The Washington PostDemocracy Dies in Darkness

Will Americans be able to trust the results of the 2020 Census?

April 26, 2021 at 7:30 a.m. EDT

Activists fear battles over the count mean results will be less accurate

The first results of the 2020 Census are about to drop, after encountering so many obstacles that some census experts privately joked that it was cursed.

Over the past four years, the decennial count was hit with funding shortages, partisan interference, legal battles and a pandemic that paralyzed it just as it was getting started, spawning new political and legal battles.

When the initial data — a tally of each state’s population that determines a decade’s worth of congressional apportionment and electoral college votes — comes out this week, and when more detailed data about race and geography comes out this summer, they will be scrutinized by statisticians, politicians and civil rights advocates, many of whom worry the setbacks could result in less accurate results than in previous decades.

“We don’t know how good the quality of the data is,” said Kimball Brace, president of Election Data Services, a political consulting firm specializing in redistricting, election administration, and the analysis and presentation of census and political data. “It is an open question.”

If the results are widely seen as flawed, it could cloud a decade of political reshuffling, federal spending, city planning, academic research, and private and public enterprise that rely on census data.

The bureau plans to release quality metrics the same day as the data release, in part to help allay concerns about problems the hurdles may have caused, said Michael Bentley, assistant division chief for Census Statistical Support in the bureau’s Decennial Statistical Studies Division.

“We are trying to be very transparent, and we do feel good about what we did with the Census 2020 ultimately,” he said on a call with reporters Wednesday. “The quality looks like it’s very comparable to the 2010 Census, but we want everyone that’s using the data to feel that way.”

No census is perfect. The ­once-in-a-decade project of enumerating every person in every household in the country inevitably results in some being missed and some being counted more than once. But generally the results are considered to be accurate enough to use (a notable exception was in 1920, when they set off so much political strife that they were never used for apportionment).

To check the count’s quality, census results are compared with other demographic data, including information from the Census Bureau’s own annual American Communities Survey and a post-census enumeration in which it returns to some areas to double-check its work.

If the state population data is significantly different from projections, that could be a hint of something amiss.

“An initial indication will be is if the reapportionment ends up having some outliers,” said Arturo Vargas, chief executive of NALEO Educational Fund, a nonprofit group that advocates for Latinos participating in the American political process. “If, for example, California loses more than one seat, that would be a red-hot burning flag . . . evidence of a serious undercount in California.”

But experts warn that even if state population totals concur with the projections, that doesn’t automatically mean the apportionment numbers are correct.

“People shouldn’t draw a conclusion — ‘Oh, everything’s accurate!’ No, we won’t know that and that’s probably not likely,” said Terri Ann Lowenthal, a former staff director of the House census oversight subcommittee. The reapportionment data “almost always masks significant over-counting and undercounting at the local and community level,” she said.

That more granular information, which is used for redistricting and which includes race, ethnicity and gender and goes down to much smaller geographic areas, won’t be released until late summer.

The 2020 count’s troubles began soon after President Donald Trump took office. Administration officials almost immediately began discussing adding a citizenship question to the census, and they officially announced it in 2018.

Civil rights groups and experts inside and outside the Census Bureau warned that the question probably would depress the count in immigrant communities and lead to an inaccurate tally, and multiple lawsuits challenged it. The Supreme Court blocked it in 2019, although Trump ordered federal agencies to provide the bureau with citizenship information anyway.

Meanwhile, Congress did not adequately fund the survey, forcing the bureau to reduce its test runs from three locations to one and sparking a separate pair of lawsuits.

As 2020 opened, the bureau struggled to hire 2.7 million temporary workers in a boom time with historically low unemployment. Then in March, just as the first mailers were sent out inviting people to fill out the census form, the coronavirus hit.

Events designed to publicize the census and encourage participation were canceled. College students left campuses, sowing confusion as to whether they would be counted as residents at college or at their parents’ homes. Plans to count homeless people had to be put off. Field offices closed and enumerators hired to go door-to-door from May through July were told to hold off.

Last April, the bureau announced a new schedule that extended the count and its delivery dates, and it asked Congress to approve a deadline extension for apportionment numbers, from Dec. 31, 2020, to April 30, 2021.

But over the summer, Trump issued a memo calling for undocumented immigrants to be excluded from being counted for apportionment. Such an action, which was unprecedented, probably would have shifted power to states with more Republicans and non-Hispanic White people. More lawsuits ensued, calling such an exclusion unconstitutional, illegal and unworkable.

Within days, the government reversed itself on the April 30 deadline, saying the bureau would stick to its original Dec. 31 date for delivering state population totals. The administration said it was doing so because Congress had never approved the later date. But keeping to the original deadline also ensured that Trump would still be in office to receive the apportionment data.

It also would mean compressing the count by a month and cutting in half the time for the post-count analysis conducted after every census. The date reversal generated additional lawsuits.

By fall, the census had become a race against the clock. The end date remained in flux as court rulings were issued and appealed. Wildfires and hurricanes raged through states that lagged farthest behind on the count, displacing residents and eating up crucial time.

Enumerators scrambled to finish counting the approximately one-third of households that had not self-responded. Many described being told to bend rules, compress in-person visits and hazard guesses as to the inhabitants of households to reach the bureau’s goal of a 99 percent count.

In mid-October, the Supreme Court ruled that the count could end, two weeks earlier than the bureau had planned. By then it was unclear how many rules had been broken.

After the bureau’s post-count analysis revealed anomalies in the data, it became clear the ­Dec. 31 deadline was untenable. Still, the administration pressed the bureau to produce the apportionment data before Trump left office, and political appointees in the bureau reportedly pushed career employees to speed up the process and produce a tally of undocumented immigrants in each state, although no such tally exists.

In the end, it couldn’t be done. On President Biden’s first day in office, he issued an executive order that ended the attempt.

But the disruptions had sown skepticism.

“The Census Bureau describes its goal for a successful decennial is to count each person once, only once, and in the right place,” Lowenthal said. “To me a successful census is one that counts all communities equally well, and the disproportionate count of certain neighborhoods and groups has always been the Achilles’ heel of the census. And I think all signs . . . suggest that this census won’t be more accurate than the last one and could be much less accurate.”

Such concerns are pervasive enough that for the first time the Census Bureau has invited an outside task force to observe and assess its processes. The group, from the American Statistical Association, recently began working inside the bureau and will release reports later this year independently of the bureau’s own quality assessments.

“I think the Census Bureau realized that there was a lot of controversy surrounding the 2020 Census and a lot of concerns among a lot of stakeholders about the quality of it, and what they wanted to do was demonstrate that it was high-quality,” said John Thompson, a former Census Bureau director and a member of the task force. “They really wanted an independent look at it.”

In doing so, the bureau is taking the chance that the group will come up with different conclusions from its own internal analysis.

“We’ll all have our fingers crossed that when it makes these reports it’s going to be able to say that the bureau had a very successful census,” said Kenneth Prewitt, a former bureau director who was recently named special adviser to the Census Bureau director. “But they won’t say it if they don’t believe it.”

The news has not been all bad for the 2020 count. For the first time, people could fill it out online, an operation that went smoothly despite initial fears of hacks or glitches.

And compared with past decennial counts, “this decade was the most transparent on record,” said Denice Ross, a senior fellow with the nonprofit group National Conference on Citizenship, noting that self-response data was released daily during the count and census officials blogged about the process.

The operational metrics the bureau plans to release this week will include how the bureau counted a household — through self-response, in-person or proxy interview, or using administrative records, for example — and how those methods were distributed across geographic areas. The bureau plans to release more detailed metrics in May.

(On Thursday, the Brennan Center for Justice at the New York University School of Law published an analysis of how the delays will affect states when data for redistricting is released later this year, several months after their original April 30 deadline).

Bill O’Hare, president of O’Hare Data and Demographic Services, said he is more optimistic than some about the 2020 count. He co-wrote an analysis this month showing that the self-response rate, a good indicator of census quality, was almost the same this time as in 2010.

Like others interviewed for this article, he said he would have liked to have more information on how the bureau closed the gap between the 67 percent who self-responded and the remainder of households, and he worried about the fact that the count took place several months after April 1, Census Day.

“I think these are factors that will affect the quality,” O’Hare said, adding that it might mean the 2020 count doesn’t measure up to the 2010 count. But, he said, “I don’t think it’s going to be a disaster some people seem to be expecting.”

Some fears also were allayed by the change in administration.

“A lot of intelligent people were very worried about the census three months ago, four months ago,” Prewitt said. “The bureau wasn’t in the position to be transparent because the commerce secretary [overseeing the Census Bureau] didn’t want them to be fully transparent. . . . This administration is giving us more leeway to do our jobs the way we’re supposed to. I don’t know what kind of conversation we’d be having if Trump had been reelected.”

More on the census

Here’s how America’s racial makeup has changed over the past decade. You can drill down by address to see how certain areas have shifted.