The U.S. Census Bureau did not release its standard 1-year estimates from the 2020 American Community Survey (ACS) because of the impacts of the COVID-19 pandemic on data collection. The Census Bureau released experimental estimates developed from 2020 ACS 1-year data.

The standard 2020 ACS 1-year estimates did not meet the Census Bureau’s Statistical Quality Standards designed to ensure the utility, objectivity and integrity of the statistical information. Unlike the ACS, the 2020 Census was able to postpone their Nonresponse Followup to a time when they could carry out the full operation, limiting the impact of the pandemic on data quality in ways the ACS could not.  

The COVID-19 pandemic posed numerous challenges to collecting ACS data in 2020, as described in our recent Adapting the American Community Survey Amid COVID-19 blog. As a result, the ACS collected only two-thirds of the responses it usually collects in a survey year and the people who did respond to the survey had significantly different social, economic and housing characteristics from those who did not. This is called “nonresponse bias.” 

Specifically, Census Bureau staff found high nonresponse from people with lower income, lower educational attainment, and who were less likely to own their home. Nonresponse bias is a natural part of sample surveys, and often statisticians can adjust for nonresponse bias by giving more weight to responses from underrepresented groups. However, Census Bureau staff found that standard nonresponse adjustments to the ACS 1-year estimates could not fully address the differences in a way that meets Census Bureau quality standards.

“The Census Bureau is committed to providing high quality data. The 1-year estimates for the 2020 ACS don’t meet our standards, so we can’t release them,” Census Bureau Acting Director Ron Jarmin said. “The ACS is one of the most comprehensive sources of information about the U.S. population, and it is relied upon by the government, business community and individual Americans. Because of that, it is essential that ACS data truly represent our communities, and the standard 1-year products from the 2020 ACS don’t do that.”

Data Collection in 2020

The ACS surveys a sample of 295,000 housing unit addresses each month on a variety of demographic, social, economic and housing characteristics. Each year, the Census Bureau typically combines the monthly responses into a set of 1-year estimates for the nation, states and communities with populations of 65,000 or more. The Census Bureau also produces 5-year estimates for geographies down to the “tract” or neighborhood level.

Response rates for March through September 2020 — half the months that would make up the 2020 ACS 1-year estimates — were severely impacted by the pandemic. To protect the safety of employees and the people who respond to the survey, the Census Bureau was forced to suspend many ACS data collection operations including mailing survey materials, following up in person with households that didn’t respond, and collecting data from nursing homes, college dorms, prisons and other group quarters. 

These challenges in collecting responses significantly impaired the quality of the resulting estimates, which were often inconsistent with benchmarks or administrative data or changed in unexpected magnitudes. The issues Census Bureau staff observed during data review were widespread across social, economic and housing characteristics, signaling a serious quality issue in the data. 

The Census Bureau released a report, An Assessment of the COVID-19 Pandemic’s Impact on the 2020 ACS 1-Year Data, detailing the issues.

Experimental 1-Year Estimates

The Census Bureau released a series of “experimental” estimates from the 1-year data on November 30, 2021. Because of the underlying quality concerns, the Census Bureau urges caution in using the experimental estimates as a replacement for standard 2020 ACS 1-year estimates. Users should evaluate the estimates and alternatives to determine if they are suited for their needs. To create these estimates, the Census Bureau applied an alternative set of weights to the 2020 ACS data to attempt to adjust for some of the nonresponse bias. Data from other surveys and administrative records informed the weights. 

The Census Bureau released the experimental estimates in the form of a limited number of data tables for limited geographies. At the same time a research paper, Addressing Nonresponse Bias in the American Community Survey During the Pandemic Using Administrative Data, detailing the methodology for the experimental weights was released. 

The Census Bureau released the 2020 ACS 1-year Public Use Microdata Sample (PUMS) files containing experimental weights on November 30, 2021. 

Technical Data Assistance

Recognizing the difficulty that the lack of standard 1-year ACS estimates has on data user communities, the Census Bureau provided a Resource Hub and data flow chart for more information. 

2016-2020 ACS 5-Year Estimates

The Census Bureau revised the methodology for the 2016-2020 ACS 5-year estimates in order to reduce nonresponse bias in data collected in 2020. After evaluating the effectiveness of the methodology, the Census Bureau determined the standard, full suite of 2016–2020 ACS 5-year data were fit for public release, government, and business uses. The Census Bureau released the standard, full-suite of 2016-2020 ACS data products on March 17, 2022.  Please visit the ACS New and Notable web page for more information.