Serafeim, GeorgeRader, KevinSerafini, Fabrizio2024-07-1220242024-06-122024Serafini, Fabrizio. 2024. Corporate Disclosures Decoded: Forecasting Real Decarbonization Rates. Bachelor's thesis, Harvard University Engineering and Applied Sciences.31144103https://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37379282Companies can report their carbon emissions via voluntary disclosure surveys, a method gaining global traction. However, the capacity of these surveys to identify actions predictive of future decarbonization for individual firms remains largely unexplored. This study examines disclosure data from the Carbon Disclosure Project (CDP) Climate Survey to forecast decarbonization rates for 3,574 firms from 2011 to 2022. Using machine learning techniques including Mixed Effects, Bayesian Ridge, and Gradient Boosting regression, this is one of the first attempts to leverage carbon disclosure data for predictive analysis at the firm-level. We identify a set of crucial predictors associated with next-year decarbonization, including the firm's sector, country, emission targets across Scope 1 and Scope 2, reporting of Scope 2 market-based emissions, verification of Scope 3 emissions, the application of internal incentives, and a Marginal Abatement Cost Curve (MACC) for emission reduction initiatives. Our study offers critical insights for investors, policymakers, and companies on firm-level actions that indicate future decarbonization, alongside recommendations to enhance the efficacy of future disclosure surveys and the predictive validity of corporate disclosure scores. Overall, we establish a foundational framework for future research to identify disclosed firm-level indicators of future decarbonization.application/pdfenCarbon DisclosuresClimate ChangeDecarbonizationFirm LevelMachine LearningMarginal Abatement Cost CurveStatisticsComputer scienceThesis or Dissertation2024-07-120009-0005-2125-2922