By Volker Tjaden and Felix Wellschmied
Standard search models are inconsistent with the amount of frictional wage dispersion found in U.S. data. We resolve this apparent puzzle by modeling skill development (learning by doing on the job, skill loss during unemployment) and duration dependence in unemployment benefits in a random on the job search model featuring two-sided heterogeneity. The model’s key parameters are calibrated using micro data on employment mobility and wages from the Survey of Income and Program Participation (SIPP). Our model is consistent with the amount of frictional wage dispersion found in the data. Skill development on the job is the most important driver behind this result. Meanwhile, firm heterogeneity never accounts for more than 20% of overall wage inequality within an age cohort.
Generating wage dispersion that compares to data is not obvious. This paper takes a labor search model and throws at it many factors that can generate dispersion (except education, which is controlled in the data I presume), calibrates it carefully and then looks what model feature sticks. It turns out unemployment spells or firm heterogeneity are relatively unimportant in this regard.