Abstract:
In a recent meta study, Disdier and Head (2008) summarize 103 empirical studies that find that the (absolute value of the) distance coefficient in a gravity estimation of bilateral exports does not fall over time. Including zero-trade flows is suggested as a solution to this "distance-puzzle" (Felbermayr and Kohler, 2006). However, as shown by Helpman, Melitz, and Rubinstein (2008), HMR henceforth, for cross-section data, including zero-trade flows will rather increase - and not decrease - estimated distance coefficients, due to omitted firmlevel heterogeneity. Using industry level data over 28 industries for a time span ranging from 1978 to 2003, we first confirm earlier results that the distance coefficient increases over time using OLS. We then show that the bias from not controlling for firm heterogeneity in OLS rises over time, whereas the non-linear estimation of HMR leads to a declining distance coefficient over time. Our estimates are also shown to capture reductions in trade costs related to distance, since the estimated distance coefficients are strongly correlated with the time pattern in freight costs reported by Hummels (2007).