Learning from experience or learning from others?: Inferring informal training from a human capital earnings function with matched employer-employee data

A model of informal training which combines learning from own experience and learning from others is proposed in this paper. It yields a closed-form solution that revises Mincer-Jovanovic's [Mincer, J., Jovanovic, B., 1981. Labor mobility and wages. In: Rosen, S. (Ed.), Studies in Labor Markets. Chicago University Press, Chicago, pp. 21-64] treatment of tenure in the human capital earnings function. We estimate the structural parameters of this non-linear model on a large French cross-section with matched employer-employee data. We find that workers on average can learn from others 10% of their own human capital on entering one plant, and catch half of their learning from others' potential in just 2 years. The private marginal returns to education are declining with education as more educated workers have less to learn from others and share the social returns of their own education with their less qualified co-workers. The potential for learning from others on the job varies across jobs and establishments, and this provides a new distinction between imitation jobs and experience jobs. Workers in imitation jobs, who learn most from others, tend to have considerably longer tenure than workers in experience jobs. Although workers in experience jobs can learn little from others, we find that they learn a lot by themselves. We document several analogies between the imitation jobs/experience jobs "dualism" and the primary/secondary jobs and firms' dualism implied by the dual labor market theory. However, our binary classification of jobs depicts the data more closely than the dual theory categorization into primary-type and secondary-type establishments. Competition prevails between jobs and firms but jobs differ by their learning technology.