Testing the augmented Solow growth model : An empirical reassessment using panel data

We estimate a conditional convergence equation derived from an augmented Solow model where human capital is defined as skilled labour. We implement Generalized Method of Moments (GMM) estimators on a panel of countries. Estimation is carried out on the model in levels instrumented by the lagged first differences of the explanatory variables. We argue that using GMM estimators on the first differenced equation, as was previously attempted, is inappropriate as lagged levels of regressors provide weak instruments for current first differences. Using Asymptotic Least Squares, we check that the GMM estimates for structural parameters are consistent with the restrictions imposed by the model under the classical assumption of constant speed of convergence. We find that countries cover half the distance to their steady state in 16 years. Nonlinear estimations carried out under the assumption of endogenous speed of convergence do not validate the Solow model, whose theoretical predictions as to the dependence of the speed of convergence on structural parameters appear to be at best fragile.