The U.S. Census Bureau this week announced it is dramatically shrinking the size and scope of a trial run this spring for the 2030 census. Experts worry the changes—notably fewer test sites, less outreach, and an English-only online questionnaire—will make it harder to test proposed improvements and could lead to a significant undercount of the country’s population in 2030.
The cutbacks mean “you’re no longer focusing on how to do a better job of counting the folks who were historically undercounted, to get a more accurate census,” says statistician Robert Santos, who led the Census Bureau under former President Joe Biden. “Instead, you’re simply testing for efficiency and some new technology,” says Santos, who resigned 3 weeks after President Donald Trump took office.
To conduct an accurate decennial national head count, the Census Bureau must first identify all occupied housing units, even those with no official address. Then it must count everyone living in those units—and avoid counting anyone twice.
That’s a huge effort—the 2020 census cost $13 billion—and one that requires massive public outreach and the cooperation of every resident. Santos says the goal of conducting a field test 4 years before the next census is to develop “better messaging, better approaches to use at someone’s door, better partnerships with community leaders to generate the type of trust that’s necessary for residents to feel comfortable participating in the census.”
Many demographers think the latest changes go in the opposite direction, starting with the decision to drop four of the six field sites announced in 2024. The six sites were selected based on 13 housing and demographic characteristics that have traditionally contributed to an undercount of poor, rural, immigrant, minority, and geographically isolated communities. The list of factors also includes fast-growing areas with large, complex housing units and group quarters such as colleges, prisons, and military bases.
None of the original sites satisfied all 13 criteria—one in western North Carolina did best, addressing eight. But together they covered all the bases. In contrast, the two remaining sites—Spartanburg, South Carolina, and Huntsville, Alabama—satisfy only three and six of the criteria, respectively. For example, neither city contains any tribal lands, unlike three of the dropped sites—western North Carolina, Arizona, and western Texas. The Texas site also satisfied another criterion—spotty cellphone and internet service. Colorado Springs, Colorado, the fourth dropped site, featured lots of new housing.
The revised field test will also be much smaller. Instead of some 631,000 residents, the Census Bureau now expects to target a pool less than one-quarter that size.
The Census Project, a nonprofit advocacy group, notes that the Trump administration reduced the number of tests before the 2020 census, citing budget constraints. It says that change, along with the pandemic, contributed to a much higher undercount than might otherwise have been the case. “The 2026 test was touted as a response to 2020 lessons learned and to help the Census Bureau improve the accuracy of the country’s upcoming once-a-decade head count,” the group said in a statement after the revised plan was announced.
The Census Bureau declined to answer questions from Science about the rationale for the changes. A one-paragraph statement on its website says the agency “remains committed to conducting the most accurate count in history for the 2030 Census and looks forward to continued partnerships with local communities.”
But census advocates question that commitment. Instead, they worry the White House has overridden the Census Bureau’s best scientific judgment to serve some political purpose—the results are used to allocate federal dollars and draw congressional districts. “Had the people at the Census Bureau [when I was director] believed they could gain enough insights from only two sites, that’s what they would have proposed,” Santos says. “The reduction from six [sites] tells you something about who’s making decisions,” he says.
Another change that troubles demographers and civil rights groups is the elimination of Spanish and Chinese versions of the online questionnaire being tested this spring. The 2020 census was the first time people could respond via the internet, and more than two-thirds of the country chose that option—offered in 12 languages other than English—right off the bat.
The Census Bureau says it is hoping to make that option even more popular in 2030 by adding a QR code to call up the census form and improving customer service for those having problems completing the 10 questions online. But an English-only version will shut out many potential users.
The main cost of the census is tracking down and obtaining answers from residents who did not respond to the initial outreach and reminders. The field test is the only chance for the Census Bureau to pilot all the planned improvements it hopes will make that follow-up exercise faster, better, and cheaper in 2030. Demographers are especially concerned about one new wrinkle: testing the use of the U.S. Postal Service (USPS) to nab hard-to-reach residents.
It’s been suggested before, but few demographers think it’s a good option. A 2011 report by a government watchdog agency said it would be much more expensive—USPS employees are paid triple what the Census Bureau pays its army of temporary enumerators. Postal carriers already have a full-time job, Santos notes, and aren’t trained to interact with residents in the same way that census enumerators do. Some experts also think relying on the carriers would undercut the agency’s promise to respondents of absolute privacy.
“A lot of people, especially in rural areas, know their postal carrier and might not be comfortable giving out personal information to them,” says Allison Plyer, chief demographer for the Data Center and former chair of the agency’s Scientific Advisory Committee, which was disbanded last year. However, Plyer and others do think USPS can play a role in helping update the master list of addresses that the Census Bureau uses as the basis for each census.
The Census Bureau did not explain its rationale for any of the changes. And that lack of transparency worries those hoping the 2030 census will be as accurate as possible. “This announcement raised a lot more questions than it provided answers,” says one veteran city planner who asked not to be identified. “And that’s not a good sign.”
Jeff Mervis tries to explain how government works to readers of Science.
Every day, the news staff of Science magazine and our contributing freelancers bring you top stories from the world of scientific research and science policy. Our offerings include breaking research news, ScienceInsider (news and analysis from the world of science policy), a weekly podcast, and Sifter, a blog that points you to the best science stories on the web. We also post a weekly podcast. And you can now find our weekly news content from Science magazine included in our daily news feed, on our category and collections pages, and on our author pages.
Click here for more information about the News from Science subscription and to purchase a subscription.
Nuclear Arms Control Era Comes to End Amid Global Rush for New WeaponsDavid E. Sanger and William J. BroadNew York TimesBeijing, Moscow and shaken American allies are seeking new warheads as President Trump ends more than a half century of nuclear arms control with Russia.February 5, 2026
