Gravity across Space and Time

How well can the standard gravity equation account for the evolution of global trade flows over the long run? This paper provides the first systematic attempt to answer this question using a newly-assembled data set of bilateral trade flows, income levels and trade frictions that spans the years from 1870 to 2005. Using this panel data set we perform a structural estimation of the gravity equation and compare the gravity-predicted trade flows with their empirical counterparts. The estimation results highlight two major puzzles: (i) the standard gravity model can explain only a small share of the variation in trade flows over time, and (ii) it requires very large time-invariant trade costs to match the average value of trade flows between country pairs. The two puzzles appear to be closely related to the assumption of a constant trade elasticity throughout the entire sample period. Allowing for modest changes in the trade elasticity across sub-periods significantly improves the time-series fit of the gravity equation and reduces to more reasonable magnitudes the time-invariant trade costs required for
gravity-predicted trade flows to match the data. These findings suggest that the key to reconciling the gravity equation with the experience of globalisation history may lie in understanding the reasons for changes in the trade elasticity over time.