����AV

Today, we count on data and data-based systems in our daily lives. The United States census is the backbone of most of these data sets that we rely on, so it is important to ensure its accuracy. 

The Census Quality Indicators Taskforce selected assistant professor of statistics to independently review the 2020 census as part of a three-person research team. The team released their findings in September.

In 2020, the pandemic, accusations of political interference, and many other factors created doubt that the census wouldn’t be accurate, says Auerbach.

Traditionally, the ASA and the statistics community spend the ten years between censuses looking over the data and offering improvements for the next census. However, because of the efficacy questions of the 2020 census, Auerbach’s research team investigated it in half a year.

“The ASA has historically been involved in advocating for the best census since 1840, right before the Civil War, which was a very formative census, and it became highly politicized. And partially because of that, the ASA as we know it today was born,” says Auerbach.

Auerbach was one of three researchers on the independent review team. He and his fellow researchers, including Paul Biemer and Joseph Salvo, had access to all census data to see how it was collected and what errors could have been introduced.

“Our major finding is what we didn’t find. We didn’t find evidence of malfeasance or politicking or any direct evidence of wrongdoing. However, we did find that largely because of the pandemic, the ways people were counted were of higher risk than previous censuses,” says Auerbach.

For example, in the 2010 census, if someone hadn’t reported their household information, the U.S. Census Bureau would send a worker to follow up. In 2020, because of COVID-19, that was not always a safe option. Instead, the Census Bureau relied more heavily on administrative records, like tax documents. “These practices could lead to more errors than the methods used in the 2010 census,” he says.

Reporting college students was another area they noticed might lead to some potential errors. “Typically, colleges will report the number of students in dorms to the census bureau, but since many students left their dorms halfway through the semester, it was unclear whether students were counted properly.”

Auerbach says that he and the research team found no evidence suggesting the 2020 Census is unreliable despite riskier counting methods. Another group at the National Academies of Sciences, Engineering, and Medicine, which includes other ASA members, will continue to help the Census Bureau improve the census and process the data, but Auerbach’s work is finished for now.

“Our census is a historical document. We really don’t have too many of them in our lifetime, so it was very rewarding to be a part of it,” he says.