The fragility of the Environmental Kuznets Curve: Revisiting the hypothesis with Chinese data via an “Extreme Bound Analysis”

This article revisits the validity of the Environmental Kuznets Curve hypothesis. Based on data for seven pollutants in 29 Chinese provinces from 1995 to 2010, we conducted statistical tests on the coefficients of the critical variables (polynomial income-related terms) by following the logic of a “General Sensitivity Test,” more often known under the name of “Extreme Bound Analysis (EBA),” which was initially proposed by Leamer (1978). We tested a set of models (6144) and estimation methods (23) using a bootstrap approach in which each model is estimated 1001 times (1000 bootstrapped database+one original database). Based on the 6144∗23∗1001 regression results, we construct distributions for the coefficients of the income-related terms and calculate the cumulative probabilities for the single coefficients with the expected signs and the regressions that obtain the “accepted” forms of the EKC (inverted-U or N forms). Our test reveals that the EKC hypothesis cannot be considered valid for any of the seven emission indicators because the probability of obtaining a negative coefficient for the quadratic income terms and the probability of finding an inverted-U-form relationship between income and pollution are all lower than the 95% critical level. Without reaching the 95% statistical significance level for emissions such as CO2 and industrial gas, our results seem to more often reveal a positive linear relationship with income.