May 22, 2023

The United States produces its official estimates of poverty and inequality from an annual survey of households, relying on the cooperation of respondents to provide an accurate account of their incomes. For the typical household, earnings accounts for over 80 percent of total income, and thus accurate reporting of earnings is tantamount to accurate reporting of incomes.

However, households are increasingly reluctant to provide earnings information to federal government surveyors, or to participate at all, and today nearly half of earnings are missing from the main survey used to produce these important statistics on American’s financial well-being.

In a recent paper published in the Journal of Political Economy, Chris Bollinger and Jim Ziliak, in collaboration with Barry Hirsch of Georgia State University and Charles Hokayem of the U.S. Census Bureau, used a unique, restricted-access dataset linking the Current Population Survey Annual Social and Economic Supplement to administrative tax records to study the patterns of earnings nonresponse across the distribution as well as its potential consequences for important labor-market issues such as earnings gaps by gender and race, and inequality.

The administrative data permit the researchers to get a second, independent report on earnings among those who respond to the survey, and a first report on those who fail to respond to the survey. They found that nonresponse across the earnings distribution is U-shaped, highest in the left and right tails, and is in evidence across gender, race, ethnicity, and employment status (hourly and full-time full-year). Moreover, this pattern of nonresponse matters for our understanding of gender and race earnings gaps and inequality, with earnings gaps and earning inequality estimates biased downward by 10-20 percent, meaning that male-female and Black-white earnings differences, as well as inequality, are worse than that reported in official statistics.