Executive Summary

The Census Bureau’s first operational plan for the 2030 Census is not a final blueprint. Known as “Baseline 1,” it provides a high-level description of the Bureau’s current approach, with more specific operations and detailed rules expected after further testing. Even at this early stage, however, the plan reveals a consequential shift in how Census expects to count the population. The Bureau plans to rely much more heavily on administrative records, linked person-and-address data, automated models, and work performed in the office rather than in the field. It also plans to process responses while collection is still underway, automate more aspects of group quarters enumeration, and reduce portions of its physical field infrastructure. Future versions of the plan will add detail, but many of its central design choices are already taking shape.

At the same time, a broader set of federal policy changes, some already adopted and others still under consideration, could substantially reshape the final census design. The Administration’s English-only Executive Order, along with the Department of Commerce directive implementing it, could affect multilingual census services. A separate Commerce order prohibits the disclosure avoidance methods Census had selected for protecting some 2030 data without yet providing a clear replacement. Other possible changes could be just as consequential, including efforts to exclude noncitizens from the population counts used for representation or to alter the questions Census asks about race, ethnicity, or household relationships.

Together, these developments create uncertainty not only about how Census will conduct the count, but also about who will be included, what information will be collected, and whether the resulting data will be sufficiently detailed and accessible to serve communities, enforce civil rights laws, and support fair political representation.

Bottom line

The 2030 Census could use new data and technology to make the count more accurate and accessible. But without strong safeguards, the same changes could reduce direct participation, deepen errors for historically undercounted communities, expose the count to legal challenge, and leave the public with far less useful local data. Advocates have an important window to shape the design before broad concepts are converted into technical rules, contracts, and operational systems.

The five takeaways below highlight the operational changes and policy decisions most likely to determine whether historically undercounted communities are reached, counted accurately, and visible in the final data. They include both serious emerging risks and opportunities for advocates to improve the 2030 Census while its design is still taking shape.

1. Census plans to rely much more heavily on government records instead of direct responses and fieldwork

The most consequential change for 2030 is the expansion of “in-office enumeration,” which would allow Census to count some nonresponding households using administrative records and other outside data. Census will build models to evaluate the available records, predict whether a household is likely to respond, determine how it should be contacted, and decide when outside data are sufficient to complete the count. A new “Person Characteristic Frame” will support this work by linking people to addresses and assembling information about their characteristics and household composition.

Administrative records may help resolve some genuinely difficult cases and could be more accurate than treating an unresolved occupied home as empty. But the change also poses serious risks for historically undercounted groups. People who move frequently, live in informal or complex households, have limited contact with government programs, or are reluctant to share information with public agencies are often missing from administrative records or linked to the wrong address. Even when a record includes the correct person, it may provide incomplete or inaccurate information about race, ethnicity, household relationships, or other characteristics. The Operational Plan itself acknowledges that traditional administrative sources often fail to capture highly mobile people, people who face language or literacy barriers, and people who have concerns about providing information to government.

The plan states that Census will use administrative records to count nonresponding households only after at least one in-person visit. That is an important minimum safeguard, but a single unsuccessful visit is not necessarily a meaningful attempt to obtain a direct response. Much will depend on the timing of the visit, whether additional visits are required, and how Census decides that field collection is complete.

In-office enumeration also creates a significant legal risk. In Utah v. Evans, the Supreme Court upheld the limited use of imputation under the Constitution’s requirement for an “actual Enumeration.” The Court emphasized that Census had made efforts to reach every household, used inference rather than statistical sampling, applied the method to only a tiny share of the population, and concluded that it would produce a more accurate count with little opportunity for manipulation. The Court did not establish a rigid test or define the outer limits of permissible census methodology. But its reasoning provides an important framework for evaluating a method that could be used earlier, more systematically, and on a much larger scale than the imputation at issue in that case. The Operational Plan does not explain how the proposed design has been evaluated against that precedent.

Worst-case risk: Census uses predictive models to reduce fieldwork in communities expected to respond at low rates, then counts those households using records that omit or misplace many residents. If the method is used broadly, a successful legal challenge after Census Day could place the validity of affected results in doubt at a point when repeating field collection would be extremely difficult. In the most disruptive scenario, the litigation could destabilize apportionment or redistricting based on those results. The available precedent does not predetermine either the outcome of such a challenge or the remedy a court would order.

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 Figure 14, p. 28. See also Utah v. Evans, 536 U.S. 452.

2. The Department of Commerce’s English-language policy could undermine access throughout the census

The Operational Plan assumes that the 2030 Census will provide substantial language support. It describes translated questionnaires and respondent materials, multilingual internet and telephone response, bilingual mailings in selected areas, and translated communications, webpages, videos, audio, and scripts. Census explains that these services improve data quality by allowing people with limited English proficiency to respond directly and accurately.

These services were used at significant scale in 2020. Online and telephone response were available in 12 non-English languages, while video and printed language guides were provided in 59. Census received 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 also completed approximately 1.9 million field interviews in Spanish. About 13 million households were initially sent bilingual English-Spanish census materials.

The Operational Plan’s language access strategies are now uncertain because the Department of Commerce subsequently issued a DAO designating English as the official language for department programs and activities. Census has not publicly explained how it will reconcile that directive with the multilingual operation described in the plan. Important unanswered questions include whether translated questionnaires will remain available, whether telephone assistance will continue in multiple languages, and whether Census can produce the materials community partners need to reach households with limited English proficiency.

The underlying presidential executive order did not require agencies to stop producing documents or services in languages other than English. It left those decisions to agency heads based on their missions. That makes Commerce’s specific implementation, and its application to the constitutionally required census, particularly consequential.

This is not merely an access or communications issue. If people cannot understand the questionnaire or obtain assistance in a language they speak well, they are less likely to respond and more likely to make errors about who should be included. Census would then need to rely more heavily on field interviews, proxies, or administrative records, all of which may produce less complete characteristic data than a direct household response.

Worst-case risk: Census questionnaires, telephone assistance, digital tools, and official outreach materials are made available only in English. Community organizations must create unofficial translations without consistent terminology or quality control. Response falls among households with limited English proficiency, while those households are increasingly counted through administrative records that may not represent them accurately.

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 3.2.8, “Language Program,” pp. 42–43; and section 3.2.11, “Census Questionnaire Assistance,” pp. 45–46. See also Executive Order 14224, “Designating English as the Official Language of the United States”, and DAO 201-46, “Designating English as the Official Language for Department Programs and Activities.”

3. The disclosure avoidance plan has been disrupted without a clear replacement

Census must protect the confidentiality of individual responses while still releasing useful statistics about communities. The Operational Plan anticipated a multiyear research process to improve or replace the disclosure avoidance methods used in 2020. Census initially selected formally private noise infusion to protect block-level population totals and planned to conduct research, produce demonstration data, engage stakeholders, and apply the proposed system to the 2028 Dress Rehearsal.

In June 2026, however, the Department of Commerce issued DAO 216-26 on disclosure avoidance for statistical products. The order prohibits noise infusion and identifies coarsening as the preferred approach. Census has since added notices to disclosure avoidance webpages stating that its prior information is no longer current and that it is evaluating alternative approaches. It has not yet published a replacement research, testing, or implementation plan.

The practical consequences remain uncertain. Without noise-based methods, Census may have to rely more heavily on approaches such as combining geographic areas, grouping demographic categories, rounding figures, publishing ranges, or suppressing data. Each can protect confidentiality, but each can also make small populations and local communities less visible. Data below the county level could face particular pressure if Census relies substantially on geographic aggregation.

The timing is also serious. Disclosure avoidance research was expected to extend through 2027, with major decisions and system development completed in time for the 2028 Dress Rehearsal. A long delay could leave Census with insufficient time to develop and test a replacement, engage data users, and correct problems before production.

Worst-case risk (not a current prediction): Census cannot identify an approved method that provides strong confidentiality protection while preserving detailed data. Some products are released only at the state level, many others only at the county level or above, and block- or tract-level statistics needed for redistricting, civil rights enforcement, disaster planning, and neighborhood research are sharply reduced or unavailable.

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 the Census Bureau’s 2030 disclosure avoidance research page, its now-superseded 2030 research agenda, and Commerce’s disclosure avoidance order.

4. Several proposed changes could improve access and data quality

Although many parts of the plan require close scrutiny, it also includes changes that could improve the count. Census is considering allowing people to request a paper questionnaire at any point during the self-response period rather than waiting to receive one through the standard mailing schedule. That could benefit people who distrust online links, lack reliable internet access, or simply prefer a physical government form. Census also plans to continue allowing online responses without a Census ID, preserving an important option for people who did not receive or cannot locate their mailed invitation.

The plan retains Mobile Questionnaire Assistance, through which Census staff work with community partners at libraries, grocery stores, senior centers, and other trusted locations to help people respond. It also proposes allowing residents of some group quarters, including some college students, to provide their own responses rather than relying entirely on facility administrators. Direct self-response could improve the accuracy of race, ethnicity, and other personal characteristics, although Census will need a strong backup plan for residents who do not respond.

Near real-time processing may present another major opportunity. Census plans to begin processing responses while data collection is still underway rather than waiting until collection ends. That could allow the Bureau to identify geographic gaps, incomplete household rosters, questionable address information, or other quality problems while there is still time to send cases back to the field. But the system must be designed carefully so that valid responses from less common households are not rejected or “corrected” simply because they conflict with administrative records or historical expectations.

These improvements should be treated as advocacy opportunities, not simply as settled commitments. Some are only under consideration, and others could be implemented in ways that primarily reduce costs rather than improve participation. Advocates should press Census to retain the strongest proposals, test them with historically undercounted communities, and measure whether they actually improve the quality of direct responses.

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 3.2.4, “Group Quarters and Special Populations Enumeration,” pp. 34–37; section 3.2.6, “Communications, Partnerships, and Engagement,” pp. 40–41; and sections 3.2.13–3.2.14 on quality assurance and response processing, pp. 48–52.

5. The decisions that matter most have not yet been made publicly

Baseline 1 describes broad operational areas, but many of the rules that will determine who is counted directly and who is counted through outside data remain unresolved or undisclosed. These include the contact strategy for nonresponding households, the standards for deciding that administrative records are sufficient, and the procedures for resolving conflicting responses. Other consequential choices include which languages will be supported, what triggers a quality-control intervention, and when field collection may end.

These are not minor implementation details. A rule defining a “completed” case could determine whether an enumerator returns to a household. A matching threshold could determine whether two records are treated as the same person. A model could direct additional outreach toward a community or conclude that further effort is unlikely to succeed. Collectively, these technical decisions will constitute the real operating rules of the census.

The Bureau plans to identify specific operations in Baseline 2, currently scheduled for summer 2027, and to release initial detailed operational plans in early 2028. The 2026 Census Test, continuing small-scale research, and the 2028 Dress Rehearsal will inform those decisions. By the time software is fully designed and major contracts are executed, however, changing the underlying rules may be far more difficult.

Advocates therefore should not wait for a final operational plan. They should seek the underlying research reports, model specifications, quality standards, decision memoranda, and testing measures while those materials can still influence the design.

Where to look: 2030 Census Operational Plan, section 1.5, “An Iterative Approach to Maturing the Operational Design,” pp. 8–11, and section 5.1, “What’s Next for the Operational Design,” pp. 81–82. See also Memorandum 2024.04, 2030 Census Operations Strategy and Roadmap, the Operational Plan Explorer, and the 2030 Census Research and Testing page.

How the risks could compound

The risks described in this guide are not independent. A politically driven change to the race and ethnicity questions could lead racial justice organizations to reduce their Get Out the Count work. Restrictions on translated materials could further reduce self-response. Census could then rely more heavily on administrative records for the same households that community outreach and language assistance failed to reach. If Coverage Estimation relies on similar records, official quality measures may not fully detect the resulting omissions. Finally, restrictive disclosure avoidance methods could make it difficult for researchers and communities to identify geographic and demographic patterns.

Combined worst-case scenario

The census removes or weakens questions that motivate community participation, reduces multilingual access, substitutes administrative records for meaningful field follow-up, and evaluates the results using data with overlapping limitations. It then releases statistics only for broad populations and geographies. The country would know less about who was counted, have fewer tools for identifying who was missed, and lose much of the detailed information needed for civil rights enforcement, political representation, and equitable program administration.

Priorities for advocates

The current uncertainty creates substantial risk, but it also creates an opportunity to influence the census before the design is locked in. Over the next two years, advocates should press Census and Commerce to:

• Explain in writing how the two Commerce DAOs will affect language services, disclosure avoidance research, tabulation geography, and the release of detailed data.

• Preserve direct self-response and meaningful field follow-up as the preferred methods of enumeration, with administrative records used as a carefully limited backstop.

• Publish decision rules, quality measures, and research underlying in-office enumeration early enough for independent legal and scientific review.

• Use the 2026 Census Test, Baseline 2, and the 2028 Dress Rehearsal as concrete accountability points for evaluating effects on historically undercounted communities.

The 2030 Census can use administrative records, automation, and new technology without abandoning the principles of direct participation, transparency, and equal representation. Efficiency, however, is not itself a measure of census quality. The central question for every operational change must be whether it makes the census more likely to count every community accurately and to produce data that allow every community to be seen.