Imperfect Markets versus Imperfect Regulation : Reevaluation of an Emissions Trading Program: Reclaim

This paper revisits an earlier estimation of an emissions market RECLAIM in Los Angeles. Different from the traditional matching approaches in Fowlie, Holland and Mansur (2012), we employ machine learning-based methods to reevaluate the average treatment effect of the program, and in this way, to capture heterogeneity. Our results differ from earlier results and are statistically insignificant in some cases. We further reexamine the program design from treatment allocation perspective. We apply policy learning to adjust the treatment assignment policy and compare the newly learned assignment policies with the original program. Results show that the newly learned policies are statistically better than RECLAIM and are statistically significant. Our conclusion is that RECLAIM is inelastic and cannot reduce emissions as expected