Census misses may have cost Florida and Texas in redistricting
Report estimates undercounts in those states were far larger than the margin for gaining another seat
Census miscounts in Texas and Florida may have been big enough to cost those states a congressional seat this decade, and may have helped states such as Rhode Island keep a seat, according to data in a Census Bureau report released Thursday.
The report estimated how much the 2020 census either overcounted or undercounted the total population in each state and showed the count largely was accurate in all but 14 states. However, the bureau found the count varied by as much as 5 percent in states like Arkansas, Florida, Texas, New York and Minnesota.
In Florida, for example, the census missed an estimated 3 percent of its overall population. The state gained one congressional seat based on the census results, but fell short of getting a second additional seat by 171,000 people, or less than 1 percent of its population.
The census missed an estimated 1.9 percent of the population in Texas, which missed out on a third new congressional seat by less than 1 percent of its population.
According to population projections the Census Bureau released before the count, Florida had been expected to gain two congressional seats and Texas had been expected to gain three.
Tim Kennel, Census Bureau assistant division chief for statistical methods, said the report results were “in line with past censuses” and would help inform efforts to make the 2030 count more accurate. The agency does not plan to use Thursday’s report on state miscounts to correct the apportionment and other results already released.
An overcount means there were more people tallied than actually live in the state, while an undercount means there were fewer counted than actually live there.
The Census Bureau estimated the amount of an overcount in eight states: 5.4 percent in Delaware, 6.7 percent in Hawaii, 2.2 percent in Massachusetts, 3.8 percent in Minnesota, 3.4 percent in New York, 1.4 percent in Ohio, 5 percent in Rhode Island and 2.5 percent Utah.
At the same time the agency estimated the amount of an undercount in another six states: 5 percent in Arkansas, 3 percent in Florida, 1.9 percent in Illinois, 4.1 percent in Mississippi, 4.7 percent in Tennessee and 1.9 percent in Texas.
Those misses contributed to the most inaccurate census in decades, which is used to determine congressional districts and guide more than $1.5 trillion in federal spending annually. In a national report released earlier this year, the agency had historic undercounts of minority populations, including missing nearly one in 20 Hispanic residents. Agency officials have argued the count is still accurate enough to be used.
Close calls
In apportionment results released last year, Texas gained two seats, while Colorado, Florida, North Carolina, Montana and Oregon each gained one. California, Illinois, Michigan, New York, Ohio, Pennsylvania and West Virginia each lost one.
In Florida and Texas, the estimated undercounts released Thursday were far larger than the margin for a state gaining or losing a seat.
The misses may have also contributed to other shuffles between the states. Rhode Island was expected to lose its second seat, and had an estimated 5 percent overcount. The state kept that second seat by 19,000 people, less than 2 percent of its population.
New York lost its 27th seat to Minnesota by a historically small margin of 89 people, which is orders of magnitude smaller than the estimated overcount in both states — about 3.5 percent.
However the estimates are just that, and have significant variance. For instance Thursday’s report estimated the undercount in Florida may have varied between 4.9 percent and 1.9 percent. Additionally, apportionment is based on the relative population of all states, rather than population of each state individually.
Census officials and advocates noted the 2020 count faced unprecedented challenges including the coronavirus pandemic, natural disasters and decisions by the Trump administration that scrambled counting efforts.
Accuracy issues
Thursday’s report, which is based on a post-enumeration survey the agency conducted after the count, follows on from a national report showing that the agency undercounted 0.25 percent of the population. Most of those misses were in minority communities, renters and the homeless population.
The agency missed almost 5 percent of the country’s Hispanic population, the largest miss for that group in decades. In 2010, the agency missed about 1.5 percent of that population, less than 1 percent in 2000, and almost 5 percent in 1990.
The 2020 census also missed about 3 percent of the Black population, also the highest such undercount in decades. In 2010, it missed 2 percent of that population, less than 2 percent in 2000, and 4.5 percent in 1990.
On the other side of the ledger, the agency estimates it overcounted the Asian population by 2.6 percent, the white population by 1.6 percent and homeowners by 0.4 percent.
Census Bureau Director Robert Santos said the agency has started researching methods to adjust the data used by federal agencies in spending decisions. A 1999 Supreme Court decision prevents the use of statistical sampling to adjust census results used for apportionment and redistricting.
Kennel said the agency would not release state-by-state race and ethnicity undercounts as the structure of the survey used for the report did not have a large enough sample size.
For the post-enumeration survey, the agency sends questionnaires to millions of people asking them about demographic information, as well as whether they responded to the 2020 census. Agency officials then compare those responses to census results to help estimate how many people it missed in 2020.
Last decade, the post-enumeration survey showed the agency had missed more than 1 million members of minority communities in the 2010 census, including nearly 5 percent of Native Americans on reservations.
The agency has missed even larger parts of minority populations in earlier counts.
Kennel noted the 2010 census did not have any statistically significant misses at the state level.