Introducing tariff data in the gravity equation

Abstract

This paper studies whether the inclusion of tariffs in the gravity estimations makes a significant change on the other commonly-used trade costs. Including tariffs in the gravity equation is not a common practice, due to the difficulty of obtaining quality aggregated data. Being aware of this pitfall, I have collected the best free available data on applied tariffs and constructed a panel of eighteen countries, along twenty-one years on which I estimated two forms of the gravity-equation, the difference between them being the scaling of the dependent variable. I used two econometric methods to estimate the equations - the least-square dummy variable approach and the Poisson pseudo-maximum likelihood estimator. The results I obtained depend on whether the dependent variable is size-adjusted or not; when just considering exports, including tariffs reduces the bias on the free-trade-agreement coefficient, while leaving the other trade costs coefficients in usual ranges. This finding is not robust to the specification of size-adjusted exports; in this case, omitting tariffs from the gravity equation does not lead to significant changes in the other coefficients.