We estimate the causal impact of the EU Emissions Trading Scheme on manufacturing firms using comprehensive panel data from the German production census. Semiparametric matching estimators yield robust evidence that the policy caused treated firms to abate one-fifth of their CO 2 emissions between 2007 and 2010 relative to non-treated firms. This reduction was achieved predominantly by improving energy efficiency and by curbing the consumption of natural gas and petroleum products, but not electricity use. We find no evidence that emissions trading lowered employment, gross output or exports of treated firms. * We owe a debt of gratitude to Katrin Rehdanz for countless discussions and her generous support at all stages of this project. We thank the research data centre (FDZ) for granting us access to the AFiD data. We thank FDZ staff members Diane Zabel, Michael Rößner and, especially, Alexander Vogel of the Statistical Offices of the Länder in Bremen, Halle and Kiel, respectively, for expert advice regarding various AFiD modules and for technical help with the data matching and estimation. Staff at the Federal Statistical Office graciously provided information on various statistical products. We thank, without implicating,