Causal emissions factor benchmarks during contingent nuclear outages

Abstract One method to decrease carbon dioxide (CO 2 ) emissions in the electricity sector is to shift the timing of electric load from periods when it causes more pollution to times when it causes less. A key step in implementing this ‘load-shifting’ strategy involves measuring changes of emissions as a consequence of changes in electricity load. While multiple modeling algorithms have previously estimated this quantity, previous methods have been forced to either calibrate against dispatch modeling assumptions, or against statistical models that might not be free of causal confounders. Further, none of these algorithms have been tested in situations where the changes in load is large, as is typical in the load-shifting context. In this work, we demonstrate a technique to estimate this consequential change in emissions using direct empirical evidence. We do so by leveraging a causally valid natural experiment involving unplanned nuclear power plant outages. The values obtained in this paper can serve as a baseline with which to help evaluate the performance of the modeling algorithms in the regimes important for load-shifting.