Convergence in per capita CO2 emissions: A robust distributional approach

This paper investigates the convergence hypothesis for per capita CO2 emissions with a panel of 166 world areas covering the period 1960-2002. The analysis is based on the evolution of the spatial distributions over time. Robust measures of dispersion, asymmetry, peakedness and two nonparametric distributional tests - shape equality and multimodality - are used to assess spatial time differences. A robust normal reference bandwidth is also applied to estimate Markov's transition laws and its subsequent ergodic (long-run) distributions. Our results point toward non-stationary, flattening and right-skewed spatial distributions before the oil price shocks of the 1970s and more stable and symmetric shapes between 1980 and 2000 at the world level and for many country groupings (similar income, geographic neighbors, institutional partners). In the latter period, group-specific convergence patterns emerge with the clearest single-peaked and compact density shapes being reached in the wealthy, well-integrated and European countries during the last years of the panel. No significant multimodality is formally detected in the world distribution over the whole period. The Markov analysis suggests more divergence and larger per capita emissions for the world in the long run, with a doubling in median emissions and stable pollution gaps during the first 50 years of the transition. A variety of steady state distributions are identified in the country subsets.