The Importance of Dynamics in Panel Gravity Models of Trade

Existing gravity models of trade based on panel data are often static, that is, they only allow for contemporaneous effects of regressors on trade. However, there are numerous economic arguments suggesting that trade is a dynamic process. Hence, we extend the static model by including lags of the regressors and lags of trade. Using a panel of 221 annual bilateral OECD trade flows over 48 years, we find that dynamics are strongly significant, so that static models are misspecified. The resulting dynamic panel gravity model leads to sensible short-term and long-term trade dynamics. We also show that the simple least squares dummy variable estimator, which is typically used in static panels, yields accurate estimates for our dynamic model and outperforms the popular generalized method of moments estimator of Arellano and Bond (1991)