XVII. Testing, Timing, and Accountability
The 2030 Census design is supposed to mature through a sequence of research, field tests, increasingly detailed operational plans, and final production decisions. In principle, this gives Census several opportunities to find problems before they become part of the census.
The usefulness of that process depends on timing. A test only informs the design if it asks the right questions, includes the populations necessary to answer them, and produces results before the relevant decision is locked in. A public report released after requirements have been approved, contracts awarded, and software built may improve understanding without creating a realistic opportunity for change.
This concern has become more important since Census substantially reduced the scope of the 2026 Census Test. The original test was designed around six sites selected collectively to include tribal areas, rural and remote communities, places with limited mail and internet access, group quarters, hidden housing units, and historically undercounted populations. In February 2026, Census narrowed the test to selected areas of Huntsville, Alabama, and Spartanburg, South Carolina, with a much more limited focus on In-Field Enumeration and a pilot involving U.S. Postal Service employees.
That does not make the 2026 Test unimportant. It does mean that Census will need other evidence for many major elements of the 2030 design. Advocates should track not only the two large field tests, but also the smaller studies, procurement milestones, requirements decisions, and software-development schedules occurring between them.
A. Planning Timeline
The current Census planning schedule leaves several major decision points before Census Day:
| Timing | Major milestone |
|---|---|
| July 2025 | Operational Plan Baseline 1: initial high-level design |
| May–August 2026 | 2026 Census Test: limited operational test in Huntsville and Spartanburg |
| Summer 2027 | Operational Plan Baseline 2: specific 2030 operations identified and refined using 2026 findings |
| Early 2028 | Initial Detailed Operational Plans: more granular descriptions to support integration and engineering |
| April 1, 2028 | 2028 Dress Rehearsal: end-to-end test of major operations and system handoffs |
| Summer 2029 | Operational Plan Baseline 3: production design incorporating Dress Rehearsal findings |
| Late 2029 | Refined Detailed Operational Plans for peak production |
| April 1, 2030 | Census Day |
| December 31, 2030 | Statutory deadline for apportionment counts |
| April 1, 2031 | Statutory deadline for P.L. 94-171 redistricting data |
| 2031 and after | Additional demographic and geographic data products on schedules still to be finalized |
Census describes this as an iterative process: each baseline is supposed to incorporate what has been learned through research and testing, with the 2028 Dress Rehearsal serving as the final major opportunity to test how the components work together.
The timeline also shows why waiting for the Dress Rehearsal is risky. Initial Detailed Operational Plans are due before the rehearsal, and systems, contracts, and requirements must be developed well before Census can test them at scale. By 2028, many design choices will already have substantial operational momentum.
The relevant question for advocates is therefore not simply when Census tests something, but whether the test occurs before the point when an unfavorable result can still change the design.
B. Is the 2026 Test Capable of Answering the Right Questions?
When Census selected the original six 2026 Test sites in 2024, it explicitly emphasized their collective diversity. The sites included tribal lands in Arizona, the Qualla Boundary and surrounding rural counties in western North Carolina, remote areas of western Texas, and metropolitan sites with universities, correctional facilities, military housing, hidden housing units, and other characteristics. Census stressed that no individual location could satisfy all of the test requirements; the strength of the design came from the combination.
That design no longer exists.
The current 2026 Test includes approximately 154,600 selected housing units in Huntsville and Spartanburg. The self-response questionnaire is online and English-only, with no paper or telephone response. Fieldwork focuses on nonresponding households and the USPS pilot. Census says the two sites include urban and rural areas, historically low-response areas, and complex or hidden housing units, which makes them useful for evaluating some In-Field Enumeration questions.
But the reduced test cannot provide equally strong evidence for many issues central to the 2030 Census.
Tribal outreach. Neither current site is on tribal land. The original Arizona and western North Carolina sites would have allowed Census to test operations and outreach in partnership with three tribal nations. That experience cannot be reproduced adequately in Huntsville or Spartanburg.
Language access. The online questionnaire for the test is available only in English, and paper and telephone response are not being offered. The test therefore cannot function as a meaningful operational test of multilingual internet response, telephone assistance, bilingual mailings, or the interaction among language services and response modes.
Rural addressing. Both sites contain some rural areas, so they can provide information about fieldwork outside urban neighborhoods. But they do not reproduce the remote addressing conditions that motivated the original western Texas, tribal, and western North Carolina selections. Census will need to rely heavily on the separate Locate Address Test and other targeted research for those questions.
Group quarters. The current public test design focuses on selected housing units and does not provide the broad group quarters testing envisioned in the original six-site design. Questions about college housing, correctional facilities, military barracks, shelters, and group quarters resident self-response will need to be tested elsewhere.
Administrative-record gaps. The two sites can help Census evaluate some cases in which records and field observations disagree, particularly around complex or hidden housing. But they cannot establish that administrative-data methods perform equally well across the much wider range of populations and living situations that will encounter them nationally.
Highly mobile communities. Huntsville and Spartanburg will inevitably contain movers and people in unstable housing, but the test was not selected to represent the full range of highly mobile populations. Separate studies should deliberately examine mobility, temporary residence, homelessness, shared custody, and other situations in which Census Day residence is difficult to establish.
The appropriate conclusion is not that Census must restore every original 2026 Test activity. That opportunity has largely passed. It is that the burden on small-scale testing and the 2028 Dress Rehearsal has increased substantially.
Census should publish a crosswalk showing every major objective removed from the 2026 Test, the alternative research that will answer it, and the date by which that evidence will be available for Baseline 2 and later design decisions.
Advocacy priority
Ask Census to identify explicitly what the reduced 2026 Test can no longer establish and how each lost test objective will be replaced.
C. Small-Scale Testing
Census has always planned for small-scale testing alongside the major field tests. That approach becomes more important now that the 2026 Test covers only two sites and a limited set of operations. Census says small-scale testing will continue through 2029 and is intended to provide faster, focused evidence about particular enhancements.
Some questions are actually better suited to targeted testing than to a single large field site.
Rural and remote addressing should continue to be tested through the Locate Address Test, which sends workers into selected rural and remote areas to determine whether they can find residences using coordinates and descriptive location information. Census plans these tests annually through 2028, subject to resources.
Tribal operations require tests developed with tribal governments and conducted in tribal communities. They should include addressing, residence rules, recruitment, communications, language, and the interaction between tribal geography and Census systems.
Language access requires operational testing in communities that actually use the supported languages. Census should test translated questionnaires, bilingual mailings, CQA, field language support, and fraud-verification materials as a connected system rather than evaluating translation quality alone.
Group quarters require separate testing across facility types. A self-response system that works in a university residence hall may not work in a group home, religious residence, military barracks, or correctional facility. Census has already conducted usability and feasibility research on online response for some noninstitutional group quarters, but production procedures will require operational testing.
Homelessness operations and confidential shelters need targeted testing with service providers and affected communities. Ordinary household field tests cannot answer questions about outdoor enumeration, shelter confidentiality, safety, timing, or engagement with people who avoid formal service systems.
Administrative enumeration should be tested deliberately among populations known to have weaker or less stable administrative records. Census should compare the administrative result with independently collected information rather than evaluating only whether an automated system can complete the case.
Accessibility and digital barriers are also well suited to focused usability studies involving people with disabilities, older adults, people with low literacy, users of shared devices, and residents with limited broadband.
Small tests should not become a way to avoid realistic operational conditions. A method that performs well with a few recruited participants may still fail when deployed through actual mailings, field systems, contractors, and staffing structures. Census should identify which findings require confirmation in the 2028 Dress Rehearsal.
Central question
For every major 2030 innovation, where will Census obtain evidence that the method works for the populations most likely to be affected by its weaknesses?
D. Test Reporting
Testing is useful only if its results are available, understandable, and connected to decisions.
Census should establish evaluation measures before a test begins. Otherwise, an operation can be judged against whichever outcomes look most favorable afterward. Study plans should identify the principal research questions, comparison groups, quality measures, and criteria for deciding whether a method should advance.
Measures should go beyond overall response rates and cost. Depending on the test, Census should examine direct versus proxy or administrative enumeration, roster completeness, address discoveries, characteristic completeness, language access, matching errors, field effort, and other indicators of differential quality.
Results should also be examined within relevant communities. A new method may improve average performance while working poorly for rural households, particular language groups, renters, tribal residents, or people in complex housing. Those differences are precisely what a national census test needs to identify before production.
Reports should be released promptly. Findings from the 2026 Test need to be available early enough to inform Baseline 2 and the systems being prepared for 2028. Dress Rehearsal findings must be available soon enough to support real changes before Baseline 3 and peak production.
Census should publish unfavorable, inconclusive, and null findings as well as successful ones. A test showing that a proposed efficiency does not save money, that two methods perform about the same, or that an innovation worsens quality is still a valuable result.
Most importantly, Census should show what happened next. For each major test finding, the public should be able to see whether the Bureau:
adopted the recommendation;
modified the design;
scheduled additional research;
rejected the recommendation; or
accepted the identified risk for another reason.
A short explanation of the rationale would make the research program substantially more accountable.
Advocacy priority
Seek a public research-to-decision crosswalk linking predetermined test measures, results, recommendations, and subsequent changes to the operational design.
E. Decision Lock-In
The public planning timeline does not show the full timeline on which decisions become difficult to change.
Operational design must be translated into requirements before software can be built or services procured. Contractors need specifications. Systems need development and security testing. Training materials must reflect stable workflows. Integration tests require multiple components to behave consistently.
A simplified decision path looks like this:
Each step narrows the range of inexpensive changes.
Requirements are especially important. Once Census has specified what a system must do, that decision begins shaping procurement and engineering even if a later Operational Plan still describes the issue at a higher level.
Procurement can create an additional form of lock-in. As of summer 2026, Census is already conducting market research or solicitations for major 2030 capabilities, including Census Questionnaire Assistance, application modernization, the temporary-worker recruiting and training system, field IT infrastructure, and fingerprinting and badging. Some of these systems must support the 2028 Dress Rehearsal as well as 2030 production.
This does not mean a contract makes every policy decision irreversible. Requirements and software can be modified. But changes become more expensive as development proceeds, and late changes can create schedule, security, integration, and performance risks.
Advocates should therefore map decision deadlines rather than publication deadlines. For each priority issue, they should ask:
When must the relevant requirement be approved?
When does it enter a solicitation or task order?
When does software development begin?
When must it be stable for integration testing?
When is it tested under realistic conditions?
What is the last practical date for a major change?
That information should guide the timing of public comments, congressional oversight, advisory-committee engagement, research requests, and FOIA activity.
The 2028 Dress Rehearsal remains essential, but it should not be treated as the moment when every major policy choice is still open. Its primary purpose is to determine whether the largely developed census works as an integrated system. By then, the most consequential opportunity to influence some components may have already passed.
Credible risk
Public planning documents make the design appear open through 2028 or 2029, while requirements, procurements, and software decisions make important elements increasingly difficult to change years earlier.
Why advocates should care
The meaningful deadline for advocacy is not always the date Census announces a final decision. It is often the date when an ostensibly preliminary decision begins driving contracts and systems.
Where to Look: The 2030 Census Operational Plan Explorer provides the clearest public planning timeline, including Baselines 1–3, the 2026 Census Test, initial Detailed Operational Plans, the 2028 Dress Rehearsal, and final production plans. The 2030 Census Research and Testing page tracks the major field tests and ongoing small-scale testing.
For the changed scope of the 2026 Test, see the 2026 Census Test webpage and Census’s February and March 2026 updates. The original 2024 site-selection announcement is useful for identifying the populations and operational questions the six-site design was intended to cover before the test was reduced to Huntsville and Spartanburg.
Finally, monitor the 2030 Census Procurements page alongside the Memorandum Series and future Detailed Operational Plans. Procurement activity can show when requirements are beginning to harden even before the corresponding operational design is fully described in public planning documents.
****