Bilateral Trade Flows: One Prediction of Several Theories Estimated in Nine Thousand and Two Samples

Abstract

Re-sampling techniques and theoretically-motivated extraneous information are used to revisit Hummels and Levinsohn's (1995) analysis of the testable implication of the monopolistic competition trade theory for bilateral trade volumes. Confronting this implication with data from over nine-thousand samples, we find a robust relationship between the estimated parameter and the summary statistics of the samples. Using this information, which is extraneous to the estimating equation, we rank the samples according to the likely share of monopolistic competition-inspired trade in total trade volumes. We then conduct a consistency check: to examine whether this testable implication is less refuted in samples where the production structures are more conducive to generating large shares of monopolistic competition-inspired trade. The monopolistic competition trade theory passes this consistency check. Next, we recognize that other trade theories can generate the same estimating equation. We argue that the assumptions, unrelated to these models' production structure, are unlikely to account for the differences in the estimated parameter across the samples. Our estimation results imply that the testable implication performed worse in samples where factor-endowment based trade is more likely to predominate. These findings indicate that the empirical performance of the trade volume predictions of the Helpman- Krugman monopolistic competition trade theory is much better than suggested by previous research.

JEL Classification: L11, C81, F49, R38