Alternative risk premia strategies in the face of extreme risk and return asymmetry (Note)

Purpose of the article:This note aims to highlight the limitations and true nature of the risk associated with quantitative Alternative Risk Premia strategies, which aim to deliver regular returns that are uncorrelated with those of traditional assets. In particular, the article will explain the main reasons that led to the collapse of these strategies during the crash of February/March 2020.

Summary:

• This note is an expanded and detailed version of the article entitled « Quantitative alternative risk premia strategies put to the test » originally published in AGEFI Hebdo No. 772, September 23-29, 2021, p. 31.
• First appearing in the early 2010s, quantitative alternative risk premia (ARP) strategies aim to deliver regular returns that are uncorrelated with those of traditional assets by taking advantage of multiple risk premiums across different asset classes.
• The decline in the performance of most of these strategies from 2015/2016 onwards and their collapse during the crash of February/March 2020 calls into question both their ability to deliver regular returns and their decorrelation from traditional assets.
• ARP strategies exploit the factors that explain the performance of risky assets beyond their sensitivity to the benchmark market (risk premiums).
• The various basic investment strategies, which aim to take advantage of the different risk premiums identified, follow algorithms based on historical market data. The ARP strategy portfolio is most often managed using a quantitative process.
• An overinvestment effect and biases affecting investment strategies based on the use of historical data certainly explain the decline in performance from 2015/2016 onwards.
• Finally, the crash of February/March 2020 served as a reminder that the search for strategies that are uncorrelated with traditional assets and deliver regular returns leads to a risk profile that is sensitive to extreme risk and characterized by asymmetry between limited gains and much higher potential losses.

In the early 2010s, in the wake of the subprime crisis and its impact on financial markets, the world of systematic and quantitative management saw the emergence of alternative risk premium strategies and funds, or « alternative risk premia » ( ARP).

These strategies seek to exploit the factors (or sources of return) that explain why, beyond sensitivity to the benchmark market (or « beta »), certain sectors/stocks or investment strategies outperform in the long term. These factors, or identifiable sources of performance in the main traditional asset classes, may be linked, for example, to the carry effect , momentum, value, or volatility (see the Definition section at the end of the article). In the case of alternative risk premiums, each source of performance is in principle exploited through long/short positions in order to cancel out exposure to the underlying market (so-called « market neutral » strategies ).

The objective of alternative risk premia funds and strategies is therefore to offer investors consistent performance that is uncorrelated with that of traditional assets by investing in a portfolio composed of multiple alternative risk premiums on different underlying assets.

The rationality of the approach and the performance recorded by these strategies in the early 2010s enabled them to enjoy a certain degree of commercial success. However, almost all ARP strategies were caught off guard by the crash of February/March 2020, even though their performance was already showing signs of slowing down prior to this event. To understand the reasons for this setback, it is necessary to revisit the very design of these strategies and the reality of the risk involved.

1) Presentation and performance of quantitative alternative risk premia strategies

1.1 Basis and operating principle of ARP strategies

ARP strategies are based in particular on the seminal work of Fama and French (1993) and Carhart (1997)[1], who highlighted that beta —or market risk premium—was not the only factor explaining the performance of risky assets. Other risk premiums could be identified as potential performance drivers.

For example, Fama and French (1993) highlighted that the factor linked to the relatively low capitalization of certain securities (known as « Small Minus Big » or SMB) could explain, beyond market risk, the performance of an equity portfolio (the authors also consider a third factor in their three-factor model). This source of performance therefore represents the risk premium paid to investors in small-cap stocks. The persistence of this source of return leads us to believe that small-cap stocks tend to outperform large-cap stocks in the long term. This is the SMB factor or premium.

To take advantage of this premium in the context of an ARP fund, the investment strategy will seek to systematically buy small-cap stocks and short large-cap stocks. In doing so, market exposure will in principle be minimal and the strategy will be said to be « market neutral. » The objective of these positions is to take advantage of the risk premium carried by small-cap stocks compared to the lower risk premium of large-cap stocks. To achieve this, the performance of small-cap stocks must exceed that of large-cap stocks. Having neutral exposure to the market should, in principle, enable the objective of offering an investment strategy that is uncorrelated with the performance of traditional assets, while still being invested in them.

This example shows how a particular ARP strategy is constructed. In practice, ARP strategy managers seek to identify a large number of risk premiums in order to invest in multiple ARP strategies. Each specific ARP strategy (exploiting a particular risk premium) then constitutes a basic investment building block in the construction of a portfolio composed of multiple ARP strategies.

The identification of performance sources to be exploited and the development of financial strategies to exploit them are based on the analysis of historical market data. Basic ARP strategies thus follow an algorithm whose performance has been tested and optimized based on historical prices (backtesting).

The management of the ARP strategy portfolio and exposures is also most often carried out through a systematic process based on quantitative analysis of historical data. These analyses should enable the creation of a diversified portfolio of alternative risk premiums (uncorrelated or weakly correlated with each other), the combination of which will, in principle, offer performance that is both persistent and uncorrelated with that of traditional asset classes.

This approach should not only enable the management objective to be achieved, but also offer an alternative to traditional hedge fund strategies (alternative management) and even active management by systematizing the capture of « structural alpha, «  in other words^, the portion of performance not explained by « beta » or the manager’s talent.

1.2 The performance of ARP strategies put to the test

By offering tailored investment solutions, investment banks and asset managers (including Eraam, BlackRock, Allianz GI, AQR, CFM, and La Française IS) have succeeded in attracting investor interest and have thus enabled strategies based on alternative risk premiums to achieve a certain degree of success.

The Eurekahedge Multi-Factor Risk Premia Index shows that the strong performance of alternative risk premia strategies between 2010 and 2015 (performance of over 40% between July 2010 – when the index was launched – and July 2015) certainly contributed to their commercial success and led to the launch of several dedicated funds in Europe in the middle of the last decade. Despite these encouraging beginnings, the index’s performance gradually slowed from 2015/2016 onwards, before entering a downward trend from 2018.

As the promise of consistent returns began to be called into question, the collapse in February/March 2020 of most alternative risk premia funds and strategies (including those that had previously managed to maintain a positive trend) shed a harsh light on their sensitivity to market fluctuations, at the worst possible moment, during the crash. Finally, the situation was made worse by the inability of most of these strategies to take advantage of the strong rebound in the equity and bond markets following the crash to the same extent.

2) Limitations of the approach and reality of the risk profile

This rout can be explained by several factors.

2.1 The overinvestment effect

First, alternative risk premia strategies were certainly victims of their own success. Concurrent and relatively large investments in multiple risk premiums reduced their potential gains while tending to increase their potential for re-correlation. By becoming overinvested or « overcrowded » ( see in particular Bouchaud et al. (2020) on this point[4]), alternative risk premia were no longer able to deliver the historical level of performance that had helped attract investors^[5].

Investing heavily and simultaneously in the same assets tended to make those that were bought more expensive and those that were sold cheaper. The performance potential of the risk premiums underlying these strategies therefore tended to decline. The reflexive effect certainly explains the decline in performance and even the negative trend from 2018 onwards.

2.2 The  » backtest overfitting  » effect

Secondly, as highlighted in particular by Harvey et al. (2019)[6] and Arnott et al. (2019)[7], biases linked to the use of historical data – « backtest overfitting » – may also explain the discrepancies between expected and actual returns. Quantitative optimization based on past data can lead to the selection of strategies chosen more for their ability to « stick » to historical data and take advantage of it than for their real potential for future performance. Even though regulations require a reminder that « past performance is no guarantee of future results, » investing on the basis of a backtest nevertheless implies hoping that this will indeed be the case…

As noted by Bailey et al. (2016)[8], the phenomenon of « backtest overfitting » can be considered the main reason why quantitative investment strategies that appear attractive on paper (i.e., in simulations based on historical data) often prove disappointing in practice. Furthermore, investing in strategies with strong historical performance can lead to « buying » certain market behaviors at their peak.

2.3 Extreme risk and asymmetry of potential gains and losses

Thirdly, the search for risk premiums that deliver consistent long-term performance (risk premiums that can be likened to building blocks used to construct the overall portfolio) leads to a search for strategies whose risk/return profile is characterized by a negative skew (asymmetry coefficient)^[9]. In other words, the strategies or building blocks that make up the portfolio will be characterized by delivering regular gains while being subject to a rare or unlikely risk of very large adverse movements. This is known as tail risk.

In the world of traditional management,fixed income strategies are an example of a strategy characterized by negative skew. These strategies invest in debt securities offering a certain rate of return against the risk of default, which could result in a loss potentially far greater than the return on the security.

Similarly, most basic ARP strategies tend to favor regular returns in exchange for a proportionally much higher but, in principle, unlikely risk of loss.

Naturally, managers will seek to select strategies that offer « sufficient » compensation for tail risk. In addition, building a portfolio based on a large number of risk premiums (which are in principle weakly correlated with each other) should help to limit the tail risk of the portfolio as a whole. Finally, certain basic strategies used (such asmomentum ) are in principle characterized by positive skew^[10].

Quantitative optimization and systematic investment processes should therefore ultimately result in a portfolio consisting of multiple alternative risk premiums, delivering a steady return that is uncorrelated with that of traditional assets and whose tail risk is, in principle, diluted. However, this risk, which is linked to rare market events (such as the bankruptcy of Lehman Brothers in 2008 or the crash of February/March 2020) and is inherently unpredictable, remains even if it does not appear in historical data.

The risks of recorreation, ineffective hedging, and black swan events affecting several basic strategies simultaneously have not disappeared simply because quantitative risk analyses cannot identify them. Lavoisier’s famous formula, « Nothing is lost, nothing is created: everything is transformed, » also applies to risk in market finance.

Risk is actually transformed and transferred. Sensitivity to market fluctuations (which is associated with a certain symmetry between the potential for gain and loss and the respective probabilities of realization) is transformed into relative insensitivity under « normal conditions » (so as to favor a high probability of regular returns) at the cost of a risk of loss that is unlikely but very high compared to the gains.

Favoring a portfolio that generates regular returns that are not very sensitive to fluctuations in traditional underlying assets therefore means accepting an asymmetry between a regular gain—which is relatively small compared to the risk incurred—and a significantly higher potential loss in the event of an unfavorable situation (even if this situation is considered unlikely).

Thus, the persistent asymmetry between potential gains and losses—as well as the risk limits sometimes reached—prevented alternative risk premia strategies from taking full advantage of the strong market rebound following the February/March 2020 crash. Indeed, the management objective of ARP strategies remains constant even at the height of market declines. To quickly offset the heavy losses incurred during the crash, it would have been necessary to favor direct market exposure (beta) without any certainty that the market would rebound. However, the objective of these strategies remains to deliver consistent returns that are uncorrelated with traditional assets, which means remaining « market neutral » and therefore minimizing market exposure. Managers and exposure management algorithms therefore continued to favor reduced or zero exposure in line with the management objective.

We might add that systematic management algorithms are most often calibrated to perform optimally over the long term and therefore under market conditions most often observed in the past. They are therefore not usually calibrated to take advantage of rare, highly improbable, and unpredictable events that do not appear in historical data.

Conclusion

Quantitative optimization—or backtesting bias — and the search for regular returns uncorrelated with traditional assets have led ARP strategies to be invested in portfolios characterized by an asymmetry between relatively low potential gains and significantly higher potential losses (although unlikely in principle). The crash of February/March 2020 thus served as a further reminder of the limitations of this type of approach and revealed the true nature of the extreme risk present in these portfolios.

Victims of their own success, certain biases inherent in quantitative management, and the radical uncertainty inherent in financial markets, alternative risk premia strategies will certainly have to reinvent themselves by focusing on better control of extreme risks and assuming greater exposure to underlying markets.

Definition

Momentum strategies:a strategy that involves buying the best-performing assets and selling the worst-performing assets based on various technical and/or quantitative indicators.

Value strategy :a strategy that involves buying securities whose market price appears low relative to their intrinsic value or book value.

Carry strategy:a strategy used in particular on the foreign exchange market, which consists of borrowing in a low-interest-rate currency and investing in a higher-yield currency.

Volatility strategy:there are several volatility strategies; one of them consists of buying stocks with low volatility and selling those with higher volatility, as low-volatility assets tend to perform better in principle.

Market neutral strategy :a portfolio managed in such a way that the market exposure induced by long positions is hedged by short positions in the same asset class.

Backtesting:performing a backtest involves using historical market data to develop or test one or more quantitative investment strategies.

Reflexive effect:as highlighted by George Soros (1998)[11], the actions of market operators based on a particular observation influence that same observation. Acting to take advantage of an investment opportunity influences the potential of that same investment opportunity. This is a reflexive effect.

Black Swan :developed and defined by Nassim Nicholas Taleb (2008)[12], a Black Swan is a random, highly improbable and unpredictable event with a particularly significant impact.