This paper casts doubt on empirical results based on panel estimations of an “inverted-U” relationship between per capita GDP and pollution. Using a new data set for OECD countries on carbon dioxide emissions for the period 1960-1997, we find that the crucial assumption of homogeneity across countries is problematic. Decisively rejected are model specifications that feature even weaker homogeneity assumptions than are commonly used. Furthermore, our results challenge the existence of an overall Environmental Kuznets Curve for carbon dioxide emissions.