Editorial: Factor Investing across asset classes

printer-friendly version


EDHEC-Risk on Factor Investing for a More Efficient Harvesting of Risk Premia Across and Within Asset Classes

Lionel Martellini

Risk factors, defined as systematic underlying sources of risk that impact a large set of securities, have long been used in investment practice for analysing risk and performance of actively or passively managed portfolios. More recently, a new approach has emerged, where factors have a more explicit role in the investment process. This approach, known as factor investing, recommends that allocation decisions be directly expressed in terms of risk factors, as opposed to standard asset class decompositions.

While the relevance of factor investing is now widely accepted amongst sophisticated institutional investors, an ambiguity remains, however, with respect to the exact role that risk factors are expected to play in the investment process. In fact, the term “factor” is used with many different meanings and this can be a source of confusion, and sometimes disappointment, about the benefits that can be expected from factor investing approaches. For factor investing to gain even wider acceptance, some clarification is needed with respect to the various definitions of factors and the benefits of factor investing in an institutional context.

In a recent research paper (Martellini and Milhau (2018))[1] supported by Amundi in the context of the "ETF and Passive Investment Strategies" research chair at EDHEC-Risk Institute, an abridged version of which has been published as the lead article in the latest issue of the Journal of Portfolio Management[2], we argue that there actually exist two main types of benefits that can be expected from factor investing. On the one hand, factor investing across asset classes allows for a better structuration of the investment process, both from an asset-only perspective and from an asset-liability management perspective. On the other hand, factor investing within asset classes allows for a more efficient harvesting of risk premia, particularly when compared to traditional approaches that focus for example on sector decompositions.

Factor Investing and Risk Allocation Decisions: A More Efficient Structuration of the Investment Process

From an allocation perspective, factor investing is the process in which investors decide how much to allocate to each factor as opposed to each asset class. Even if factor allocation decisions must of course be eventually translated back into asset weights in order to be implemented, there are reasons to believe that this approach allows for a better structuration of the investment process, both from an asset-only perspective and from an asset-liability management perspective.

To provide some initial intuitive support for this claim, let us note that the first rationale for factor investing is that the focus on factors allows investors to gain a more holistic understanding of risk and performance by allowing them to analyse the commonalities in the key drivers for the returns on seemingly disparate asset classes.

One of the key problems experienced by institutional investors during the subprime crisis and the severe market downturn that has followed is indeed that even a seemingly well-diversified allocation to multiple asset classes can hide an extremely concentrated set of factor exposures. In this context, the use of risk factors with relatively high explanatory power, is critically important since it is only by framing the allocation exercise in the factor space as opposed to the asset space that investors will be able to understand how well- or how poorly-diversified their performance portfolios actually are.

Because macro-economic factors typically have low explanatory power for asset returns, implicit factors appear to be the most natural option for representing underlying sources of risk and the more appropriate tools in risk budgeting exercises. Implicit factors, extracted in a multi-asset universe via principal component analysis or minimum linear torsion techniques for a more robust outcome, can be used to build efficiently diversified portfolios. Explicit macro-economic factors like the GDP or inflation, on the other hand, can still be useful as state variables characterising market conditions, or more precisely state variables defining various regimes of economic activity, and these regimes are relevant for asset allocation if expected performance and risk of assets and liabilities vary across these states of the economy. There are well-known advantages to identifying and employing economic regimes within a consistent asset allocation framework, in particular because various asset classes have contrasted performances in different economic environments.

It can then be shown that equities tend to do best when inflation remains modest, and suffer when inflation is above the median, especially with growth below average. Government bonds, on the other hand, do reasonably well except for when growth and inflation are above average. Treasury inflation-linked bonds do best when both inflation and growth are below average, a situation in which real estate tends to have excellent returns. Commodities outperform when growth is above average, but suffer when inflation is above average and growth is below average. These historical patterns, backed by sound economic theory explaining the patterns, confirm that the construction of a well-developed portfolio should take into account the behaviour of asset categories under varying economic regimes. These relationships also demonstrate the opportunities to take advantage of patterns to improve risk-adjusted returns at the asset allocation level, and even more so when liabilities are to be taken into account given that liabilities for most pension plans are also sensitive to changes in economic growth and inflation.

Factor Investing and Benchmarking Decisions: A More Efficient Harvesting of Risk Premia

Factors are not only useful for understanding the cross-sectional and time-series determinants of risk and returns across asset classes, but they can also be used as investable building blocks within asset classes. Individual securities earn their risk premium through exposure to rewarded factors, while the remaining risk goes uncompensated. The academic literature has identified a number of rewarded factors, the existence and persistence of which seem to be robust over different time periods and across different regions – this includes the size, value, momentum, low volatility and quality factors.

The outstanding question from an investor perspective is to determine the best way to harvest such multiple risk premia. This decision is in fact embedded within the choice of a benchmark, which is then used as a reference portfolio for passive or active mandates. While cap-weighted (CW) indices are typically used as default investment benchmarks by asset owners and asset managers, they have in fact been shown to suffer from two main shortcomings. On the one hand, CW indices are ill-suited investment benchmarks because they tend to be concentrated portfolios that contain an excessive amount of unrewarded risk. On the other hand, CW indices represent bundles of factor exposures that are highly unlikely to be optimal for any investor, if only because they have not been explicitly controlled for. For example, CW equity indices by construction show a large cap bias and a growth bias, while the academic literature has instead shown that small cap and value were the positively rewarded risk exposures. In practice, it has been shown that the use of investable forms of smart factor indices allows for substantial improvements in risk-adjusted performance compared to these traditional benchmarks (see for example Amenc et al. (2014)[3]).

It is fair to say that the smart beta approach is now firmly grounded in equity investment practices, and the key question for an increasing majority of institutional investors is not whether one should use smart beta, but instead which and how much smart beta to use. In contrast, the concept of smart beta in the fixed-income space is still relatively less mature, despite the obvious importance and relevance of the subject. Over the recent years, a number of concerns have been expressed, however, about the (ir)relevance of existing forms of corporate and sovereign bond indices offered by index providers, both in term of lack of diversification and absence of control of underlying factors exposures. More generally, it appears that existing bond indices can be regarded as more "issuer-friendly" than "investor-friendly". The rationale behind such thinking is that these bond indices passively reflect the collective decisions of issuers regarding the maturity and size of bond issues, with no control over risk factor exposures associated with such choices nor over the reward that investors should deserve from holding a well-diversified portfolio of such factor exposures. In this context, there is an increasing understanding that improved bond benchmarks are required, which will provide adequate answers to investors' needs through the construction of investable proxies for rewarded risk factors. This would include interest rate risk factors (level of the yield curve, slope of the yield curve, curvature of the yield curve) and credit risk factors, but also liquidity risk factors, low risk factors, carry factors, value factors, momentum factors, etc., that are suitably defined in fixed-income markets. This question is the subject of ongoing research at EDHEC-Risk Institute, also supported by Amundi.

Smart Beta and Beyond: Maximising the Benefits of Factor Investing

A comprehensive risk allocation framework that can be used by institutional investors to formalise the factor investing process across and within asset classes in a coherent manner. In a nutshell, the framework involves two steps. In a first step, the combined use of implicit factors that explain cross-sectional differences in risk and return parameters, as well as explicit macro-economic factors that explain their changes over time, allows asset owners or asset managers servicing them to implement more efficient allocation decisions. In a second step, when the allocation decisions are translated back into more traditional asset class decompositions, the use of smart factor indices as building blocks allows for a more efficient harvesting of risk premia.

While these two steps have been related in this discussion to a factor investing effort across and within asset classes, respectively, let us note in conclusion that the situation is in fact more subtle. One the one hand, factor risk parity techniques can be applied to smart factor indices but not to broad asset classes, so implicit and macro-economic factors can and should impact the allocation to investable proxies for rewarded micro-economic factors within asset classes. On the other hand, as opposed to using investable proxies for risk premia in equity and fixed-income markets, one may seek to identify investable portfolios replicating factors that influence returns in several classes. These questions represent a fertile area of investigation both for academic research and investment practice, and we expect more innovation in the years ahead in terms of how factor investing can to lead to further developments in welfare-improving investment solutions for institutional and individual investors alike.

We refer interested readers to the paper Factor-Based Commodity Investing, by Nikolaos Tessaromatis and Athanasios Sakkas.

[1] Martellini, L. and V. Milhau. 2018. Smart Beta and Beyond: Maximising the Benefits of Factor Investing. EDHEC-Risk Institute Working Paper.

[2]  Martellini, L. and V. Milhau. 2018. Proverbial Baskets are Uncorrelated Risk Factors! A Factor-Based Framework for Measuring and Managing Diversification in Multi-Asset Investment Solutions. Journal of Portfolio Management 44(2): 8-22.

[3] Amenc, N., R. Deguest, F. Goltz, A. Lodh and L. Martellini. 2014. Risk Allocation, Factor Investing and Smart Beta: Reconciling Innovations in Equity Portfolio Construction. EDHEC-Risk Institute publication.