Distributionally Robust Online Portfolio Selection with ESG Scores

Online portfolio selection (OPS) is gaining increasing attention since it responds better to financial market volatility and efficiently averts investment risk through real-time updating. To alleviate the impact of financial environment uncertainty on online decision making and improve investment efficiency, we propose a novel distributionally robust online portfolio selection (DROPS) strategy by two stage optimization. In Stage 1, a sector portfolio selection is performed, considering various sectors with different financial market characteristics. Specifically, two distributionally robust Mean-CVaR models are constructed for determining the allocation weight of each sector in each month, where risk preference parameters are dynamically adjusted based on past investment performance. In Stage 2, a daily portfolio selection is conducted on individual stocks. Given that environmental, social, and governmental (ESG) factors exert an influence on returns, the daily ESG scores are first incorporated into the auto-regressive integrated moving average (ARIMA) model for boosting return prediction accuracy. The ARIMA-ESG-Cost algorithm is then proposed to update the portfolio for maximizing net returns. Numerical experiments demonstrate that the DROPS strategy achieves higher cumulative wealth and outperforms a wide range of OPS strategies on multiple composite metrics of risk and return, exhibiting strong practicability in real investment activities.