Convergence or divergence in cross-country growth?

In the traditional empirical convergence literature, a negative coefficient on initial income in a cross-country growth regression is interpreted as evidence of poor countries growing faster than richer ones. A key assumption in this work is that the relationship between initial income and income growth is linear. The linearity assumption is challenged in some new growth theories, and studies adopting an alternative (semi-parametric or nonlinear) econometric methodology provide support for a nonlinear specification. This paper finds evidence for nonlinear convergence. Using semi-parametric estimation we find that convergence occurs among countries with very low and very high initial incomes, suggesting that convergence clubs characterize the cross-country growth process. Our results provide further evidence for multiple-regime steady states.