GRAViTY
(with Rodolphe Desbordes, Strathclyde University)
Abstract: This paper makes two contributions to the literature on structural gravity modelling in long panel datasets: first, by adopting a common factor framework we show that multilateral resistance (MLR) can be addressed in empirical models which do not pool all the available observations. We offer alternative specifications for exporters or exporter-importer dyads while still accounting for 'global' MLR, further enabling us to highlight differences in the impact of policy variables across economies. Second, we demonstrate that a straightforward extension to our modelling framework is able to account for a large share of the significant bias inherent in standard log-linear gravity regressions arising from the omission of zero trade flows. Experimenting with the observable information on economic mass and accessibility for zero flow dyads enables us to address this selection bias maintaining the simplicity of a linear model estimated with ordinary least squares. We illustrate the performance of these two aspects of panel gravity modelling in Monte Carlo simulations and empirical applications.
Sir Clive Granger BuildingUniversity of NottinghamUniversity Park Nottingham, NG7 2RD
Enquiries: hilary.hughes@nottingham.ac.uk