Exploiting the low-risk anomaly using machine learning to enhance the Black–Litterman framework: Evidence from South Korea

Abstract

Many studies have revealed that global financial markets are experiencing low-risk anomalies. In the Korean market, for example, even the portfolios of high-risk stocks recorded a loss of about 70% between 2000 and 2016. In this study, we construct a low-risk portfolio that responds to low-risk anomalies in the Korean market using the Black–Litterman framework. We use three machine-learning predictive and traditional time-series models to predict the volatility of assets listed in the Korean Stock Price Index 200 (KOSPI 200) and select the best-performing one. Then, we use the model to classify assets into high- and low-risk groups and create a Black–Litterman portfolio that reflects the investor's view where low-risk stocks outperform high-risk stocks. The experiment shows that reflecting the low-risk view in the market equilibrium portfolio improves profitability and that this view dominates the market portfolio.