Skip Header
U.S. flag

An official website of the United States government

Addressing Nonresponse Bias in the American Community Survey During the Pandemic Using Administrative Data

Jonathan Rothbaum, Jonathan Eggleston, Adam Bee, Mark Klee, and Brian Mendez-Smith

Due to the challenges of fielding a household survey during the COVID-19 pandemic, household nonresponse increased substantially in the American Community Survey, with evidence of increased nonresponse bias in many statistics. Specifically, higher socioeconomic status households became relatively more likely to respond during the pandemic. This likely biased estimates of many statistics, including building structure, marital status, educational attainment, Medicaid coverage, citizenship, income, and poverty. We use extensive administrative, third-party, and decennial census data to identify household and housing unit characteristics for respondent and nonrespondent households. We show that the pattern of survey nonresponse was unique during the pandemic period. For example, nonresponse was more strongly associated with income than in 2019. Second, we create new weights to adjust for nonresponse bias using entropy balancing, a form of empirical calibration. We evaluate the impact of our nonresponse adjustment in both 2019 and 2020 compared to the normal survey weighting. We estimate large impacts of nonresponse bias, particularly in 2020. For example, with the standard weights, real median household income increased 5.5 percent between 2019 and 2020, compared to 0.2 percent using the entropy balance weights. Overall, entropy-balance reweighting significantly reduced 2019-2020 changes in many estimates.


Back to Header