Asset Allocation Via Clustering or, Just How Useful Is My Stylebox? And Are Most Hedge Funds Really the Same?
SBV Research Working Paper, 2020
25 Pages Posted: 19 Nov 2020 Last revised: 7 Jan 2021
Date Written: October 30, 2020
Abstract
In order to take advantage of the “one true free lunch” in investing, namely the increase in compound return due to the reduction of volatility by regular re-balancing among uncorrelated assets, it is necessary to first establish what those uncorrelated asset classes are. In practice, many investors are disappointed to find their attempts at “style-box” diversification (e.g. large vs. small, and growth vs. value) failed to provide much overall portfolio benefit in terms of this re-balancing effect. Machine-learning algorithms such as cluster identification via hierarchical trees have seen some success in identifying truer sector and industry classifications among individual equities; can such algorithms provide any insights as to the diversification benefits of standard asset classes? We apply various clustering algorithms to asset classes and hedge fund strategies from 1990 to the present to investigate cluster stability and compute the returns of a risk-parity investing approach. Some surprising insights emerge with actionable implications for portfolio construction (a few examples: there are three types of bonds; TIPS are not different; High Yield is mostly Equity; MLPs and Precious Metals are interesting; and all hedge funds except Merger Arbitrage and certain sub-categories of Macro have morphed over time into Equity).
Keywords: clustering, diversification, portfolio construction, asset allocation, hierarchical risk parity, hedge funds, asset classes, machine learning
JEL Classification: C38, D81, G11
Suggested Citation: Suggested Citation