The Gravity Model for International Trade: Specification and Estimation Issues

The Poisson gravity model along with pseudo maximum likelihood (ML) methods has become a popular way to model international trade flows. This approach has several econometric advantages that we outline in the paper. We argue that estimating the parameters by ML would only be justified statistically if the trade flows were independent. Such an assumption, however, is generally not valid, and a failure to account for spatial dependence may lead to biased parameter estimates and misleading inferences. To overcome this estimation problem we suggest eigenvector spatial filtering variants of the Poisson gravity model (without and with zero-inflation) along with pseudo ML estimation.