Masking Real Unemployment: The Overall and Racial Impact of Survey Non-Response on Measured Labor Market Outcomes

Abstract

A large and growing percentage of households are missed in the monthly Current Population Survey (CPS). For the survey as a whole, the rate of nonresponse is roughly 13 percent. This is higher for Blacks, with the share for young Black men being about 30 percent. The BLS’s current methodology effectively assumes that, with adjustment for various characteristics, people who are not included in a follow-up survey may not differ systematically from those who are included. The present paper, however, provides evidence that this may not be the case. With the rotation panel structure of the CPS data from 2003 to 2019, we investigate bias from nonresponse in CPS and its association with one’s prior labor market status, paying particular attention to how the relationship differs by race, ethnicity, and gender. Our analysis suggests that people are considerably more likely to be missing in a subsequent observation if they are unemployed or not in the labor force in the prior observation. We also estimate what the real labor market outcomes might have been when adjusting for nonresponse and undercoverage. Findings indicate that the current methodology may underestimate the national unemployment and labor force participation rates by about 0.7 and 1.2 percentage points, respectively. The gap between observed and adjusted unemployment rates tends to grow beginning in 2015. The unemployment rate is more understated for Blacks than for whites, particularly with a gap of about 3.3 percentage points for young Black men (age 16 to 34).