A Spatiotemporal Fixed Effects Estimation of U.S. State-Level Carbon Dioxide Emissions

One of the major shortcommings of past environmental Kuznets curve (EKC) studies is that the spatiotemporal aspects within the data have largely been ignored. By ignoring the spatial aspect of pollution emissions past estimates of the EKC implicitly assume that a region’s emissions are unaffected by events in neighboring regions (i.e., assume there are no transboundary pollution emissions between neighbors). By ignoring the spatial aspects within the data several past estimates of the EKC could have generated biased or inconsistent regression results. By ignoring the temporal aspect within the data several past estimates of the EKC could have generated spurious regression results or misspecified t and F statistics. To address this potential misspecification we estimate the relationship between state-level carbon dioxide emissions and income (GDP) accounting for both the spatiotemporal components within the data. Specifically, we estimate a dynamic spatiotemporal panel model using a newly proposed robust, spatial fixed effects model. This new estimation scheme is appropriate for panels with large N and T. Consistent with the EKC hypothesis we find the inverted-U shaped relationship between CO2 emissions and income. Further, we find adequate evidence that the underlying economic processes driving carbon dioxide emissions and state-level GDP are temporally and spatially dependent. These findings offer policy implications for both interstate energy trade and pollution emission regulations. These implications are particularly important for the formulation of national policies related to the 2009 Copenhagen Treaty in which the U.S. has committed to significantly reduce greenhouse gas emissions over the next twenty years