Learning about Comparative Advantage in Entrepreneurship: Evidence from Thailand JOB MARKET PAPER

Entrepreneurial activity has been argued to be an important stimulus of growth, especially in less developed countries. However, measuring the returns to entrepreneurship is made difficult due to potential selection on the basis of unobservable abilities and the agents’ imperfect information about their own comparative advantage in entrepreneurship. I develop a novel extension to projection-based panel data methods to estimate a model of returns to entrepreneurship that accounts for heterogeneous abilities, learning, and financial constraints. My approach has two main advantages: 1) it is robust to arbitrary relationships between latent heterogeneity and sector and input choices; and 2) using this method, I can test between the full model and nested models featuring homogeneous returns and/or perfect information, allowing the data to determine the need for the additional complexity. I estimate the model using panel data from Thailand, and find strong evidence of selection into entrepreneurship on the basis of comparative advantage. The results show that the households with low ability in the default sector have high productivity gains from switching to entrepreneurship, and suggest that these households learn about their comparative advantage through lower-than-expected productivity in the default sector and/or positive productivity realizations in entrepreneurship. Importantly, the structural results do not suggest a salient role for savings or credit constraints in entrepreneurship decisions. I conduct additional analysis using agricultural output prices and their interactions with the household’s farm acreage as instruments for savings and self-reported credit constraints and find no significant effects of these on entry into entrepreneurship. ∗Yale University, Department of Economics; [email protected]; http://nyshadham.squarespace.com/ †I would like to especially thank Mike Boozer for his invaluable guidance and support. I would also like to thank Ach Adhvaryu, Joe Altonji, Prashant Bharadwaj, A.V. Chari, Penny Goldberg, Fabian Lange, Costas Meghir, Chris Udry, and participants of the Yale Labor and Development Prospectus Workshops for their helpful comments. All errors are my own.