This paper proposes a systematic comparison of parametric (i.e. OLS regressions and related generalizations) and non-parametric (i.e. kernel regressions and regression trees) log-linear gravity models for reproducing international trade. For this aim, experiments were carried out to estimate a log-linear gravity model reproducing import and export trade flows in quantity between Italy and 13 world economic zones, based on a panel estimation dataset. The best parametric regression model was first estimated, so as to define a baseline reference model. Some specifications of non-parametric models, belonging to the categories of kernel regressions and regression trees, were also estimated. In this paper we first contrast the performance of parametric and non-parametric models by means of the comparison of goodness of fit measures (R2, MAPE) both in estimation and in hold out sample validation. Furthermore, in order to assess the differences in model elasticity and forecasts, both parametric and non-parametric models are applied to some future scenarios and the corresponding results compared.
File in questo prodotto:
Non ci sono file associati a questo prodotto.