Competing specifications of the gravity equation: a three-way model, bilateral interaction effects, or a dynamic gravity model with time-varying country effects?

We investigate bilateral trade flows across the EU-15 countries from 1962 to 2003 by different specifications of the gravity model. We augment the basic gravity model with population and exchange rate variables, and then include time-varying country fixed effects, to account for Anderson and van Wincoop (Am Econ Rev 93(1):170-192, 2003) multilateral resistance terms. Then, following the previous theoretical derivations of the gravity model in the presence of panel data in a dynamic setting we change the specification of our gravity model. We compare the results of different specifications showing the improvement in each case. We claim the comparative superiority of the dynamic gravity model with time-varying exporter and importer fixed effects due to its higher explanatory power. Finally, we compare out-of-sample forecasting performance of different specifications of the gravity model.