A note on gravity models and international investment patterns

We show that recent methodological advances in econometric theory raise questions about the results obtained by some influential contributions on the determinants of international investment patterns, since the seminal paper by Lane and Milesi-Ferretti (2008) (LMF). In most such contributions, estimated equations are affected by heteroscedasticity, which may be shown to lead to inconsistent estimates in log-linearized models. Thus, the empirical findings of these works may need to be reassessed. By taking the results in LMF as a benchmark, we use a different methodology, which produces consistent estimates even under heteroscedasticity and report substantial differences with respect to the traditional methods. Moreover, we extend the data-set over time (over years from 2001 to 2009) to estimate a panel gravity model, which allows to properly account for unobserved heterogeneity through country-pair fixed effects and further improves on the cross-section analysis, by also reconciling empirical evidence with economic theory. Our panel estimates suggest the relevance of a diversification motive in driving international equity purchases.