Industrial Targeting, Experimentation and Long-Run Specialization

Abstract

The government?s policy of selecting industries for targeted promotion is formalized as a multi-armed bandit process. The welfare maximizing targeting strategy is determined by a profitability index attached to each industry. A government updates its beliefs about its country?s comparative advantage in a Bayesian way after observing realizations of random productivity of new firms in a targeted industry. The sequence of posterior beliefs is a martingale with respect to a natural filtration generated by a sequence of random productivity parameters. By applying martingale convergence results, the paper analyzes the pattern of specialization to which a country will converge in the long-run given its government?s optimal targeting policy.

JEL Classification: D83, 014, F13, C44