methodological concern: the system may be less effective at identifying people who are missing from both the census and the administrative records used to evaluate it.

Coverage Estimation works by comparing two sources. Traditionally, Census matches people found through an independent post-enumeration survey with people included in the census. The differences between those sources help estimate how many people the census missed or counted incorrectly. The method works best when the likelihood of appearing in one source is sufficiently independent from the likelihood of appearing in the other. When the same factors make a person less likely to appear in both systems, the resulting estimate can understate the true number missed, a problem Census describes as correlation bias.

That concern is particularly relevant to the compounding risks described in this guide. People who move frequently, live in informal housing, face language barriers, distrust government, or have limited contact with public programs may be difficult to count directly. Many of those same circumstances can also make them less likely to appear accurately in administrative records. Even when Coverage Estimation uses records that are technically separate from those used in the census itself, the datasets may have overlapping limitations because they draw from institutions that interact with many of the same people in similar ways.

The planned targeting of field interviews could reinforce this problem. If a person appears in the administrative roster but not in the census, or appears in the two systems at different addresses, the discrepancy may generate follow-up. But a person who appears in neither source does not necessarily create a visible disagreement. Unless the Coverage Estimation design has another way to identify that omission, the household may never be selected for an interview precisely because both systems missed the same person.

For example, consider a family living temporarily with relatives. The family may not include everyone on the census because the household is uncertain about whom to count. Administrative records may still link some family members to an earlier address or fail to show that the two families are living together. If the census and the independent administrative roster reflect the same incomplete household, their apparent agreement could look like confirmation even though both are wrong.

Operational independence therefore does not automatically guarantee independence of error. Census can operate Coverage Estimation separately, protect it from information collected during census operations, and use different personnel and procedures. Those safeguards remain important. But if both systems depend heavily on records that underrepresent the same communities, the evaluation may reproduce part of the error it is intended to measure.

The 2030 design also includes Demographic Analysis, which provides a separate check using sources such as birth and death records, Medicare enrollment, and migration estimates. That program may help identify some errors that Coverage Estimation does not. Its estimates, however, are strongest for particular age and demographic groups and generally provide less geographic and population detail than advocates need to assess coverage for every historically undercounted community. It should complement, rather than substitute for, a strong Coverage Estimation design.

The possible chain of consequences is:

Barriers to direct participation cause Census to miss or misplace members of historically undercounted communities.
In-office enumeration fills some cases using administrative records that contain similar omissions or outdated information.
Coverage Estimation constructs much of its independent population roster from administrative and supplemental data with overlapping limitations.
People missing from both systems do not create the discrepancies most likely to trigger field interviews.
Official coverage estimates understate the full extent or unequal distribution of census error.
Census, policymakers, and the public receive false reassurance that the count performed better for affected communities than it actually did.

Each step is contingent. Census is still developing the design, plans to conduct field interviews, and may establish methods specifically intended to detect omissions shared across sources. Administrative records may also improve coverage measurement for many populations. The concern is not that a records-supported system cannot work. It is that Census must test whether the independent data are genuinely capable of finding people the census is likely to miss, rather than simply confirming the people both systems can already see.

Current commitment

Coverage Estimation will remain an independent operation, continue dual-system estimation, and include both administrative and supplemental data and field interviews. Census also plans to use administrative data to develop substate coverage estimates.

Open design question

Census has not publicly explained how it will identify people and households missing from both the census and the administrative sources used for Coverage Estimation, how broadly field interviews will be conducted, or how it will measure and correct for correlated errors.

Credible risk

Similar or overlapping gaps in census collection and administrative records could cause official estimates to understate undercounts for highly mobile people, households facing language barriers, people in informal living arrangements, and other historically undercounted groups.

Worst-case scenario

Census and Coverage Estimation miss many of the same people. Because the two systems appear to agree, those omissions do not trigger field investigation and are not fully reflected in official coverage estimates. The quality-measurement system provides false reassurance, making serious disparities harder to demonstrate and less likely to be corrected in future census planning.

Where to look: 2030 Census Operational Plan, section 3.3.2, “Demographic Analysis,” pp. 56–57, and section 3.3.3, “Coverage Estimation,” pp. 57–58. See also the Census Bureau’s overview of Post-Enumeration Surveys and its materials on dual-system estimation and correlation bias.

E. Disclosure Avoidance Could Make Census Data Less Useful and Remaining Errors Harder to See

The value of the census depends not only on counting people accurately, but also on releasing data detailed enough to be used. Census data support redistricting, civil rights enforcement, funding decisions, program administration, emergency planning, infrastructure investments, and community research. If disclosure avoidance requirements force Census to combine geographies or population groups, round figures extensively, or withhold data, the consequences will extend far beyond researchers’ ability to evaluate the quality of the count. Communities may lose access to the information they rely on to secure representation, distribute resources, and understand local needs.

Disclosure avoidance is the set of methods Census uses to prevent published statistics from revealing confidential information about individual respondents. These protections are essential. Census must safeguard responses, and detailed tables about small populations or small areas can create particular confidentiality risks. The challenge is to protect privacy while still releasing data that are sufficiently detailed, accurate, and timely to serve the purposes for which the census is conducted.

That balance has become substantially more uncertain. The 2030 Census Operational Plan anticipated that Census would research several possible disclosure avoidance approaches, consult data users, select a system, and test it through demonstration products and the 2028 Dress Rehearsal. Census initially selected formally private noise infusion for protecting block-level population counts. The Department of Commerce subsequently prohibited noise infusion, identified coarsening as the preferred disclosure avoidance approach, and permitted suppression only as a last resort. Census now states that its earlier 2030 research plans are no longer current and that it is evaluating alternatives.

Coarsening protects confidentiality by reducing the specificity of published information. It can include combining geographic areas, grouping population categories, rounding numbers, or publishing ranges instead of exact values. Suppression protects confidentiality by withholding particular figures altogether. These methods can provide strong privacy protections, but they do so by reducing the amount or detail of information available for public use.

The practical effects could be extensive. Redistricting depends on population data for small geographic areas and on detailed race and ethnicity data needed to evaluate compliance with the Voting Rights Act. Funding formulas and program decisions often rely on census-derived population estimates, geographic classifications, or measures of community need. State, tribal, and local governments use small-area data to decide where to locate services, plan transportation and infrastructure, prepare for disasters, and evaluate whether programs are reaching intended populations. Civil rights organizations use detailed data to identify discrimination and disparities that disappear in statewide or countywide averages.

If Census can publish information only for larger geographic areas, communities may be unable to show that their population has grown, that a neighborhood has particular needs, or that resources are being distributed inequitably. If detailed racial and ethnic groups are combined into broader categories, agencies may be unable to identify disparities affecting a particular community. If values are suppressed entirely, decision-makers may have no reliable federal data about the population at all.

These effects would be especially serious for small racial and ethnic groups and rural communities. A large racial or ethnic population may remain visible even after some categories are combined, while a smaller group may disappear into a broad aggregate. Similarly, data for a large city may remain publishable at several geographic levels, while residents of a small town or sparsely populated rural county may receive only highly aggregated information. Tribal geographies and communities that cross conventional political or geographic boundaries may face similar losses.

The effects on funding and program administration may not always be immediate or uniform. Many federal programs do not allocate money directly from a single decennial census table. But census data are built into population estimates, survey weights, geographic definitions, eligibility measures, and other statistical products used throughout government. A loss of detailed census data can therefore flow into later decisions about where programs operate, who qualifies, how funding is distributed, and whether services are meeting community needs.

Reduced detail would also make census errors harder to identify. Coverage Estimation and other evaluations may find that particular populations or areas experienced higher undercounts, overcounts, or placement errors. But those findings are useful only if Census can release enough information for communities, researchers, and policymakers to understand where the problems occurred.

Census errors are rarely distributed evenly. A national or state-level estimate may show a small net undercount even when particular neighborhoods or populations experienced much larger errors. Undercounts in one place can also be offset by overcounts elsewhere. Broad totals can therefore appear acceptable while concealing serious problems in a rural community, urban neighborhood, tribal area, or small racial or ethnic group.

For example, Coverage Estimation might find evidence that a population was missed at a higher rate in several communities. If Census can publish the estimate only for the state as a whole, the local pattern may disappear. If detailed racial and ethnic categories are combined, an undercount affecting one group may be absorbed into a much larger population. If the relevant value is suppressed because the population is small, advocates may receive no official information about the disparity.

Detailed public data also provide an external check on Census’s conclusions. Local organizations may know that an apartment complex was omitted, a shelter population was counted incorrectly, or a fast-growing rural community appears much smaller than expected. Small-area data allow them to compare official results with local knowledge, investigate discrepancies, and raise concerns with Census, Congress, courts, and program administrators.

When data are available only for broad areas or population groups, that oversight becomes much more difficult. Patterns are harder to demonstrate, independent experts cannot fully test Census’s methods, and concerns about the count are easier to dismiss as anecdotal. Census may retain detailed confidential records internally, but the public cannot evaluate information it is not permitted to see.

The possible chain of consequences is:

New disclosure avoidance requirements lead Census to combine geographies or population categories, round values, or suppress detailed statistics.
Small communities and population groups lose access to information needed for redistricting, funding, program administration, civil rights enforcement, and local planning.
Census errors affecting those same communities become harder to identify in published data.
Advocates and independent researchers have fewer tools to challenge inaccurate results, evaluate official quality measures, or pursue available remedies.
Communities that are already difficult to see in broad statistics become less visible in public decision-making.
Operational failures may persist in later population estimates, surveys, programs, and future census planning because the evidence needed to document them is unavailable.

Each step remains uncertain. Coarsening does not necessarily eliminate all useful local data, and the consequences will depend on the standards Census adopts for each product. Some tabulations are required by law or are essential to core census functions, including redistricting. But Census has not yet explained how it will preserve those products, what detail will remain available for other uses, or whether its replacement approach can be fully developed and tested before 2030.

The concern is therefore not that any use of coarsening or suppression will make census data unusable. It is that broadly restricting the methods Census may consider could leave the Bureau with fewer ways to protect both confidentiality and data utility. The costs would fall most heavily on communities that depend on detailed information because they are too small, too rural, or too geographically concentrated to remain visible in broad totals.

Current commitment

Commerce DAO 216-26 prohibits noise infusion, identifies coarsening as the preferred disclosure avoidance approach, and permits suppression only as a last resort. Census has stated that its earlier 2030 disclosure avoidance plans are no longer current.

Open design question

Census has not published a replacement research and testing plan or explained what geographic and population detail will remain available in the major 2030 Census products.

Credible risk

Coarsening and suppression could reduce the availability of data needed for redistricting, civil rights enforcement, funding, program administration, and community planning. They could also obscure concentrated census errors affecting small populations and local areas.

Worst-case scenario

Census can release only limited statistics for small populations and geographies. Detailed racial and ethnic groups disappear into broader categories, rural and tribal communities lose important local data, and some figures are withheld entirely. Communities lack both the information needed to secure representation and resources and the evidence needed to show when the census counted them inaccurately.

Where to look: 2030 Census Operational Plan, section 3.4.1, “Data Products Creation and Dissemination,” pp. 60–62, and Appendix B, “Security, Privacy, and Confidentiality,” pp. 86–90. See also Commerce DAO 216-26, “Disclosure Avoidance for Statistical Products”; the Census Bureau’s 2030 disclosure avoidance page; and the notice attached to the Bureau’s now-superseded 2030 disclosure avoidance research agenda.

F. Combined Worst-Case Scenario

Each of the risks described above could cause serious harm on its own. Their combined effect, however, could threaten not only the accuracy and usefulness of the census but also the legal validity of the count itself.

Reduced trust or less useful questionnaire content could weaken community participation. Restrictions on language access could make direct response harder for millions of households. Census could then rely more heavily on in-office enumeration and administrative records for the same communities that outreach, language services, and field operations failed to reach. If the records used to evaluate the count have similar limitations, official quality measures may detect only part of the resulting error. Disclosure avoidance restrictions could then make the remaining disparities difficult to see in the data available to the public.

This would create a reinforcing cycle. The communities most likely to face barriers to participation would also be the most likely to be counted through indirect methods, the least likely to have resulting errors fully measured, and the least likely to remain visible in published statistics. The harms would affect political representation, civil rights enforcement, funding, program administration, and communities’ ability to demonstrate their own needs.

The cycle could also create a constitutional crisis. In Utah v. Evans, the Supreme Court upheld a limited use of imputation after Census had made substantial efforts to reach every household, used inference rather than sampling, applied the method to only a tiny share of the population, and determined that it would produce a more accurate result with little opportunity for manipulation. The Court did not define the outer limits of permissible methodology, but those considerations provide a framework for evaluating whether in-office enumeration qualifies as an “actual Enumeration.”

If Census uses administrative records for a much larger share of the population, ends fieldwork before making meaningful efforts to obtain direct responses, or cannot demonstrate that the method improves accuracy for the people it is used to count, a state or another plaintiff could argue that the design falls outside that constitutional framework. Visible discrepancies in the results, weak quality measurements, or an inability to explain how individual households were counted could make such a challenge more likely and make the method more difficult to defend. This is an inference from the factors emphasized in Utah, not a settled legal rule.

Combined worst-case scenario

The Administration removes or weakens questions that help communities see value in census participation, while restrictions on multilingual services make it harder for many households to respond. Trusted organizations reduce their Get Out the Count work, and direct self-response falls in communities that already face barriers.

Census responds by relying extensively on administrative records and other indirect enumeration methods. Those records omit, misplace, or inaccurately describe many of the same people who were difficult to reach directly. Coverage Estimation relies partly on data with overlapping limitations, causing official evaluations to understate the resulting errors.

Disclosure avoidance requirements then sharply reduce the geographic and demographic detail Census can publish. Small racial and ethnic groups, rural communities, tribal areas, and neighborhoods with concentrated errors become difficult to identify in the public data. Communities and independent researchers cannot adequately validate the results or demonstrate where the count appears implausible.

A state or another plaintiff successfully challenges the broad use of in-office enumeration, arguing that Census did not conduct an “actual Enumeration.” The court invalidates the method or the affected results and orders corrective action. Depending on the scope and timing of the ruling, that remedy could require Census to revisit large numbers of households, conduct additional field enumeration, revise state totals, or develop another method of correcting the count.

By that point, the field workforce and local offices may have been dismantled, appropriated funds may have been spent, and the statutory deadlines for apportionment and redistricting may be imminent or already passed. Federal law generally requires the state population totals used for congressional apportionment to be completed within nine months of Census Day and redistricting tabulations within one year. Census could be unable to afford or operationally complete a new nationwide field effort within those deadlines.

The country would then face a choice among deeply unsatisfactory outcomes: relying on results produced through an unlawful or unreliable method, attempting a rushed and incomplete corrective count, delaying apportionment and redistricting, or asking courts and political officials to construct a remedy without a credible replacement dataset.

The precise judicial remedy would be uncertain. A court might prohibit further use of the challenged method, require revision of particular results, order additional enumeration, or provide another form of declaratory or injunctive relief. Federal law expressly contemplates declaratory, injunctive, and other appropriate relief when an unlawful statistical method is used for congressional apportionment or redistricting, but it does not prescribe a single remedy for a violation discovered after the count. A court would also have to consider the scale of the problem, when the challenge was resolved, and whether any feasible corrective method remained.

This scenario is not a prediction that every risk will occur, that a legal challenge would succeed, or that a court would necessarily order Census to repeat fieldwork nationwide. It illustrates why the legal design cannot be separated from questions of data quality and public credibility. If Census can demonstrate that in-office enumeration is carefully limited, follows meaningful attempts to obtain direct responses, and produces accurate and auditable results, it will be better positioned to defend the method. If the final data appear unreliable and the Bureau cannot show how the count was produced, courts may be more likely to conclude that judicial intervention is necessary.

The consequences of getting the design wrong may therefore extend far beyond an elevated undercount. A census that cannot be trusted, validated, or successfully defended in court could disrupt the constitutional process that allocates political representation for the following decade.