Returns to type or tenure?

A regression of wages on firm tenure is likely to yield biased estimates of the returns to tenure because tenure and wages are confounded by unobserved attributes of the job and the unobserved quality of the match between the firm and the employee. Previously, the within-job variation in tenure has been used as an instrument to estimate the average returns to tenure. We propose to use instead an easy-to-implement control function estimator for the returns to tenure and their dependence on unobserved heterogeneity. The results obtained for Germany indicate that there is a substantial amount of unobserved heterogeneity in the returns to tenure and that good job matches are characterized by higher returns to tenure in the first 5 years and lower returns thereafter.