This paper examines how robust economic, political, and demographic variables are related to water and air pollution. Employing Bayesian Averaging of Classical Estimates (BACE) for a cross section of 47 countries, 34 variables and 3 proxies for air and water pollution over a period from 1980 to 2000 we confirm the environmental Kuznets curve hypothesis and highlight the relevance of variables that are not directly related to production.