The gap between the rich and the poor: Patterns of heterogeneity in the cross-country data

Persistent income inequalities across nations have led to the emergence of a voluminous literature, focusing on cross-country growth regressions. Using the same regression model for all countries in the sample, the majority of these studies ignore the inherent heterogeneity that can actually lead to different regression models for different countries. This paper explores whether this assessment is valid, and in doing so, provides a way to overcome it. Bayesian classification analysis is used to reveal patterns of heterogeneity and to identify groups of countries with similar growth processes. Standard growth regressions can then justifiably be performed on each subsample. The method is illustrated using a cross-country data set that includes the Solow growth model variables.