Parameter Heterogeneity In The Neoclassical Growth Model: A Quantile Regression Approach

In this study we examine the issue of parameter heterogeneity in the neoclassical growth model using a quantile regression estimator. Using cross-sectional data on 86 countries covering the period from 1960 to 2000, we estimate a version of the growth model of Mankiw, Romer and Weil (1992). We first estimate the model by OLS. We find that the model is quite successful in explaining the growth empirics of the ¡°average¡± country. We next estimate the model using quantile regression. The results of quantile regression are at odds with the OLS results. We find evidence of partial parameter heterogeneity. Countries whose growth rates are in the higher quantiles respond differently to investment in human and physical capital than do countries whose growth rates are in the lower quantiles. The neoclassical model predicts conditional convergence. The results from the quantile regression do not fully confirm this prediction. We find that convergence is not a generalized phenomenon across the conditional growth distribution, and, in particular, is not characteristic of countries in the lower quantiles. This suggests that an endogenous growth model, where government policies play a more decisive role in shaping the growth process, may be more suitable to describe growth in the lower tail of the distribution, whereas growth in the middle and higher quantiles is better described by the neoclassical model.