The Determinants of CO2 Emissions Differentials with Cross-Country Interaction Effects: A Dynamic Spatial Panel Data Bayesian Model Averaging Approach

This study analyzes the importance of a large number of possible determinants of CO2 emissions per capita during the period 1991-2014 for a sample of 123 countries. They key contributions are methodological given that we consider the effect of a great number of economic, institutional, demographic and socio-cultural factors that could affect CO2 emissions employing Spatial Bayesian Model Averaging techniques while accounting for different concepts of cross-country interactions and different spillover processes. Over the different type of interactions considered: geographical, genetic, linguistic and religious we find that traditional geographical interactions outperform the others. Spatial Bayesian Model Averaging analysis enable us to compute the PIPs for the different indicators to generate a probabilistic ranking of relevance for the various CO2 determinants. Our findings suggest that CO2 emissions are mainly determined by economic factors such as the sectoral composition, the prices of gasoline, the intensity of fossil fuels consumption and the level of output. In a intermediate level of importance we find social and demographic factors such as the age composition, the religious attitudes or the social globalization of the population.