Human Capital, Saving Rate, Population, and Growth - Consistent and Robust Estimation of Speed of Convergence

In general, Baumol (1986)’s growth regression is a starting point in growth empiric. But, as De long(1988) pointed out, it has serious defects. In this study, we apply Arellano and Bond (1991) to this regression using the data of PWT 9. We extend the econometric model also using instrumental variable estimation for solving endogeneity due to measurement error problem and correlation between error and right-hand variables. Differenced panel IV (using instruments 2) yields the estimate of speed of convergence 13.2%, and differenced panel GMM (using instruments 2) 19.2%. These coincide with those of previous researchers such as Cellini (1997) which reveals the higher estimated speed when controlling the endogeneity of explanatory variables in growth regression. In addition, we adopted the semiparametric and nonparametric approach for getting robust estimator for the speed: quantile regression, partially linear regression and kernel estimation for regression function. We compared the results with those by traditional methods.