| Whether world income levels converge or diverge, and what the long term sources of economic growth have been, remain the most important and controversial issues in economic growth. To these issues, traditional neoclassical growth theory and new endogenous growth theory have provided contradictory explanations. The conflict centers on the question of convergence.;This study reinvestigates the convergence hypothesis using improved LSDV-fixed-effect analyses including five endogenous growth variables, extended sets of data in pooled time-series, more careful designs of nonlinear/cyclical patterns inside each data set, and critical analyses of data measurements.;No evidence is found for conditional convergence when tested using five endogenous growth variables for conditioning, when tested in different subsample sizes, when tested in different subsample periods, and when tested using growth rates instead of levels. The seemingly-appropriate evidence for conditional convergence appears to result from model misspecification and two measurement artifacts, 'downward-sloping-bias' and 'nonstationarity' associated with the use of level forms for control variables. On the other hand, the evidence for 'divergence' is strong and robust in all respects.;This study therefore concludes that endogenous growth theory is the theory that is most consistent with the observed pattern of economic growth. World income levels diverge as predicted by endogenous growth theory. The long term source of economic growth is technological innovation embedded in five major endogenous growth factors, physical capital with R&D effects, human capital with knowledge spillover effects, trade liberalization with learning-by-doing effects, industrialization with big-push effects, and urbanization with agglomeration effects. |
