Can Portfolio Construction Considering ESG Still Gain High Profits?

This paper proposes a new approach to constructing investment portfolios based on Environmental, Social, and Governance (ESG) factors. The approach uses machine learning models and a Markov Switching Variance (MSV) model to predict returns and volatility, and a Multi Criteria Decision-Making (MCDM) model, named TODIM, to select superior stocks. An optimization model based on the Mean-Variance model considering ESG is then used to find optimal weights for each asset. The goal is to find more profitable and stable investment opportunities by considering a broader range of factors for investors, and the proposed approach is compared to several benchmarks and outperforms them in terms of profitability and risk metrics. Results in two different markets show that ESG-oriented investment can be profitable in the short term, and the use of features such as regime separation and future-encompassing information in addition to historical data enhances the outcomes of the proposed framework. [Display omitted] •Designing a highly profitable portfolio construction approach based on ESG factors.•Training predictors on the separated regimes through MSV, can positively affect the performance of the model.•Compared to classic portfolio construction models, the proposed model considers more criteria than historical prices, such as predicted return, volatility, correlation, and the relationships among different assets.