Segmentation in Urban Labor Markets: a Machine Learning Application and a Contracting Perspective

Michael Kaiser (LMU Munich - Department of Economics)

Abstract: A significant fraction of the labor force in developing countries is “informally” employed outside the “formal” sector. Much of academic thinking uses this dichotomous view to study (urban) labor markets in developing countries, although in the presence of imperfectly enforced institutions employers and employees could choose any contract that satisfies their preferences and meets their constraints. I employ labor force survey data from Tanzania and unsupervised Bayesian machine learning to estimate the latent structure of observed contracts in the urban private sector of Dar-es-Salaam. The results suggest that around 30% of the relevant population cannot be adequately captured using a dichotomous view of the market. Controlling for employees’ observable characteristics, I estimate wage distributions that suggest that workers are willing to trade off formal protections against higher pecuniary remuneration. Taken together the results suggest that a non-negligible fraction voluntarily chooses non-formal employment. An economic contracting framework is presented that formalizes the argument’s underlying principles.

Download the paper