US Funds’ returns-based ESG extraction and implementation: a multifaceted quantile regression approach

Abstract

ABSTRACT This study introduces a novel ESG intrinsic-based return factor and its application in asset pricing. This factor is extracted using a parallelized rolling window estimation and extreme value-weighted quantile portfolios. It carries a positive risk premium, indicating that investors are willing to assess its risk exposure. We further show that higher returns can be obtained in the top 30% quantiles using a long-only trading strategy. We apply a Monotone Composite Quantile Regression Neural Network (MCQRNN) model to explain US fund returns and address the needs of investors seeking to optimize their investment strategies. This model surpasses traditional benchmark models by performing deep quantile estimation and considering the nonlinear relationships between fund returns and six firm-based characteristics. This approach empowers investors by explaining the core principles of impact investing and highlighting how our constructed ESG risk factor can generate competitive returns even in volatile markets when its risk is well assessed. Highlights A sustainable smart beta factor construction is ensured ESG Long-only trading strategy grants higher returns in the top 30% quantiles Using ML, we further assessed the need to implement ESG factors into asset pricing Positive ESG risk premium reveals the investors’ interest in impact investing US market remains promising for positive ESG investment signals