Skip to main content

Can Machine Learning Improve Portfolio Risk-Adjusted Performance? A Smart Beta Case Study

Investors have often asked how best to allocate into smart beta strategies to improve risk diversification. In our latest whitepaper, we demonstrate how hierarchical clustering techniques could play a role in this pursuit.

Head of Quantitative Research and Analysis, SPDR ETF Model Portfolio Solutions, EMEA & APAC
Senior Quantitative Research Analyst, ETF Model Portfolio Solutions at SPDR EMEA & APAC
Quantitative Research Analyst

In long-only equity allocations, the dominant risk typically emanates from market beta. In a bid to achieve better diversification, investors have grown increasingly accustomed to allocating into smart beta — or factor — strategies to smooth their investment journey. Aside from diversification, investors also attempt to generate outperformance by timing their allocations across different smart beta strategies, aware that the performance of individual factors are sensitive to the prevailing macroeconomic cycles (for example, see Ung and Luk (2016)1).

Once the objectives of the allocation are established, there is the additional consideration of how best to build the portfolios to achieve these objectives. In recent years, a growing body of literature has emerged on whether the allocation should be built bottom up at the stock level or top down at the strategy level. The aim of this paper is neither to assess the merits of pursuing a top-down or bottom-up strategy, nor evaluate whether smart beta is best used to diversify risk or enhance performance. Instead, our goal is to examine whether machine learning insights can play a role in improving risk diversification in a multi-factor smart beta allocation through the use of the hierarchical clustering techniques.

More on Smart Beta