Trade and Income -- Exploiting Time Series in Geography

Abstract

Establishing a robust causal relationship between trade and income has been difficult. Frankel and Romer (1999) use geographic instruments to identify the causal effects of trade. Rodriguez and Rodrik (2000) show that these results are not robust to controlling for missing variables such as distance to the equator or institutions. This paper solves the missing variable problem by generating time varying geographic instruments. The quantity of world trade carried by air has been increasing over time. Estimates from a gravity model show an increase in the elasticity of bilateral trade with regard to air distance over time while the elasticity with regard to sea distance has declined. This change has heterogeneous effects on the trade between pairs of countries depending on the relative sea and air distances between them. This heterogeneity in geography can be used to generate geography based predictions for bilateral trade that vary over time. These predictions can be aggregated and used as instruments for trade in a regression of income on trade. The time series variation allows for controls for country fixed effects, eliminating the bias from any omitted time invariant variables such as distance from the equator or historically determined institutions. Trade has a significant effect on income with an elasticity of roughly one half. Differences in predicted trade growth can explain roughly 17 percent of the variation in cross country income growth between 1960 and 1995.

Keywords: trade, income, output, geography, growth

JEL Classification: O3, O4, O11, F43, F4, R1