WASHINGTON — The Census Bureau, long the gold standard for nonpartisan probity and statistical rigor in the federal government, is rushing toward the close of the most imperiled and politicized population count in memory with two huge issues in dispute.
The first is the overall accuracy of a hurried count, buffeted by the coronavirus on the one hand and partisan interventions by the White House on the other. The second is whether the use of that count will result in figures for congressional reapportionment shaped by political considerations instead of an objective count of all the nation’s residents as the Constitution requires.
For all its problems, the Census Bureau has mounted a dogged effort in recent days to fill in some of the count’s many gaps. It has plucked more than 16,000 census takers from places where it considers the count completed and dispatched them to states where the response is lagging, notably in the Deep South. The deployment is well above those of past censuses, and appears to reflect the bureau’s use of its new digital capabilities to post workers where they are most needed.
But it comes after the bureau has curtailed counts in many areas, despite a federal court order to continue them until the end of October, and a barrage of complaints by census takers and others that tallies are inaccurate and incomplete.
The most consequential decision appears yet to come. For decades, the Census Bureau has calculated the allocation of seats in the House of Representatives with its fresh population totals. The process has been walled off from partisan influence, its methodology set by law and its results available for anyone to double-check.
But this time, court documents indicate that, as one Justice Department legal filing stated, the bureau “will provide the President with information” for the calculations.
That has further fueled fears that partisan politics will taint a reapportionment process that the White House already has vowed to recast by removing unauthorized immigrants from state-by-state population totals. Analysts say that probably would increase Republican ranks both in the House and in state legislatures.
Neither the Census Bureau nor the Commerce Department, its overseer, responded to questions about the bureau’s role in the next reapportionment.
“The only interpretation I can give you is that the Census Bureau itself will not control the allocation of population numbers for apportionment,” Kenneth Prewitt, a Columbia University professor who ran the bureau during the 2000 census, said. “The Census Bureau is simply not in charge of what it has been in charge of.”
He added: “I hate the word unprecedented, but this is an unprecedented situation.”
Depending on how it was performed, a reapportionment directed by a Trump White House could push the census into uncharted political and legal territory. If Joseph R. Biden Jr. wins the presidency, he could face limited options to reverse it, potentially leaving the fate of the census and reapportionment in the hands of the House.
“We don’t know what would happen if the House of Representatives rejected the data as being incomplete or inaccurate,” said Jeffrey M. Wice, a senior fellow at New York Law School and staff chair of the redistricting and elections committee at the National Conference of State Legislatures. “There are so many different possibilities that it’s not possible to predict.”
Those scenarios will turn on the outcome both of legal battles and of the November election. But the political shadow cast over the 2020 count has created more urgent problems.
Even before the tally reaches its scheduled end on Oct. 31, leading experts are questioning the credibility of its population totals. Perhaps the major reason is the sudden order by Commerce Secretary Wilbur L. Ross Jr. in August to end the 2020 count a month early, on Sept. 30, further upending a census schedule already mangled by the pandemic.
Mr. Ross’s order, since struck down by a federal court, came despite warnings by career census officials that a shortened census would be deeply flawed. Census experts said it had one purpose: to ensure that population totals could be error-checked and processed quickly enough to get them to President Trump before the end of his first — and possibly last — term.
The Commerce Department appears determined to do that, given that the appeals court ruling also gave it a green light to try to do so, overriding the lower court’s order to ignore a Dec. 31 deadline and continue the processing into the spring of 2021.
Processing and checking census data usually takes five months. In a Supreme Court filing, the Justice Department argued that it would cut that almost in half, to three months or less, if the court would order the count stopped — a pace census experts say would require stripping away quality checks essential to an accurate count.
To meet the Sept. 30 deadline Mr. Ross had ordered, the Census Bureau scaled back some verification procedures. And it eased reporting rules, allowing census takers to accept less reliable “proxy” information — such as a neighbor’s word on who lives in an uncounted house — earlier.
That allowed the bureau to report on Saturday that it had tallied 99.9 percent of all the nation’s households. That so-called completion rate, a headline feature of the administration’s legal arguments to end the count early, is new; past censuses used more conservative measures.
But experts say it may mask gaping holes in the count’s coverage. While the bureau has not detailed what households the measure covers, it does not represent households that have actually filled out census forms. Rather it appears to include those checked off the list of uncounted households by any means, however inaccurate.
Nor does it show variances between and within individual regions. The bureau carves the country into 248 “area census offices,” none of them yet fully completed and some lagging badly, including an 88.1 percent rate in the Shreveport office, which covers much of Louisiana.
Because the hardest-to-reach households are disproportionately occupied by minorities, young people and the poor, a rushed census compiled with more guesswork would produce especially inaccurate counts of those groups, experts say.
“You don’t know what makes up the 99.9 percent,” said John Thompson, a former career census official who was director of the bureau from 2013 to 2017. “It’s not how many households you complete. It’s the quality of the completions.”
Consider Westlake Village, with more than 3,000 apartments and townhomes south of San Francisco. Counting there has been difficult, because building entrances were often locked.
But in late September, an order appeared on the census takers’ government-issued iPhones: Stop interviewing Westlake residents to get details like ages and races, it said; instead, enter an estimate of a household’s occupants provided by the complex manager.
“They only had the names of the people who signed the lease,” said one census taker, who spoke on the condition of anonymity because the bureau prohibits speaking with reporters. “You didn’t know if they had roommates or family — nothing.”
Supervisors stopped interviews after deciding to wind up census-taking in that area, standard practice when a deadline looms and the remaining uncounted households seem unreachable. In this case, however, a federal judge had barred the Census Bureau from ending the count on Sept. 30, “including but not limited to winding down or altering any census field operations.”
A federal appeals court later refused to stay that order, and the Census Bureau ordered regional offices to obey it. But it is clear that in the rush to finish, some officials misunderstood the order or were told to wrap up counting regardless.
Court documents recount a flood of emails from door-knockers saying they had been ordered to “close down remaining cases by whatever means necessary” and that “we are wrapping up everything tomorrow.”
In Philadelphia, a census taker provided The New York Times with a text message from early October in which a supervisor stated that “there was no discussion regarding an extension” of the count and that the court order “doesn’t apply to us.”
The bureau has said such messages either were mistaken or reflected normal procedures in areas nearing completion. But Judge Lucy H. Koh of the Federal District Court for the Northern District of California has suggested that the Commerce Department could face sanctions for “repeated violations” of the directive.
Deliberate or not, the push to finish the 2020 count has left the census substantially complete, at least technically, even though the bureau is legally bound to continue counting for nearly three more weeks. Some enumerators said in interviews that they have little or no work to do, but some counting has continued.
Census forms filled out online, by phone or by mail — the most accurate part of the count — can still be filed until Oct. 31. About 67 percent of the nation’s households have responded on their own so far, eclipsing the rate in the 2010 census.
Yet even as the Census Bureau boasts of nearing a 100 percent completion rate, the stated rates in scores of census offices have been rolled back in recent weeks, said Steven Romalewski, the director of the mapping service at the City University of New York’s Graduate Center. That suggests that the bureau is still finding places where its count is problematic and sending door-knockers back to do more work, he said.
Which makes outside experts even more skeptical that 99.9 percent is anything more than a public-relations gimmick to bolster the case for ending the count early, coronavirus or not.
“This happy talk in the context of the pandemic is completely misplaced,” said Margo J. Anderson, a census expert and historian at the University of Wisconsin-Milwaukee. “Aside from whatever manipulation the Trump administration is up to, a more honest bureau right now would be saying, ‘You know, we’re having a really hard time.’”