Texts: US census manager told counters to use fake answers

The texts from an Alabama census supervisor had an urgent tone. "THIS JUST IN ...," one of them began. It then laid out how census takers should fake data to mark households as having only one resident even if they had no idea how many people actually lived there.

The goal of the texts from October, obtained by The Associated Press, was to check off as many households as possible on the list of homes census takers were supposed to visit because residents never had filled out census questionnaires. The supervisor wanted the census takers to finalize cases — without interviewing households — as the Trump administration waged a legal battle to end the once-a-decade head count early.

The texts are the latest evidence suggesting census accuracy was sacrificed for speed as census takers and supervisors rushed to complete a head count last month. Critics contend the schedule was shortened by two weeks so the Trump administration could enforce a presidential order excluding people in the country illegally from the numbers used for apportionment of congressional districts.

The texted instructions said that if two failed attempts were made to interview members of the households, along with two unsuccessful tries to interview landlords or neighbors about the homes' residents, then the census takers should mark that a single person lived there.

"You are to clear the case indicating occupied by 1," said the text from the census supervisor in the small city of Dothan, Alabama.

The texts were shared with the AP by a census taker from Florida who traveled to Alabama among groups of enumerators dispatched to areas lagging behind in the count. The existence of the texts suggests that falsification of census data may be more widespread than previously known.

The census taker who provided the texts asked for anonymity because of privacy concerns and said she refused to follow the texted guidance because she felt doing so would falsify data. She declined to name the supervisor, who was identified only by her first name in screenshots of the texts seen by the AP.

Census Bureau spokesman Michael Cook said the agency is investigating the Alabama case and has not identified any data irregularities. When there appear to be problems with data collection, the bureau can take steps such as revisiting households to improve accuracy, he said.

"We take falsification allegations very seriously," Cook said.

More than two dozen census takers and supervisors have contacted the AP since the beginning of the month, telling similar stories about corners cut in the rush to close cases as the Trump administration sought to end the census before the Oct. 31 deadline set in response to the pandemic.

The most recent cases also include a census supervisor in Baltimore who said that thousands of addresses were manually marked completed without evidence that residents had been interviewed.

The Alabama supervisor in her text included a photo of her hand-written instructions listing the 15 steps she said would allow the census takers to mark in their bureau-issued iPhones that only one person lived in a home without interviewing anyone about the household's demographic makeup or the number of people living there.

The instructions for the census takers in Alabama were sent a week before the Supreme Court made a ruling that allowed the Trump administration to end field operations for the 2020 census on Oct. 15 instead of Oct 31.

The Census Bureau has said it compiled information for about 99.9% of U.S. households in the U.S. during field operations. At the height of the door-knocking phase of the census in mid-August, there were more than 285,000 temporary census takers on the bureau's payroll.

In Baltimore, census supervisor Amanda Colianni said she believes 5,300 cases in neighborhoods she managed were closed prematurely and removed from the door-knocking effort after only one attempt by census takers to interview members of households in mid-to-late September. The Census Bureau was working toward what officials believed at the time would be an Oct. 5 early finish for the count.

Colianni said she does not know why the cases were removed or how they were resolved, though she says it's possible that government administrative records were used to fill in the information gaps when detailed records from the IRS, the Social Security Administration or other agencies existed for the households.

An outside census advisory group warned this month that filling in large numbers of households with administrative data late in the census process suggests no high quality data existed for the addresses. If that had been the case, the group said, it would have been used earlier to save census takers time.

Colianni filed statements with the Commerce Department's Office of Inspector General, the U.S. Office of Special Counsel and a coalition of local governments and advocacy groups that have sued the Trump administration over its attempts to shorten the 2020 census schedule. The coalition's case led to the Supreme Court decision allowing the Trump administration to end the head count.

The coalition's lawsuit in San Jose, California said the deadline for finishing the count was changed from the end of October to the end of September to ensure that number-crunching for the census would take place while President Donald Trump was still in office, no matter the outcome of the presidential race.

That could guarantee the enforcement of an order Trump issued in July seeking to exclude people in the country illegally from numbers used to determine congressional seat distribution. Trump's order has been found unlawful by three courts — in New York, California and Maryland. The Justice Department is appealing.

Whether the Census Bureau can meet a Dec. 31 deadline for turning in the apportionment numbers to Trump is now in jeopardy after the agency said Thursday that it found anomalies in the data during the numbers-crunching phase.

The coalition contesting the early end to the count is seeking to extend the numbers-crunching phase of the census from the end of December to the end of next April, especially since the Census Bureau is relying on large numbers of administrative records to fill in gaps from data collection.