This dissertation contains three essays relating to learning in the workplace. In the first essay, I develop a framework to distinguish empirically between two learning processes that take place in the workplace: learning-by-doing and learning about ...
This dissertation contains three essays relating to learning in the workplace. In the first essay, I develop a framework to distinguish empirically between two learning processes that take place in the workplace: learning-by-doing and learning about match quality. Learning-by-doing is the process of a worker learning how to do a particular job, while learning about match quality is the process of finding out whether the worker is well suited for the particular job. I estimate the model using a worker-firm matched data-set from France. I find that learning about match quality is the more important process, as it takes a longer time and leads to the accumulation of more match-specific capital.
The second essay explores one policy implication of this finding. It looks at the policy of removing separation costs for low-tenure workers in an environment where there are large separation costs, possibly through the use of fixed-term contracts. I show that, when learning-by-doing is dominant in an employment relationship, this policy reduces average worker productivity, since it encourages the employment of low-tenure workers who are less productive. When learning about match quality is more important, however, this policy allows for more experimentation, leading to better-quality matches, hence increasing average worker productivity. Using the estimates from the first essay, I quantify this productivity effect and show that the positive effect of learning about match quality dominates.
The third essay shows that differences in the amount of learning about match quality can be an important source of unemployment rate differences across groups. Groups that have more heterogeneity in match quality and hence undertake more learning about this quality accumulate more match-specific information. This shields them from the adverse effects of negative shocks, thus leading to a lower unemployment rate. The essay develops a model that describes this effect and shows that for reasonable parameter values the effect can be quantitatively significant. I propose this as one possible explanation for the unemployment rate differences observed across skill groups.