The carbon Kuznets curve: A cloudy picture emitted by bad econometrics?

We discuss several major econometric problems that have been ignored in the empirical environmental Kuznets curve (EKC) literature thus far. These are: First, the use of nonlinear transformations of integrated regressors and second, in a panel context, cross-sectional dependence in the data. Both problems fundamentally invalidate the use of widely applied time series and panel unit root and cointegration techniques. We use the important special case of the relationship between GDP and CO2 (and SO2) emissions to show and discuss in detail that the seemingly strong evidence for an inverted U-shaped relationship between these variables obtained with commonly used methods is entirely spurious and vanishes when resorting to estimation strategies that take the discussed problems into account.