Poster Paper: Explain State Environmental Policy Making By a Spatial Model

Thursday, November 7, 2013
West End Ballroom A (Washington Marriott)

*Names in bold indicate Presenter

Shuang Zhao, Indiana University and Jialu Liu, Allegheny College
Explain State Environmental Policy Making by a Spatial Model

There has been a substantial shift in the devolution of federal government responsibilities to the states since the 1970s (Donahue 1997). This movement is especially strong in the case of environmental protection (Engel and Rose-Ackerman 2001). The environmental federalism features a structure that the federal government returns the responsibility of environmental regulation to the states while retaining the ultimate authority to judge the adequacy of state actions (Woods 2005). The Clean Air (CAA) and Clean Water (CWA) acts passed by the federal environmental legislation in the 1970s delegates a substantial implementation of environmental regulation to the states.

This federalism structure raised concerns among scholars who argue that competing for businesses will lead to lax environmental regulation, a phenomenon of “race to the bottom.” To the contrary, some scholars argue that instead of racing to the bottom we may actually observe a race to the top phenomenon through supply chain management or market pressure (Vogel, 1995). In this paper, we argue that state environmental policy making may be a mixture of its own political, economic and environmental conditions as well as interactions with external environment – federal government’s pressures and other states’ competition.

One of the challenges of this research is to model spatial interdependency. Without counting for spatial dependency, estimations from a multivariate model will be biased. Franzese and Hays (2007) argued that political shocks would be overestimated if interdependence processes are ignored or inadequately modeled. In our project, we will first perform spatial correlation test to determine whether a spatial autocorrelation exists and whether it is strong enough to be considered. Second, we apply a conditional autoregressive (CAR) model. The simultaneous regressive (SAR) model uses a regression on the values from the other areas to account for the spatial dependence while the CAR model mainly relies on the conditional distribution of the spatial error terms (Bivand and Gomez-Rubio V., 2008). Whereas the neighborhood matrix W in the SAR model is asymmetric, the matrix in the CAR model is symmetric. One of the disadvantages that the SAR model is that the εi will be correlated with {Z (Aj): j ≠ i}, which violates the assumptions of OLS models. A CAR model could correct this correlation problem, in which the error term ζi will not be correlated with {Z (Aj): j ≠ i}. Another advantage that the CAR model offers is that any SAR model can be represented as a CAR model; however, a CAR model could not be written as a SAR model. Therefore, the CAR model provides more flexibility than the SAR model. 

We use state level inspection and penalty data to measure the variation of state level policies. Political, economic and environmental variables are included to evaluate internal factors that drive state level policymaking. External pressure from the federal environmental protection agency (EPA) is measured by the proportion of population in nonattainment area in each state, and the CAR model estimates the interactions among states.