Exploring the equity–bond relationship in a low-rate environment with unsupervised learning

Lucas Baynes and Giulio Renzi-Ricci

Abstract

Some investors have become concerned about the low-interest-rate environment and its impact on the role that bonds play in a multi-asset portfolio. In order to analyze the equity hedging property of government bonds, we apply a simple but powerful machine learning technique called k-means clustering to periods with low interest rates. Our findings show that government bonds have historically acted as intended in an equity–bond portfolio, with typically positive bond performance in equity-down scenarios. Although there are some periods in which both equities and bonds fall, these can be viewed as ordinary parts of market volatility and distinct from the typical outcomes that can be considered as recurrent market states. The results of this alternative and complementary approach – which has not, to the best of the authors’ knowledge, been used before to study equity–bond diversification benefits – supplement the existing literature by providing further evidence of the added value that bonds bring to a strategic multi-asset portfolio.