Misspecification and cross–country growth regressions

Abstract

Cross–country growth studies are commonly used to empirically investigate the neoclassical convergence hypothesis, the determinants of growth and the impact of policies on growth performance. While questions have been raised concerning the reliability of data and the specification of the explanatory variables, little attention has been paid to the derivation of growth rates. The standard method employs time series regressions to estimate average annual growth rates, which are then used as the dependent variable in the cross–section analysis. This procedure is only valid if the growth rate equations are well specified and the growth rates are stable over the time period in question. Using data for 75 countries for the period 1950 to 1992, we present much evidence to show that neither of the above propositions is supported; growth rate regressions are commonly misspecified and suffer from parameter instability. These results cast doubts on the validity of cross–country growth studies.