The 2020 American Community Survey and the Pandemic: Trouble is Brewing
11/22/21 / Kevin Raines
As demographers and market researchers, we rely heavily on US Census data to inform and underpin our work. In particular, we frequently use American Community Survey (ACS) data, which is the annual Census sample data collection process to examine the characteristics of the American population. Over 3 million addresses are surveyed each year.
As we all know, the year 2020 was marked by the chaos of economic shutdowns, eviction moratoriums, layoffs, and other Covid-19 ripple effects. So we were particularly anxious to see what the data told us once the 2020 ACS 1-year summary data was released. Unfortunately, that will not occur as the US Bureau of the Census has decided not to release the data due to concerns about data quality and integrity.
The problem is unfortunate, but not surprising. The ACS is conducted over the course of each year, with 12 overlapping three-month data collection cycles. Unfortunately, as the Covid-19 shutdown began, the US Bureau of the Census was forced to shut down and/or limit several critical elements of data collection between March of 2020 and April of 2021. Both mail and in-person outreach were suspended from March through June of 2020, and limited through April of 2021. Those limitations also impacted the ability to gather information via online and telephone data collection.
These issues led to a non-response bias in the results, particularly since they impacted the ability to conduct follow-up contacts with households that did not respond to initial outreach. These harder to reach populations disproportionately have lower incomes, lower education levels, and lower home ownership, so their absence impacted the accuracy of the results. Additionally, access limitations to group housing such as dormitories, senior facilities, and prisons limited the ability to accurately survey those groups.
Overall, response rates to the 2020 American Community Survey were the lowest response rates in the history of the survey at 71 percent, with response rates much lower than even that level during the peak pandemic months. Compare that to response rates of 86 percent in 2019 and 92 percent in 2018.
In a bit of good news, though, the decision currently affects only the 1-year population characteristics. The US Bureau of the Census is still evaluating whether combined 5-year estimates will be released as usual. Further, the Bureau is also considering providing limited sets of 1-year tables for certain geographic areas using experimental weighting methods, with a warning to be cautious in interpreting findings. So, we may yet be able to better understand the turbulent state of the US in 2020, albeit likely with a fog of uncertainty that appears to be the theme of such a tumultuous year.