“Overall, the results are less accurate than in 2010,” she said.
By the bureau’s estimates, the 2020 tally incorrectly counted 18.8 million residents, double-counting some 5.2 million people, wrongly including another two million and missing others entirely, even as it came extremely close to reaching an accurate count of the overall population. Many of the people whom the census originally failed to count were picked up and added to final census totals through a process called imputation — a statistical guess, using complex algorithms, of who was living in places that census takers could not reach.
Those imputations, however, did not guess the race or ethnicity of missing residents. That partly explains why the grand total count of residents was substantially accurate, even as many minorities were missed.
Although the bureau did not say how many people it missed entirely, they were mostly people of color, disproportionately young ones. The census missed counting 4.99 of every 100 Hispanics, 5.64 of every 100 Native Americans and 3.3 of every 100 African Americans.
In contrast, for every 100 residents counted, the census wrongly added 1.64 non-Hispanic whites and 2.62 ethnic Asians.
Young people in general were undercounted: by 2.79 percent for children under 5, and, among those ages 18 to 29, 2.25 percent for men and 0.98 percent for women. Older residents were overcounted, most significantly by 2.63 percent for women 50 and older.
The estimates released on Thursday — in essence, a statistical adjustment of totals made public last year — are based on an examination of federal records and an extensive survey in which the bureau interviewed residents in some 10,000 census blocks, the smallest unit used in census tabulations. Bureau experts then compared their answers to the actual census results for those blocks.
The survey enabled the bureau to estimate how many residents it missed entirely in the 2020 count, how many people were counted twice and how many people — such as short-term visitors and people who died before or were born after the official April 1, 2020, census date — were counted mistakenly.
