Risk & Portfolio Construction: Revenge of the quants
The ‘risk-factor’ revolution offers the tantalising prospect of getting exposures for which investors used to pay alpha fees as bargain beta. But Brendan Maton notes that it raises as many questions as answers when it comes to investable product
Call it the revenge of the quants. After years out of the limelight quantitative managers are back with a bang. What they are offering today is a reconstructed bundle of risk premiums promising improved returns with lower total risk. The promise might be hoary and the sources of those premiums – momentum, size, quality, and the like – have been discussed academically for decades. What is new are real products specifically designed to access these risks.
Let us consider some of the risk premiums themselves, which can be thought of as the ‘nutrients’ within the asset ‘food’. Or, as the academics Andrew Ang, William Goetzmann and Stephen Schaefer wrote in a seminal paper evaluating the active management of Norway’s Government Pension Fund Global in 2009: “Assets are essentially conduits for factor risk premiums, like insurance.”
The first would be size, which is broadly the propensity of smaller stocks to outperform larger stocks over the long term. Two good reasons are that small caps are less liquid and less likely to survive harsh conditions, both of which merit some premium.
The second would be value, the higher long-term return for stocks with low fundamental valuations metrics like the ratio of price to book value. One explanation for this premium is that markets erroneously perceive low-value stocks as risky, an error that corrects itself over longer periods.
Momentum can be complementary to value because it reflects markets’ tendencies to bid up stocks that have fared well in the immediate past, especially in bullish periods (or dump those that have fared badly). Although bull and bear trends are intuitive, the momentum premium was one of the last to be documented by academics, in the early 1990s.
A fourth popular premium is low-volatility, which may be explained by investors’ propensity to confuse high-risk assets with high-return and subsequently underweight low-volatility stocks to their own disadvantage. Some practitioners believe that the excess return to low-volatility is not a premium but, in fact, an ‘anomaly’, something that ought not to persist. As we shall see, the whole subject of factor risk premiums, also known as alternative betas, is fraught with such ambiguity. There is, for example, no universally-agreed number of premiums, nor is there a single means of capturing them.
“Theory is silent on what the set of factors is, but statistical studies suggest a limited number, usually fewer than 10, are enough to capture most variation in expected returns,” say Ang, Goetzmann and Schaefer.
This lack of a common tradition might explain why pension funds and insurers have not been offered such alternative betas before. And while there almost certainly are other risk premiums, the quartet above provides a good beginning – not least because they apply to equity and in traditional portfolios equity is where most risk finds itself.
For a dedicated small-cap or value manager, however, the idea of their stated bias being rechristened as a factor-risk premium or alternative beta might seem unnecessary or even predatory. The world of institutional investment, especially the US, has recognised style and size investing for decades, and there are appropriate benchmarks to boot. How has this turned into anything particularly new?
The answer would be that existing style indices nod towards the factor risk premiums, notably in their names. But if investors believe that extant style investing has offered much diversification, they should think again – the labels are misleading. These size and style indices are strongly correlated to the market-cap-weighted index. As such, they have not supplied much ‘ballast’ during the worst of market storms. One could broaden the argument to all active managers benchmarked to the market index: without a new term of reference, it is unlikely that genuine diversification is possible.
This was the most startling point made by Ang, Goetzmann and Schaefer – that active management had made but a negligible contribution to the track record of Europe’s largest institutional investor. They found that the fund was more similar to an index fund than an actively-managed fund, and that “a significant part” of even the very small component of the total return represented by active return was linked to “a number of well-recognised systematic factors”. The contribution of active management to the overall return that is genuinely idiosyncratic was “extremely small indeed”, they concluded.
Here then is the academic endorsement of a change in investing – not mere baby steps to style-tilted market cap indices, but a thorough reconstruction of the portfolio.
To give one example of the magnitude of difference between traditional small-cap investing and exploiting the size-risk premium, here is a simple, generic trading idea: divide the broad equity market into quintiles. Then short the fifth quintile (the largest stocks) while going long the second quintile. This idea is from a study of long-term risk premiums from the X-asset team at SEB in Stockholm. The team estimates real returns of 2.3% after trading costs and with no leverage for this strategy, backtested over 85 years. If it sounds like a hedge fund, that might be no surprise: SEB’s implementation team, which is developing a product for pension funds from the research, has been conducting hedge-fund analysis for a decade. Part of their job has been to identify factor exposure in hedge funds. What is new here is the clear attempt to isolate and label the risk premiums as such.
“Many clients are tired of paying 2-and-20 [hedge fund fees],” says SEB fund manager Andreas Johansson. “We are not afraid of cannibalising our hedge fund business. Our aim has always been not to pay hedge funds for systematic factor exposures that can be accessed cheaper elsewhere.”
He is quick to add that clearly labelled factor-risk premiums will not sound the death-knell for all hedge funds that exploit these factors. To a large extent it’s the quality of interpretation of the data that matters. Some alternative investment managers, such as AQR, are actually evangelical about the importance of factor risk premiums in their process and the clarity they bring to portfolio construction.
Like SEB, AQR aims to avoid market direction by going long and short. “Investors get market exposure from going long traditional asset classes, including less-liquid asset classes such as real estate and private equity,” explains Ronen Israel, a principal and portfolio manager at AQR. “Our styles focus on liquid markets across assets, including currencies and fixed income. But the strategies are uncorrelated, so there is no tie-in with market beta.”
Israel, AQR colleague Antti Ilmanen and University of Chicago professor Tobias Moskowitz have co-authored a paper suggesting that 30% exposure to a blend of ‘styles’ – in other words, factor-risk premiums – would improve the Sharpe Ratio of a traditional portfolio of equities and bonds from 0.30 to 0.74.
At this point, however, readers might be asking themselves a couple of questions. Are factor risk premiums merely the new hedge funds? And is there any more to them than mere backtests – where are the real products?
Quant folk instinctively love backtests but commercially they have little influence: when conducting due diligence on potential providers, few pension funds and insurers will entertain a product without a real track record. Quite right, too: as Dimitris Melas, head of research at MSCI, wryly points out: “I’ve never seen a bad backtest.”
Let’s take the first point. Market neutrality is not essential for factor-based investing; most products labelled as such do not short-sell. But it can be a more effective way of implementing these ideas. In a forthcoming paper in the Journal of Financial Economics, Israel and Moskowitz show that when it comes to classic styles like momentum, profits from going long and short are basically equal. This supports the notion that there are benefits to these styles in a long-only context but double the benefits in a market-neutral version.
Looking back at the four risk premiums, the one frequently associated with shorting is momentum, because of the theory that profits can be made from the dogs as much as the winners. But houses such as Robeco now offer even momentum as a long-only product.
Regarding back-testing, Pim van Vliet, a senior portfolio manager in quantitative equities at Robeco, agrees that observing factor-risk premiums in hindsight is insufficient. He emphasises that managers have to make forward-looking estimates to optimise portfolios and reduce implementation costs – but optimisation in the alternative-beta world is problematic for the reason identified by Ang, Goetzmann and Schaefer: there is no agreed best way to implement these ideas.
This applies to the index providers as much as the active managers involved in this arena – so that even the question of finding a suitable benchmark for the managers one might mandate is fraught with difficulty.
As David Blitz, head of quantitative equities research at Robeco points out, there are considerable differences between the providers themselves. For example, Blitz notes that in its Low Volatility index, S&P does not have sector constraints. This currently leads to a big overweighting in telecoms and utilities. MSCI, by contrast, constrains sector deviations to five percentage points under or over their weight in the MSCI market cap-weighted index. Such variety between would-be reference indices leads Blitz and Van Vliet to describe the indices themselves as active strategies.
This might explain why many providers have taken it upon themselves to license indices based on their own methodology alongside running conventional mutual funds and mandates. Some, such as Research Affiliates, have made a fortune from so doing. But none of this aids a broad swathe of investors in settling on a definitive reference point for factor-risk premiums.
This is the first part of the revenge of the quants: investors have more-or-less to abide with the quants’ chosen approach for exploiting the factor-based premiums.
Which is all very well, but like all optimisations, as Van Vliet indicates, these require inputs about expected risk and return which will inevitably contain errors, and the models will maximise those errors. Counter that by dragging the model back towards market cap-weighted optimisation and you end up offering just another version of ‘enhanced indexation’ – a neither-fish-nor-fowl fad of the last decade which has not lasted.
The new wave of quant managers running factor-based investing models are keen to say that this time is different. One means by which they are making it different is by co-mingling factors. Thus, while Robeco is associated with low volatility, Van Vliet points out that its Conservative Equity strategy has exposure to value and momentum as well.
“When we first launched in 2005, the seed investor’s main worry was the likelihood of huge drawdowns from volatility alone,” he explains. The nature of premiums to disappear for long periods – as the size premium did in the 1990s or momentum did this century – is well-documented. No one involved in this sector denies these tendencies.
“Some factors, such as momentum, are pro-cyclical, while others, such as low-volatility, are anti-cyclical,” explains Melas. Hence the common practice of blending premiums to a greater or lesser extent to achieve more reliable returns.
This is the ultimate revenge of the quants: the role of allocation. While any presentation on these betas begins with a distinct list of them, in practice another factor, or risk gets introduced in their combinination – manager skill.
So how does all of this fit into a traditional portfolio of stocks and bonds?
“We’re still working on that,” says Tom Goodwin, director of research at Russell Indexes. Thomas Thygesen, head of X-asset strategies at SEB, finds that it has been much more a case of tentative evolution rather than outright revolution: “Every pension fund is moving in the same direction. They are looking at this first in the context of equities and adding some factor exposure there.”
Indeed, following Ang, Goetzmann and Schaefer’s report, the Norwegian Government Pension Fund Global itself now has introduced the size and value premiums into its operational reference portfolio for equities (while, it has to be said, shying away from many of the other factors that have been identified).
The important lesson for other funds looking to Norway is that raising or lower exposure to such factors is a risk-attribution and risk-management exercise for the Global fund: it serves to clarify where risks and returns emanate in changing markets. Investors naïvely buying alternative beta products statically, without reference to broader macroeconomic conditions, could well form the next generation of disgruntled clients of quant managers.