1. Introduction
Studies of the environmental Kuznets curve (EKC), which posit that an inverted-U relationship exists between a measure of wealth and environmental degradation, have attracted increasing attention in the literature.1 Numerous studies have examined the issue (e.g. Hettige et al., 1992; Shafik and Bandyopadhyay, 1992; Panayatou, 1993; Cropper and Griffiths, 1994; Selden and Song, 1994; Antle and Heidebrink, 1995; Grossman, 1995; Grossman and Krueger, 1995; Holtz-Eakin and Selden, 1995; and the special issue of Ecological Economics, 1998), but perhaps the most convincing sign of the EKC’s significance is that interest has extended well beyond academic circles (see, e.g. Arrow et al., 1995).
A plethora of the academic studies find that some pollutants adhere to the inverted-U hypothesis. With this evidence in mind, it may be tempting to generalize such results and argue that the ‘way to attain a decent environment in most countries is to become rich’ (Beckerman, 1992). Although this premise is appealing, the EKC model has noteworthy limitations. First, the inverted- U relationship appears to hold for some pollutants, but it has not been found to be a particularly accurate depiction for all pollutants. For example, Shafik (1994), Holtz-Eakin and Selden (1995), and Roberts and Grimes (1997) find that carbon emissions fail to follow an inverted- U path. Second, if the estimated turning points occur at exceedingly high levels of wealth, the environmental benefits of economic growth may be unachievable for many countries. Third, some studies find that when alternative variables are included in the EKC specification, the estimated coefficients of the EKC equation either diminish in significance or no longer adhere to an inverted-U (Kaufmann et al., 1998; Rothman, 1998; Torras and Boyce 1998).
Another limitation of the existing EKC literature rests with the nature of the data under examination. Due to the lack of available data, studies have traditionally estimated EKCs with crosscountry panel data. Given that the quality of such data is often questionable, the empirical results obtained may be suspect. Furthermore, since the common method of estimation with panel data assumes that all cross-sections adhere to the same EKC, if cross-sections vary in terms of resource endowments, infrastructure, etc., it may be unreasonable to impose isomorphic EKCs (see Unruh and Moomaw, 1998). We address these and other issues using a new panel data set on state-level sulfur dioxide (SO2) and nitrogen oxide (NOx) emissions from 1929–1994. Our analysis focuses on two hypotheses. First, do emissions at the state-level follow the inverted-U shape proposed by the previously cited cross-country studies? Yes, we find that US states have undergone the familiar environmental degradation followed by environmental amelioration found in many recent cross-country studies. Second, is it appropriate to restrict states to follow isomorphic EKCs? No, empirical results suggest parameter estimates will be miscalculated if the modeler assumes interstate slope homogeneity.