Robust inference concerning recent trends in US environmental quality

Understanding trends in environmental quality is important for individuals and policymakers. Typically, trends in environmental quality are determined solely through comparisons of unconditional or conditional mean pollution levels. However, reaching unambiguous conclusions on the basis of only the first moment of the (unconditional or conditional) distribution is problematic since it ignores what is occurring in different regions of the distribution. Even relying on indices that incorporate both mean and variance is suspect to the extent that relative rankings are not typically robust to index choice. Addressing these concerns, we adapt recent developments in the stochastic dominance literature to test for unambiguous relations between current and past distributions of toxic releases. Using EPA data from 1988-1999, we find statistically significant evidence that the unconditional 1999 distributions of air, land, underground and total toxic releases dominate in a first-degree sense their respective 1988 distributions. While some of this improvement is explained by economic growth, pollution net-of-income improved over the sample period as well. Finally, we document robust differences in the distribution of pollution across regions in the US. Copyright © 2004 John Wiley & Sons, Ltd.