This thesis contributes to the burgeoning research field of sustainable finance. The first part of the thesis aims to address some of the criticisms raised by investors, regulators, and stakeholders on ESG ratings. Chapter 1 develops a backtest methodology to evaluate ESG ratings. Our empirical applications indicate that the informativeness of ESG ratings strongly depends on the location of firms' headquarters and the level of consensus among rating agencies. Chapter 2 proposes a new methodology for the production of ESG ratings based on supervised learning. The methodology has the advantage of eliminating any confusion regarding what ESG ratings truly measure. Our empirical applications highlight its usefulness, but also the limitations of...