The Output Convergence Debate Revisited: Lessons from Recent Developments in the Analysis of Panel Data Models

This paper examines the empirical basis of the output convergence debate in light of recent developments in dynamic heterogeneous panels with trended latent factors. Popular tools such as Barro’s cross-country regressions and two-way fixed effects (TWFE) estimators — which assume parallel trends and homogeneous dynamics — substantially underestimate the speed of convergence and yield mislead-ing inference. Dynamic common correlated effects (DCCE) estimators (Chudik and Pesaran, 2015) instead provide consistent estimates and valid inference robust to nonparallel trends and correlated heterogeneity, even with breaks, trends, or unit roots in the latent technology factor. The paper also proposes a way to estimate the effects of slowly moving determinants of growth. Using Penn World Tables data, the empirical results show that cross-country convergence in per capita output is very slow, while within-country convergence is reasonably fast. Capital accumulation is an important determinant of cross-country output differences, and slow-moving indicators — potential for conflict and protection of property rights — are statis-tically significant determinants of steady-state output per capita levels. The paper replicates a positive effect of democratization on output, but finds its statistical significance declines once nonparallel trends and dynamic heterogeneity are allowed for.