Network-based online portfolio selection with ESG scores

Online portfolio selection is crucial for asset enhancement and risk control in contemporary financial market where information is renewed every second. Traditional approaches tend to ignore the impact of environmental, social, and governmental (ESG) factors on online investment decision making. This work studies the sustainable online portfolio selection problem by innovatively proposing the network-based online portfolio selection (NetOPS) strategy, which considers ESG factors to satisfy investors’ demands for high returns, effective risk management and strong ESG performance. Specifically, the mean–variance model with network constraint is constructed to improve investment return and avert risk, constraining that any high-correlated risky assets cannot be selected simultaneously. In the meanwhile, the ESG based mean–variance model with network constraint is proposed to enhance sustainable performance. The integrated Return-Risk-ESG model incorporating financial and ESG performance of portfolio is subsequently formulated, and transformed into a mixed-integer quadratic programming. To formulate the network constraint which is used in the integrated optimization model, the dynamic asset network is first constructed by exploring the correlation of risky assets in terms of returns and ESG scores, where triangulated maximally filtered graph algorithm and Bootstrap resampling technique are used to filter uncritical edges. Then, a network-based exponential smoothing (NES) method is designed for predicting assets’ future returns and ESG scores, incorporating the mutual impact among assets. Numerical experiments demonstrate that the NetOPS strategy exhibits an effective integration of robust profitability and sustainability throughout the investment cycle, providing investors with an option to strike a balance between financial returns and sustainable performance.