'The hardest census': Historic challenges could mean more errors in 2020 data

In March 2020, hundreds of thousands of people were preparing to fan out across the country to knock on doors to collect data for the federal government’s pivotal, once-a-decade population count: the Census. Then, the pandemic struck. “Everyone else is bulk buying toilet paper and my first response was, ‘Oh my god what are they going to do with field work?’” said Beth Jarosz, a senior research associate at the Population Reference Bureau, a non-profit demography organization. “This was the hardest census the U.S. has ever had to do, and there were times when I wondered whether we would be able to do it.” After unprecedented challenges, the U.S. Census Bureau plans to publish more detailed information from its 2020 count on Thursday afternoon, revealing population numbers at the county, city and census tract level, including race, age and housing information. The data will be used by states to draw political lines for federal and state offices and it will shape the flow of more than $1.5 trillion in federal funding. No census is perfect, experts said. It’s meant to count every person in the nation once, and in many years it’s fallen short of that goal, in particular by undercounting children and people of color. But a number of factors in 2020 have scholars and government watchdogs worried the 2020 census is at risk for more error than usual. The pandemic changed multiple data collection processes. Debate over adding a citizenship question to the census created fear among undocumented immigrants and other groups that is believed to have lowered participation. Data processing was completed under tight timelines and with changing administrative directives. The agency is under-funded, according to some researchers. And a new privacy initiative has obscured some data to the public. The Census Bureau says ensuring the quality of its results is built into every step of conducting the census. “Once we release the results, we continue to do an even deeper dive to evaluate the quality of the census,” the Bureau said on its website. “When we release numbers, we make sure they meet our quality standards.” Experts agreed the Bureau is deeply committed to data quality and has numerous checks in place. But 2020 delivered a historic strain on the agency. Just how far off could the results be? The answer to that question is likely to take years to unravel, multiple experts said. Zack Almquist, assistant professor of sociology at the University of Washington, said it could take another 10 years — until the 2030 census — to really understand it. “There are a number of scholars who spend a bunch of time trying to understand the errors in the census in a normal period,” Almquist said. “This [census] is inducing such a different set of errors this year, combined with the data collection strategies, that pulling apart exactly how the directions go and where things happen is going to take a lot of work.” Response hesitancy In most census years, some population groups are undercounted by the census and some groups can be over-represented in the data. Generally, children under 5 years old, people of color, people in dense urban areas or extremely rural areas and people in transient housing are most likely not to be counted. Well before the 2020 census count started, former President Donald Trump pressed to add a question to the survey asking whether the respondent was a citizen. The U.S. Census is intended to count all people who live in the nation, whether or not they are a citizen, legal resident or living in the country with undocumented status. Multiple groups sued to block Trump from adding a citizenship question, mostly motivated by concerns that the question would drive down participation among undocumented immigrants who might be concerned their response would be used to deport them and causing an undercount of those groups. The U.S. Supreme Court ultimately ruled in June 2019 that the Trump administration could not include the question. “Even though the question wasn’t on the census, many people still believe that it was,” Jarosz said. “[There was] also worry that Hispanic, Latino or Asian communities might be targeted anyway. That even just responding to the race and ethnicity question might be enough to get you on [Immigration Customs and Enforcement’s] target list.” In a challenge for all surveyors, it’s becoming a general trend that some people are less willing to share information with the federal government, said Jan Vink, extension associate with the program on Applied Demographics at Cornell University. Some respondents may skip the entire census or individual questions within it for that reason. Pandemic counting Then COVID-19 swept the globe. The Census Bureau suspended door-to-door enumeration for three months in the spring of 2020 and when it resumed the process — a key part of counting those hard to reach populations — it shortened the work by two weeks. Although the census has other fall-back methods, this change could have contributed to an undercount of some groups, especially populations that were facing natural disasters like wildfires and hurricanes in 2020, Vink and Jarosz said. At the same time, the Census Bureau deployed new techniques for the first time to ensure more people were counted, even with the pandemic limitations. The 2020 Census was the first in which the Census Bureau allowed people to complete the survey online. Responding online became the most common way to answer the survey and it may have helped drive up overall participation and participation among younger people, said Jennifer L. Van Hook, professor of sociology and demography at Penn State University. The Census Bureau cancelled in-person events, but advertised the survey in online, radio and other media ad campaigns to urge Americans to participate. The Census Bureau also collected more information than ever before from colleges and universities about their students to ensure students were counted as residents of the city where their school is located. Many college students returned home when the pandemic struck so later the Bureau had to ensure these students were not counted twice — at their home and at their schools. It’s possible some college students were missed depending on the quality of information shared by schools or they were over-counted. The AP reported the Census Bureau collected no data for as many as one in five college dorms, nursing homes and prisons, requiring the agency to use statistical methods and backstopping to fill in the gaps. Additionally, the Bureau used administrative records like tax filings and Medicaid documents as a further way to learn about people who did not self-respond to the Census. That’s a good way to catch non-responders, but it’s more likely to capture wealthier, white and older people than poor, young people of color, Jarosz said. Data collection was “chaotic” during the pandemic, but the Bureau did make big efforts to count everyone it could, Van Hook said.“There were genuine efforts to improve what was going on despite the headwinds they were facing,” Van Hook said. “There is reason to be concerned. [But] I’m hopeful that many of the other things that the Census Bureau did to make up for that will have helped.” Data cleaning After data collection, the Bureau embarked on cleaning, processing and analyzing the data. Trump ordered the agency to meet a Dec. 31 legal deadline to deliver reapportionment totals to Congress, despite the delays it faced in the spring. That would have left Census staff scrambling to get the work done in half the time they usually took. One reason Trump pressed the agency to get the work done quickly, experts said, was he wanted to eliminate undocumented immigrants from the state populations for apportioning seats for the U.S. House of Representatives. Although that effort was not ultimately accomplished, Census Bureau staff worked on it, transferring resources away from other Census work, according to Van Hook, who serves on the Census Scientific Advisory Committee. “All of sudden, all of these people and staff were shifted to a whole different project, which hadn’t been vetted and the quality was unknown,” Van Hook said. “They were squishing that all into a matter of months. It was just a little crazy.” Eventually, the Census Bureau blew past the December deadline and kept working on the data until its first release of apportionment data in April. Moreover, as it cleaned up the data, the Census Bureau did find problems. In Nov. 2020, Census Bureau Director Steve Dillingham said “During post-collection processing, certain processing anomalies [had] been discovered.” The next month, the U.S. House Oversight Committee obtained internal documents from the bureau showing 15 different data errors, impacting more than a million records and every state, U.S. Rep. Carolyn Maloney, D-N.Y. said. The documents laid out an 11-step intensive process to fix the errors, Maloney said, but rushing that procedure could lead to further problems. Jarosz said she is confident the Census Bureau corrected these errors before any data was released by virtue of the fact that the problems were identified and announced. “The errors that were not known are a bigger concern than those that were on that list of anomalies,” said Jarosz. Finally, when releasing the data to the public, the Census Bureau plans to use a new technique called differential privacy to obscure some of the numbers in low population areas so they could not be used to identify individuals. “In the beginning, [I was] very afraid of the usability of the data,” Vink said. Later the Bureau made some improvements, but researchers still noted some concerns about how differential privacy will impact how some data can be used accurately. A dated snapshot Demographers emphasized that even with errors, the census data is important; it’s the best possible picture of the composition of the nation. The Census is a snapshot of the population on April 1, 2020, however, and the population has changed a lot since then — even more than it usually does in a 16-month period due to the pandemic. The nation saw more deaths, fewer births and migration out of cities, among other trends, demographers agreed. The 2021 American Community Survey released by the Census Bureau next year will be able to capture some of these trends. “The census data is still useful because the vast majority of people are still alive,” Jarosz said. “Most people are still where they were in the spring of 2020.”

'The hardest census': Historic challenges could mean more errors in 2020 data

Don’t Miss This Deal!
26 WEEKS FOR 99¢

Enjoy the many benefits
of being a subscriber.

Keep reading with
unlimited digital access.

SALE ENDS AUG. 22 | CANCEL ANYTIME