Analyst vs. Machine Learning in Implied Cost of Capital Estimations by Minghui Chen :: SSRN

This paper advances the estimation of the implied cost of capital (ICC) by improving analysts' multi-horizon earnings forecasts using machine learning algorithms, which exhibit reduced bias and higher forecast accuracy. The model-based ICC demonstrates stronger cross-sectional relationships with future realized returns, offering a better proxy for ex-ante expected returns. Additionally, stocks with underestimated ICCs outperform those with overestimated ICCs. Weaker information environments primarily drive the ICC differences between analyst-based and model-based ICCs. Furthermore, larger ICC differences are associated with lower earnings announcement returns. This study underscores the potential of machine learning to refine ICC estimates and address the importance of ICC differences.