Dynamic rebalancing portfolio models with analyses of investor sentiment

Abstract

This study extends the risky portfolio models that use historical returns while incorporating investor sentiments in optimizing asset allocations. To ensure the practicality, the portfolio models that are suitable to strategize a large number of assets are developed by advancing two linearized objectives, the Omega ratio and Conditional Value-at-Risk (CVaR). Transaction costs, short selling, and adjustment of lower bounds of weights are considered in dynamic portfolio rebalancing. More than ten million messages from Twitter are analyzed to generate sentiment scores in managing portfolio rebalancing. Using the data of the S&P 500 composite stocks, our empirical results show the sentiment-triggered dynamic rebalancing portfolios ex post outperform their corresponding fixed-period rebalancing models and the naïve diversification portfolio. The flexibility in adjusting the asset allocations according to investor sentiments improves realized portfolio performance. The proposed models demonstrate self-correction by detecting investor sentiments, resulting in more effective asset management.