Regional Convergence in Europe: A Dynamic Heterogeneous Panel Approach

This paper studies the effects of allowing for heterogeneous slope coefficients in the Mankiw, Romer and Weil (1992) model, based on panel data for 193 EU-15 regions from 1980 to 2005. We first estimate the model using conventional pooled panel data estimators, based on data at five-year intervals, allowing at most intercepts to differ across regions. Then we relax the restriction of homogeneous slope coefficients by estimating separate time-series models for each region based on annual data, using Pesaran and Smith's (1995) mean group estimator. To account for spatial dependence, we employ the common correlated effects approach of Pesaran (2006). Our empirical analysis indicates important differences across regions in the speed of adjustment to region-specific long-run paths for the level of income per capita. Allowing for heterogeneous coefficients doubles the speed of adjustment to 22% per year on average compared to the homogenous case, which suggests downward bias in the latter. We also find a positive and significant effect of the rate of investment, although the implied capital elasticity and the estimated long-run effect of investment are smaller than expected.