We compare the most common reduced-form models used for emissions forecasting, point out shortcomings, and suggest improvements. Using a U.S. state-level panel data set of CO 2 emissions, we test the performance of existing models against a large universe of potential reduced-form models. We find that leading models in the literature, as well as models selected based on an emissions per capita loss measure or different insample selection criteria, perform significantly worse compared to the best model chosen based directly on the out-of-sample loss measure defined over aggregate emissions. The Review of Economics and Statistics, February 2012, 94(1): 172-185 © 2011 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology FORECASTING THE PATH OF U.S. CO 2 EMISSIONS 173 4The EIA uses the NEMS model to forecast the national energy system and CO 2 emissions, which are published in the "Annual Energy Outlook." Since these forecasts are provided by an agency of the U.S. government, they are considered the official forecasts. The EIA conducts an annual outof-sample evaluation of its forecasts, but it is not used to reparameterize the NEMS model. For a discussion of the NEMS model forecast evaluation, see O'Neill and Desai (2005) or Auffhammer (2007).