Abstract

The main aim of this paper is to investigate the role of distance in gravity regressions in relation to the missing globalisation puzzle cited by Coe et al. (2002). These authors claimed that the fact that distance coefficients do not fall over time is another puzzle that should have been added in Obsfeld and Rogoff's (2000) paper, in which six major puzzles in international macroeconomics were cited. In order to do so, a linear and a non-linear gravity model are estimated for a cross-section of 65 countries over the period 1980-1999. Non-linear specifications of the gravity model have been recently used in the literature to address the issue of non-declining transport costs over time (Coe et al., 2002 and Croce et al., 2004). However, we claim that there is no missing globalisation puzzle once the linear specification results are correctly interpreted. The main findings are that distance has a different effect for developed and developing countries and that the grounds for using a non-linear specification are not clear since the linear specification, in general terms, shows better performance than the non-linear one.