The task force also encountered no irregularities indicating the 2020 figures were unfit for use in the apportionment of congressional seats or were of lower quality than those in 2010. However, the information given by the Census Bureau for review by the task force was too limited for a thorough assessment on the quality of the data. Due to tight deadlines facing the Census Bureau, the task force only was provided with state-level population counts lacking demographic information on race and Hispanic origin, the report said.
“That is not a thorough assessment,” said Nancy Potok, a former chief statistician of the U.S. who chaired the task force. “We didn’t find any anomalies that were immediate causes of concern that you couldn’t use the data.”
The task force was created last year to examine the quality of the initial data, because of concerns raised by the obstacles the census faced, until a more in-depth study could be conducted. The association's president, Robert Santos, was a task force co-chair until he was nominated by President Joe Biden to be the next Census Bureau director.
As part of the task force's review, the Census Bureau allowed three outside statisticians to look for potential opportunities for errors in the census numbers that were greater in 2020 than they were in 2010. The statisticians assigned a ranking for those potential error risks by each state using 10 measurements.
The statisticians found that the states with the highest potential risks for more errors in 2020 than 2010 were Alaska, Montana, New Jersey, New Mexico, New York, Texas and Utah. The states with the lowest risks were Delaware, Idaho, Indiana, Kentucky, Nebraska, South Carolina, South Dakota and Tennessee.
“These states have very different populations and range from mostly urban to mostly rural ... which indicates that the error risks apply to very diverse populations and not only to either densely or sparsely populated areas of the country," the statisticians wrote.
However, the task force ultimately decided that reaching any conclusions about data quality from the 10 measurements would be premature at the state level.
The statisticians also found out that data was collected using the most accurate methods — either from households filling out the questionnaire on their own or having a household member answer questions from a census taker — in 90% of households in 2020, a decrease from 93% in 2010.
This was the first head count in which the Census Bureau used administrative records from the IRS, Social Security Administration and others to fill in missing details about households lacking information. After reviewing studies on their use before the 2020 count, a task force member said he was unable to conclude whether using administrative records is more accurate than having census takers gather information about homes’ residents.
Two other panels of outside experts also are reviewing the quality of census data, and early next year the Census Bureau plans to release its own study of how good a job it did.
One of the panels of outside experts, overseen by The National Academy of Sciences Committee on National Statistics, is expected to provide a much more detailed assessment.
That group should also look at why so many people didn't answer questions on the form in 2020, whether there were increases in undercounts of Blacks, Hispanics and children in 2020 and how administrative records were used to fill in gaps about households that didn't answer their census forms, the task force said.
Follow Mike Schneider on Twitter at https://twitter.com/MikeSchneiderAP