Modeling Parameter Heterogeneity in Cross Country Growth Regression Models

Given the failure of the conventional linear Solow growth model to establish reliable results in the analysis cross-country growth performance, this paper proposes a new framework using the concept of hierarchy of time-scales. By hierarchy of time scales, I mean that slower moving variables such as culture, play a major role in determining medium moving variables such as institutions, and which in turn play a major role in determining faster moving variables such as the conventional determinants of economics growth. This approach provides a systematic way of thinking about the heterogeneity in the cross-country growth performance. In the context of the Solow growth model the hierarchical approach suggests a local generalization of the Solow growth model in the form of a semiparametric varying parameter model along the lines of Hastie and Tibshirani (1992). Using the varying coefficient model, this paper studies two examples. In the first example the parameters of the model vary according to initial human capital while in the second they vary according to a measure of ethnic diversity. The results suggest that there exists substantial parameter heterogeneity in the cross-country growth process.