The Dynamics of Comparative Advantage

Gordon H. Hanson, Nelson Lind, Marc-Andreas Muendler

This draft: Mar 18, 2018
First draft: Nov 30, 2013

University of California, San Diego


This paper characterizes the dynamics of comparative advantage and draws implications from these dynamics for quantitative analysis. In cross-section data, we establish that the distribution of export capabilities across industries is approximately log normal. This heavy-tailed shape is similar across 90 countries and stable over 40 years. Over time, there is mean reversion in export capability and this mean reversion, rather than indicating degeneracy, is instead consistent with a stationary stochastic process. We develop a GMM estimator for a Markov process whose stationary distribution nests many commonly studied distributions, and show that the Ornstein-Uhlenbeck (OU) special case closely approximates the dynamics of comparative advantage. The OU process implies a log normal stationary distribution and has a discrete-time representation that can be estimated with simple linear regression. Incorporating stochastic comparative advantage into the counterfactual analysis of changes in trade costs, we document the transitory nature of policy effects: churning causes targeted trade-policy changes to decay markedly, with most impacts fully dissipated within 10 to 20 years. These findings speak to the importance of incorporating dynamic comparative advantage into quantitative trade analysis.

keywords: International trade; comparative advantage; generalized logistic diffusion; estimation of diffusion process

jel: F14, F17, C22


  • supporting files
  • data sources
  • nber working paper [21753] version (html)
  • cesifo working paper [5622] version (pdf)
  • cage working paper [252] version (pdf)