The Environmental Performance of Polluting Plants: A Spatial Econometric Approach

This paper uses plant-level data from EPA and Census databases to incorporate spatial components in models of the environmental performance (compliance and emissions) of manufacturing plants. The spatial approaches include creating explanatory variables based on information related to the plant’s location and the location of nearby plants, testing for spatial autocorrelation in environmental performance and the explanatory variables, examining whether spatial factors in the explanatory variables explain the spatial factors in environmental performance, and directly modeling the spatial component of environmental performance with spatially lagged dependent variables. Our analysis uses data for 521 manufacturing plants in the compliance analysis and 102 plants in the emissions analysis. Our results indicate a significant, but limited, role for spatial factors in modeling environmental performance. We find positive spatial autocorrelation in the analysis of compliance status, but no such spatial effects are observed for emissions. In fact, very few variables in our model show significant impacts on emissions. The positive spatial autocorrelation of compliance status holds in most cases, whether or not we control for other explanatory variables, though these results depend in part on the way in which the spatial effects are assumed to operate across plants. Some of the models indicate that the spatial autocorrelations in the explanatory variables help explain part of the spatial autocorrelation observed in compliance status, though much of the autocorrelation in compliance usually remains unexplained. Models which directly incorporate spatially-lagged compliance status in the estimation often (but not always) find significant positive effects for those spatial lags. Much of the explanatory power of the compliance models comes from plant-specific characteristics, with larger, older, and more pollution-abatement-intensive plants, and those in single-plant firms, showing less compliance. Neither the local demographic characteristics nor measures of political activity have much impact on compliance. Our measures of inspection activity tend to have the expected signs – having more inspections at the plant, at nearby plants, and at plants in the same state is associated with greater compliance – but these effects aren’t always significant. As expected, inspections at nearby plants in other states don’t seem to increase compliance.