PRB

This Tuesday, the U.S. Commerce Department published a federal register notice with details about a 2026 operational test in support of the 2030 Census. The notice reveals significant reductions in the scope and rigor of census preparation activities. While this test is not intended to be a full end-to-end “dress rehearsal” for the decennial census, operational tests are designed to surface problems with field operations, technology, data collection, and processing before the census is conducted nationwide.

These tests can unearth real problems before they take root. Case in point: After the end-to-end test before the 2010 Census uncovered failures in the handheld devices designed to collect data electronically, plans to use over half a million of the devices were ultimately scrapped—preventing errors from being multiplied on a national scale. Beyond cost and quality concerns, a bad census has far-reaching consequences, affecting congressional representation, federal funding distribution, and policy decisions for the next decade.

The Census Bureau’s planned 2026 operational test for the 2030 Census raises several concerns:

1. Only Two Test Sites Are Proposed—and Both Are in the South

In July 2025, the Census Bureau published their 2030 Census Operational Plan, the guide that tells the public how they will conduct the census. The original plans were to conduct operational tests in six locations: western NC, Spartanburg, SC, Huntsville, AL, western Texas, tribal lands in Arizona, and Colorado Springs, CO. The newly released notice reduces the testing sites to just two: Spartanburg and Huntsville.

Spartanburg and Huntsville are not representative of our diverse country. Both are mid-sized southern cities where over 90% of households speak English as their primary language. Huntsville has experienced tremendous population growth in recent years, which may help the Census Bureau identify gaps in household records there. But limited testing to these two sites would leave other critical populations completely unexamined.

Rural households, tribal lands, and people with limited English proficiency are typically undercounted in the decennial census and were stated as populations of interest in the Census Bureau’s 2030 Operational Plan. By eliminating Western North Carolina, tribal lands in Arizona, Western Texas, and Colorado Springs from testing, the Census Bureau has abandoned its stated commitment to testing operations in the very communities its plan identified as most vulnerable to undercounting.

There are also no stated plans to test a non-English version of the 2030 Census questionnaire in the new notice. Without such testing, the Bureau will have little insight into how proposed systems and methods will perform for non-English-speaking households.

2. USPS Staff Will Perform Tasks Typically Done by Census Workers

Calls to reduce federal spending are common, but the decennial census requires significant resources to achieve a full and accurate count. The notice states that the Census Bureau plans to test whether certain operations can be performed by U.S. postal workers to “increase efficiency.” But this suggests that the Commerce Department doesn’t understand the complexity of either postal workers’ or census enumerators’ jobs.

Setting aside whether this is even possible—postal and census staff work for different agencies—the notice leaves roles and responsibilities unclear. Would postal workers simply deliver census materials, or actually conduct the enumeration? Such critical distinctions affect the feasibility of this entire approach.

Beyond interagency coordination concerns, this proposal would hamper data collection in rural and remote areas, which are already famously hard to count. If the Census Bureau plans to conduct a USPS-based enumeration approach, the decision to exclude a rural area from the proposed testing sites is a tremendous shortcoming.

3. The ACS Is Not the Census

The notice also reveals that the Census Bureau plans to use the American Community Survey (ACS) questionnaire for this operational test, rather than testing with the decennial census form. The Census Bureau conducts the ACS continuously across the country, producing a geographically and topically rich data source. But it is not equivalent in design or protocol to the decennial census.

A key distinction is that the ACS is based on a sample of the U.S. population, unlike the census, which aims to count every person at a single point in time. ACS survey methods and field approaches rarely change, and its most recent methods panel test was slated to test emerging issues related to questionnaire timing, response modes, contact strategies, and question changes.

The decennial census is its own animal—often cited as the largest non-wartime hiring effort in the country—and so complex that it cost $13.7 billion to conduct the 2020 Census. It requires opening local field offices, conducting nationwide advertising and outreach campaigns, and processing vast amounts of data.

Without testing the entire census infrastructure, we can’t identify critical failures before they multiply into costly problems at go time, including issues with internal systems and the larger survey ecosystem. The new notice emphasizes testing hiring and payroll processes—not the full operation of the census— leaving a lot of the intricacies of data collection, for example, unexamined.

4. Critical Questions About Data Collection Are Unanswered

Rather than detailing how data collection will be tested, the notice focuses primarily on testing backend operations like hiring and infrastructure. For example, it doesn’t specify whether the census questionnaire itself will be tested. But given the choice to use the ACS methods panel, it’s unlikely the Bureau will be testing anything other than the ACS survey and its questions.

The notice also focuses exclusively on testing housing units, suggesting they are excluding tests of group quarters (GQ) housing such as colleges, universities, nursing homes, correctional facilities, and military barracks. While this population is relatively small on the national scale, it includes some of the most vulnerable U.S. populations, where high-quality data is sorely needed.

Finally, the notice does not describe how administrative records will be tested or integrated, despite their central role in the 2030 Census Operational Plan. Administrative records from sources like the Internal Revenue Service, Social Security Administration, and other federal agencies are essential tools for filling data gaps in the census, particularly as response rates to traditional data collection methods decline. However, these records are known to underrepresent some vulnerable populations, raising questions about how this increasingly important component of census operations will be evaluated if it isn’t explicitly tested.

A Census Untested Is a Census at Risk

Overall, this notice speaks to the continued attempts to devalue our federal data systems. With every passing decade, the census becomes more complex, and robust testing becomes more—not less—important.

The Census Bureau should expect considerable pushback from research and policy communities. The scaled-back approach announced in this notice lacks the scientific rigor necessary to ensure an accurate count and raises serious questions about whether operational and political considerations are taking precedence over data quality. Stakeholders should make their concerns known during the public comment period, and Congress should demand a more comprehensive testing plan that adequately prepares for the complexities of the 2030 Census. Decisions made now will shape the quality and credibility of the final count.

The public comment period for the Department of Commerce federal register notice closes on March 5, 2026.