Behavioral Portfolio Management with Layered ESG Goals and Ai Estimation of Asset Returns

Abstract

Portfolio managers and individual investors alike are in quest of efficient asset allocation models that simultaneously express environmental, social, and governance (ESG) considerations along with investor behavioral biases. The current study presents a novel approach to optimize the behavioral portfolio management model in the presence of investor biases for ESG sustainability, loss aversion, and cognitive dissonance. We extend the factor pricing literature by implementing a factor extraction protocol to identify three unique and pervasive ESG factors. Upon examining the interconnectedness of the factors, machine learning methods are applied to a production-theoretic six-factor Fama and French model to predict individual asset returns. Enumeration of efficient asset allocations is obtained by solving a hierarchical multiobjective portfolio optimization model. simulation results from solving alternate specifications of the layered goal ESG-driven model corroborate and extend emergent research on portfolio sustainability, network theory, and the interconnectedness of financial returns. Additionally, we provide results to amplify the existence of a hump-shaped ESG efficiency frontier. The results provide new information about the trade-offs available for resolving cognitive dissonance when investors are conflicted between holding ‘green’ versus ‘brown’ asset diversification plans.

Keywords: Behavioral Portfolio Management, Multiple Objective Optimization, ESG-efficient frontier, Networks, Asset Returns

JEL Classification: G11, C61, C58, C45