Description:
A comparison of measures of work time in the CPS-ASEC data file (based on recall) with contemporaneous measures reveals many logical inconsistencies and probable errors. About 8 percent of ASEC respondents report weeks worked last year that contradict their work histories in the Basic monthly interviews; the error rate is over 50 percent among workers who move in and out of the workforce across their monthly interviews. Over 20 percent give contradictory information about whether they usually work a full-time weekly schedule (35 or more hours per week). A small part of the inconsistency arises because an increasing fraction of observations in the ASEC (over 20 percent by 2018) consists of people whose record was fully imputed. The errors and imputations are not random: The levels and trends differ by gender and race, and they affect the calculation of wage differentials between 1978 to 2018. After adjusting for the measurement errors and excluding imputations, we find that gender wage gaps among all workers narrowed by 4 log points more than is commonly reported, and that residual wage inequality decreased by 6 log points more. The biases also exist in measures of wage gaps and residual inequality among full-time year-round workers. Using a more carefully defined sample of such workers shows that gender and racial wage differentials have narrowed slightly less than previously estimated using ASEC data, but much more than indicated by commonly used estimates from CPS Outgoing Rotation Groups.