Census is planning on hiring fewer temporary workers for the 2030 count, watchdog reports

The Census Bureau wants to hire fewer temporary workers for the 2030 count of the U.S. population and rely more on technology and data from other agencies, according to a recent Government Accountability Office report. 

A civil rights nonprofit, however, is concerned about how the Trump administration is developing its census plans and that such changes could undercount certain populations. 

“If people are not represented in the data, they are not represented in our democracy. They're also not represented in the economy. They're not represented in decision making,” said Meeta Anand, senior director for census and data equity at the Leadership Conference on Civil and Human Rights. “It sounds like a wonky and boring topic, and yet [the census] is perhaps the most essential thing we can ensure that we have to get right.”  

Census, in December 2024, approved the first iteration of its high-level operational plan for the 2030 national count, which hasn’t yet been publicly released. The GAO, however, issued a report on Thursday highlighting major changes the agency proposed in its initial plan. 

For one, Census does not intend to hire temporary workers to create its address list of all known living quarters in the U.S., which informs where the agency sends enumeration forms and conducts follow-up visits when there is no response. 

Instead, it is going to rely more on artificial intelligence to analyze satellite imagery and records from state, local and tribal governments as well as commercial sources. 

“The bureau believes that, since the quality of the above data sources has improved sufficiently over time, the bureau will not need to send a large temporary workforce door to door but instead will use trained geographic professionals to verify individual addresses only when needed and as a last resort,” according to the report. 

For the 2020 census, the agency went door to door to verify 35% of the country’s households; however, the practice has been on the decline since 2010. In that year, about 150,000 temporary Census employees canvassed roughly 145 million addresses compared with 32,000 workers who recorded 53.5 million residences in 2020. 

The agency also plans to more frequently use records from other agencies to count people. Census did this to enumerate 450,000 non-responsive households in 2020. As a result, officials predict that they will need to set up fewer local offices. 

“The bureau intends to eliminate large-scale canvassing to build the address list and may be able to enumerate or delete as nonexistent more non-responding households earlier in data collection,” GAO said in its report. “If these efforts are successful, the bureau could greatly reduce the number of temporary census workers needed to follow up, as well as the number of field offices needed to support them.”

There were 248 local Census offices for 2020, which was a decline from 494 such offices in 2010. 

Anand said these changes are not intrinsically bad, but she questions their effectiveness.

“Vulnerable populations are going to be the ones that are in hidden housing units [and] doubled up housing units, and we don't know whether or not these technologies are going to be effective in picking up those housing units,” she said. “So I could say it's automatically bad, but I'm not going to be that person. But I'm going to say I need to see that they have tested to make sure that this is going to capture the population, not in an efficient way…but rather that they're making sure they're capturing the entire population, which is their remit.”

Anand also said she has concerns about an overreliance of technology in conducting the census, which is widely considered to be an authoritative source of data. 

“It makes me start worrying about algorithmic bias,” she said. “If we're using census numbers to build out future algorithms, and we've used an algorithm to build the census, then what strange sort of loop have we started creating in that process?” 

While the number of local Census offices has already been on a downward trend, Anand stressed that the agency found that in the 2020 enumeration it undercounted people who are Black, American Indian or Alaska Native living on a reservation, Hispanic or Latino or reported being another race. 

“There are many factors that fit into that, but I don't think the answer then is to get rid of field offices,” she said. “To me, it feels very similar to other efforts we're seeing across the government to reduce the touch points that the government has with the population.” 

Several agencies have sought to reduce the number of buildings they operate out of in an effort to save money. 

Regarding the use of government records to count people, Anand warned that individuals who are unhoused, formerly incarcerated, live in rural areas or are distrustful of the government are then more likely to be excluded. 

“Has the bureau done the sufficient amount of testing to understand which populations get left out when you do it this way, and if they know there are populations that are being left out, what are the plans to ensure that those populations are going to be counted?” she said. 

In the coming years, Census will test out these plans and potentially change them ahead of the 2030 count. But Anand said she is concerned about the public’s ability to be involved in that process because of the elimination of many of the agency’s advisory committees. Trump officials have shuttered advisory councils at several agencies as part of an effort to shrink the size of the federal government. 

“These committees serve as a place for the public to ask questions of the bureau and for the bureau to sort of stress test their plans and figure out whether or not it'll work,” she said. “So I am currently looking at everything with a lot of skepticism because that back-and-forth is gone.”