Equities

Is factor investing still working?

Stock market exchange on the computer screen.

A large number of long-only multi-factor strategies have performed disappointingly over the past three years. Some have called into question the usefulness of solutions based on factor diversification but recent research by EDHEC suggests this doesn’t hold up against an even remotely serious investigation.

A large number of long-only multi-factor strategies have performed disappointingly over the past three years. This shift has led some commentators to call into question the very usefulness of solutions based on factor diversification, and notably the fact that the crowding effect was supposed to be the source of the disappearance of factor premia. However, we find in recent research that this suggestion does not hold up against an even remotely serious investigation. Rather it is the non-control of market beta exposure in a bull market context that has prevented factor indices from benefitting fully from the important market risk premium. It is this poor market conditionality, rather than variations in factor returns, that explains the recent disappointing performance of long-only factor offerings.

What are the drivers of the performance of factor strategies?

Commentators’ criticisms have tended to be based on assertions that have not been proven empirically, are not supported by serious academic research and ignore the very nature of factor strategy performance drivers. The latter is based on three main elements:

  1. Exposure to rewarded factors. While there are a large number of risk factors that can explain the variation in a stock or a portfolio of stocks’ returns over a period, there is a very limited number of factors that are considered to be rewarded in the sense that they not only have explanatory power over the variations in returns, but also explain the cross-sectional differences in returns of stocks or portfolios of stocks. These factors have been identified by academic research as being six in number, namely the value, momentum, size, low volatility, high profitability and low investment factors. The final two are often called quality factors.
  2. Good diversification of unrewarded idiosyncratic risk. Academic research shows that it is important for investors to strongly reduce idiosyncratic risks, those that are specific to each stock and which do not correspond to exposure to a systematic factor, because these risks are not rewarded. The usual way of reducing this in modern portfolio theory and construction is to diversify it. This allows the risk premia to be captured more efficiently.
  3. Management of systematic, non-factor risks. These risks are the undesired or implicit consequences of explicit choices of factor exposure or weighting schemes. When these implicit risk choices are not anticipated and controlled, they have significant consequences for the risk and performance profiles of factor strategies and can lead to strong differences in performance and risk for the same choice of factors over a given period.

These performance drivers have been the subject of many publications. However, one cannot but notice that in many critiques of the recent disappointing performance of factor strategies, they have been largely ignored in favour of highly-sample-dependent anecdotes and explanations that tend not to be based on rigorous observations.

Some of the explanations proposed were that some rewarded factors were no longer really rewarded (size) or that certain academic factor definitions were no longer appropriate (value) or indeed that the negative performance of factors was due to a crowding effect related to the very popularity of factor investing.

This is not the case. Rather, it is not the factors but the non-factor risks that are the source of the disappointing performance of the last three years. After analysing the performance of the factors and their consequences for the performance of a long-only multi-factor portfolio to place them in a long-term investment context, we will show the very simple impact of controlling or not controlling the non-factor risk to which all factor strategies are exposed, namely the market beta risk, on the performances of these same factor portfolios.

Performance and contribution of the factors

Within the US universe, three of the six long/short factors mentioned above have performed negatively over the past three years, namely size, value and momentum, and delivered much worse than the average negative performance observed since inception (21 June 2002). For momentum, this performance is even below its worst 5 per cent three-year rolling returns. For value, the performance is slightly above the worst 5 per cent and for size, the loss is close to the average of negative performance. On the developed ex-US universe the observation is quite similar.

However, in terms of both its value and its frequency, this negative performance is absolutely not abnormal and in no way constitutes a reason to call the premia associated with these factors into question.

With three factors out of six generating negative performance, it is expected that the analysis of the contribution of the factors as part of a multi-factor strategy will give mixed results. We therefore constructed single factor indices by using cap-weighted indices and adding a market-neutral long/short overlay and then built a multi-factor construction by aggregating these single long-only factor sleeves in the form of an equal-weighted six-factor index. This illustrated the previous results again, namely that three out of the six factors underperformed the broad cap-weighted index in the US region over the last three years, and three out of six in the Developed ex-US region.

However, if we instead use a perfect equal-weighted index, we observe that in both the US and Developed ex-US regions, this long-only factor construction outperformed the cap-weighted index over the same timespan (though admittedly to a lesser extent than over 15 years – 0.51 per cent compared to 2.45 per cent annual relative return for the US region and 0.55 per cent compared to 3.73 per cent for developed ex-US).

Hence, contrary to what is said by critics, a pure multi-factor construction continued to provide positive performance in the last three years.

It is therefore not so much the factors that are to blame in the performance but the construction choices of the multi-factor indices or portfolios offered by the providers. Among these design choices, some are constrained by the long-only regulatory framework of many funds or institutional investors. It is often difficult to set up pure factor strategies due to the inability to implement long/short strategies and the construction of long-only indices through the use of long/short overlays is difficult. Other choices correspond more to a lack of consideration of non-factor risks in the design of the index, where it involves documenting this risk in order to then manage it properly. In the next section, we will show that it is the lack of integration of the main non-factor risk that is the cause of the disappointment with factor strategies. 

Integrating the contribution of non-factor elements into the performance of single and multi-factor indices

The vast majority of long-only factor strategies were rarely neutral from a market exposure viewpoint; market betas are generally defensive and unstable. For the same construction of long-only indices where, if instead of taking market-neutral indices we take long/short dollar-neutral indices, it is clear that over the last three years, these indices without market-beta control have considerably underperformed their equivalents with a market-neutral market beta long/short overlay. For the US, the 0.51 per cent outperformance over the past three years flips to an underperformance of -1.4 per cent, and to
-1.02 per cent from 0.55 per cent for the developed ex-US. For both regions, the long-only multi-factor assembly does not allow the broad cap-weighted index to be outperformed, as was the case previously.

It is therefore indeed the uncontrolled market conditionality of factors that, in a context of strong bull markets this year, led to disappointing performance, and not the choice of factors or the traditional proxies that represent them. Comparing the conditional performances of dollar-neutral long/short factors over 15 years illustrates this point well. It is easy to observe that in a context of strong bull markets, particularly in the US, the poor conditionality of dollar-neutral long/short was highly penalising compared to the market-neutral version.

In addition to design questions, performance is a matter of fiduciary choice

The analysis that we have conducted on the performance of factor and multi-factor portfolios has allowed us to observe that even though negative performance has been observed for some factors in recent years, it has been possible to offset this through the good performance of other factors. In the strict sense, the factor contribution to the performance of the strategies is not overall negative.

The source of the poor performance therefore needs to be sought elsewhere. We observe that only multi-factor indices which benefitted from a risk-control option that guaranteed alignment of the market beta with that of the reference cap-weighted index were able to significantly improve relative returns. Naturally, taking this market variation risk into account is a fiduciary decision that falls outside of the remit of an index provider, as long of course as the index provider offers these options, which is rarely the case.

Daniel Aguet, head of indices, Scientific Beta; Noël Amenc, chief executive, Scientific Beta and Associate Dean for Business Development, EDHEC Business School; and Felix Goltz, research director, Scientific Beta

 

Comments
    Avatar
    Scientific Beta

    Dear anonymous commentator,

    Your message unfortunately illustrates our introductory point on the lack of attention of many commentators to investment matters and the wrong conclusions drawn from partial or one-sided observations.

    Before replying to you on the substance, I would like to reply on the form of your conclusion. Contrary to what you have stated, the point of our article is not to present reality from the angle that suits us, but to return to that reality. If we did not make explicit reference to our multi-factor indices to speak about factor performance, it is simply because we think that contributions to a website such as Top 1000 Funds that is a reference for investors should remain essentially, if not scientific, then at least neutral from a commercial point of view. The research article that served as a reference for our contribution analyses the conditional performances of factors that are commonly defined in the academic literature and the construction method advocated for building long-only proxies for these multi-factors is also fairly common. We chose to communicate on these proxies to illustrate a simple point, which was that it was the conditionality of the factors, and not so much the factors themselves, that was the cause of performance that we qualified as disappointing over the last three years.

    As you have been able to observe, since you refer to a document published on our website and distributed widely to the press, we always communicate in a detailed manner on the performance of our indices and we have nothing to hide. Accusing us of lacking transparency while referring to one of our published documents to do so is, to say the least, fairly inconsistent in terms of the form.

    On the substance, since our article relates to the poor performance of factor strategies over three years, we would have liked there to have been at least a reference to what we wrote, and if there were comparisons to be made, perhaps use this same three-year period.

    As you know, measurement of returns is unfortunately highly sample-dependent and choosing very short periods of six months only amplifies the phenomenon. This poor statistical representation should at least lead commentators to avoid drawing overly hasty and definitive conclusions on performance observed over short periods, especially when the justification for the usefulness of the strategies involved is that they collect factor premia, which are long-term premia.

    As far as the USD48bn (as of June 30) invested in our indices is concerned, it is essentially in multi-factor indices. As we show in the abovementioned study, this factor diversification provided positive average performance over three years for the six factors in which our long-only multi-factor indices are invested. Nonetheless, and we think it is an advantage of our construction method compared to many other competing methods that do not allow this risk to be controlled, we observe a strong difference in performance over the last three years between our multi-factor indices that benefit from market-beta neutrality in long-only (market beta 1) and those that do not benefit from it. Scientific Beta offers this fiduciary choice to investors who can thereby choose the market conditionality of their factor investment strategies. De facto, in the context of a strong bull market and in our clients’ main investment regions (US, Developed ex-US and Eurozone), the Scientific Beta indices that benefit from the market beta adjustment (CAPM Beta 1) have performed positively compared to the cap-weighted index over the past three years.

    Moreover, since you seem to be concerned about our indices’ investability conditions, the same document to which you refer provides the live performance of our indices in the main investment regions (Table 2a, page 6) and it is easy to observe therefore that in real investment conditions, whatever the variants of our indices and the risk options chosen, these indices have also largely outperformed the reference cap-weighted indices. Like you, we like facts, but it is necessary to refer to them when one wants to speak of “real investment” in our indices.

    Finally, concerning the Sharpe ratio, we do not believe that Sharpe ratio objectives of 1 or greater than 1 are realistic in the long-only space. What we know is that over the long-term, as expressed in the document to which you refer (Table 2c, page 9), the Sharpe ratio of our multi-factor strategies has been 57% greater than that of the reference cap-weighted indices.

    Noël Amenc, CEO, Scientific Beta

    Avatar
    John Peterson

    The most ‘remotely serious investigation’ for investors – and the only one of any relevance – is how do the real investments deliver in terms of returns and risks.

    A simple look at the performance tables for Scientific Beta’s indexes (on which USD 43 Billion of assets are replicated – i.e. managed) is illuminating.

    Scientific Beta’s Smart Beta Index Performance Report to June 2019 can be found at:

    https://ml-eu.globenewswire.com/Resource/Download/4701aef5-043b-42fa-b33e-9b1fb8d27191

    In short, every single strategy (Index) in every single market, underperformed the Cap Weighted indexes in the year to June 2019.

    Every strategy also has Sharpe ratios over longer periods of less than 1, which suggests that they are not a particularly effective use of the risk budget.

    The question is not whether Factors / Smart Beta’s / Risk Premia / etc., delivers returns – yes of course they do

    The question is whether they deliver them in a manner useful for investors, and consistent with the way that they have been included into portfolios (i.e. sold to Trustees).

    The simple reality is that Risk Premia strategies do not meet that ‘investability threshold’ in general, and in recent years have failed to delivered the easily and consistently available return premia that was sold as being just lying around waiting to be ‘harvested’ by all these ‘smart’ strategies.

    In reality markets do not give away free lunches, and the purveyors of the fallacy that they do need to adjust to that reality, rather than trying to bend reality to fit with their preferences.

    John

Join the discussion