Updated November 18, 2021 at 12:58 PM ET
The U.S. Census Bureau is again extending a final round of door knocking into early 2022 for a key survey that is expected to help determine the accuracy of last year's national head count.
The change is the latest in a series of delays to the little-known but critical follow-up survey. The disruptions have raised concerns about whether the bureau can produce important indicators about who was missed and which groups were over- or undercounted in a census that was upended by both the coronavirus pandemic and interference by former President Donald Trump's administration.
In the past, the results of the Post-Enumeration Survey have been factored into some of the population statistics that guide how federal funds are distributed to local communities. The survey also helps the bureau determine how to better carry out future once-a-decade counts that are used to reallocate each state's share of congressional seats and Electoral College votes.
What was expected to be a monthlong operation for gathering information on housing units starting in late October is now set to begin on Nov. 29 and end in March, Judy Belton, the bureau's assistant division chief of special enumerations, confirmed to NPR during a virtual public forum before the bureau's announcement. As NPR had reported, that operation was previously expected to wrap up a month earlier in February.
"We adjusted the start date and operational length as a result of the impact of the COVID-19 pandemic on the schedule of the preceding census operations," the bureau said in a statement in October about changes to the Post-Enumeration Survey, which does not involve college dorms, prisons or other group-living quarters and is not conducted in remote areas of Alaska.
About 1,100 of the bureau's field representatives — who, like all federal government employees, must be fully vaccinated for COVID-19 by Nov. 22 — will try to interview people at some 14,000 housing units while wearing masks and practicing social distancing.
The bureau has already finished gathering the demographic information it needs to produce the over- and undercount rates by population group, which are expected to be released within the first three months of next year.
"All the data collection and matching for the report in the first quarter of 2022, that's all done. And none of the housing-unit work will impact those reports," Timothy Kennel, assistant division chief of statistical methods, said at the forum organized by the National Academy of Sciences Committee on National Statistics.
Still, some census watchers outside the bureau say they're worried the difficulties of conducting in-person interviews during the pandemic could limit the usefulness of the survey's findings.
Were people of color undercounted by the census?
Decade after decade, the U.S. census has overcounted people who identify as white and not Latino, while undercounting other racial and ethnic groups. That unevenness often translates into inequities when census data is used to redraw voting districts and inform research and planning.
"I'm just worried that we're going to have a starting point for the next 10 years of enumeration counts that undercount people of color," Robert Santos, the president of the American Statistical Association who has been confirmed by the Senate to be the next Census Bureau director, told NPR in an interview before his April nomination by President Biden.
Santos, who is set to be sworn in early next year, added that not having reliable PES results runs the risk of baking racial inequities into other government statistics that rely on census data.
"I don't think that that's acceptable," Santos said.
Many people don't want to talk to Census Bureau workers because of COVID-19
Door knocking for the PES, which takes place in multiple phases, started as originally planned in January 2020. It was supposed to wrap up by mid-2021.
But COVID-19 quickly intervened.
Lower-than-expected levels of participation from the public in the final months of 2020 led the bureau to add another round of interviews about people's demographic information.
Those early response rates were hurt by a devastating new reality — "people don't want to open their doors to talk to a stranger during a pandemic," the bureau acknowledged in a March presentation to its scientific advisory committee. That raises the risk of the bureau missing certain people not only in the census, but also in the follow-up survey that determines who was not counted.
"When the PES comes out, I worry about its possible quality because the survey had to be conducted in the teeth of the winter COVID surge," says Connie Citro, a senior scholar with the National Academy of Sciences Committee on National Statistics who served as the senior study director of the committee's evaluation of the 2000 census.
Could COVID-19 stop the bureau from releasing survey results?
The delays to the survey make it harder for the bureau to collect accurate data. Tallying for the 2020 count has been over for nearly a year, and some people interviewed for the PES may have a hard time remembering exactly where they were living on Census Day, which was April 1, 2020. People who moved during the pandemic may not know who used to live at their current address.
And many households are experiencing census fatigue.
"They're not ready to see another census person come and knock at the door," Bob Leibowitz, a Census Bureau survey technician in the New York region, told the committee's panel on 2020 census quality during a virtual meeting in August.
Data quality issues brought on by the pandemic have already forced the bureau to cancel a release of American Community Survey results this year and replace them with "experimental" estimates. Some census watchers fear the Post-Enumeration Survey may be headed toward a similar fate.
There are early signs of a likely undercount of Black people
In the meantime, researchers outside the bureau have been comparing the latest census numbers with a set of benchmark data based on birth and death certificates, Medicare enrollment files and other government records about the country's residents.
"It does look like the 2020 census had some undercount problems for some groups," says Citro, who recently conducted an independent analysis that was not part of her work with the committee.
Using publicly available data and a method that is different from what's used for the PES, Citro estimates that nationally, the 2020 census may have produced a net undercount rate for Black people similar to what the bureau's PES estimated for the 2010 count (2.05%) or more than two times as high (4.36%).
"The Census Bureau did just a heroic and really outstanding job, but they faced a combination of circumstances for conducting a census that was unprecedented in our history," Citro, who once worked as a social science analyst at the bureau early in her career, says of the pandemic and the push by Trump officials to end counting early.
Children were likely undercounted in 2020, too
There are also signs the 2020 census likely didn't correct a decades-long flaw with the national tally: undercounting children.
"All the evidence I'm seeing from the 2020 census suggests that that's going to be a continuing problem," says Bill O'Hare, a demographer and former research fellow at the bureau who wrote the book The Undercount of Young Children in the U.S. Decennial Census.
O'Hare, who is currently consulting with the Count All Kids Campaign, estimates that the net undercount rate for children bumped up to 2.1% last year, while adults had a net overcount rate of less than a percent for the 2020 census, according to a report released this week. The report also cites preliminary estimates by Citro that suggest the net undercount rates for Black and Latinx children were about double that for all children.
Still, both Citro and O'Hare say they're waiting for the PES results to reveal a more comprehensive look at the count's accuracy.
Jeri Green, a former senior adviser for civic engagement at the bureau, remains concerned that the 2020 census — which the agency recently estimated to cost around $14.2 billion — will repeat the undercounting of Black people and Latinos, as well as Native Americans who live on reservations, in the 2010 census.
"The American taxpayer is being cheated, the congressional appropriators who funded the census also are not getting their dollars' worth, if the PES and the undercount are not accurate," Green says. "And we have to live with this for the next 10 years."
Copyright 2022 NPR. To see more, visit https://www.npr.org.
