On the Estimation of Classical Human Capital with Two Independent Sources of Data on Actual Work Experience

Linda L. Moeller


The BLS multifactor productivity series decomposes labor productivity growth into components associated with increased capital intensity of production and shifts in the skill-composition of the work force attributable to changes in the level and distribution of human capital. The use of administrative record data on actual accumulated work experience as an indicator of workers' current productivity is one of the distinguishing features of this series. The current procedure relies on a one-time match of Social Security Administration (SSA) data to records from the March 1973 CPS to develop a proxy that is entered in a human capital wage equation. The parametric relationship between accumulated work experience and the demographic characteristics of the work force is unlikely to remain stable over time, however. Therefore the BLS has undertaken a long-run research project to update the work experience data at regular intervals.

Two independent sources of information on accumulated work experience were examined: SSA administrative record data on quarters of covered employment, and data from the 1984 Survey of Income and Program Participation (SIPP) on the number of years in which the respondent was employed for 6 months or longer. Direct comparison of the SIPP and SSA data show that the SSA data generate biased estimates of total actual work experience due to incomplete program coverage at the beginning of the Social Security program, in 1937, and subsequent changes in SSA coverage rates. Contemporaneous experience profiles estimated with SSA data are flatter than their SIPP counterparts, while the 1973 CPS-SSA experience profiles are flatter than their 1984 counterparts. Wage profiles estimated with the 1973 CPS-SSA experience proxy are too high for younger and older workers. The result is that weights assigned to the hours growth rates for younger and older workers are too large, while those assigned to hours growth rates for prime-aged workers are too small. These results tell something of a cautionary tale with respect the assumption that the coefficients of the experience equation are stable over time.