Noël Amenc, finance professor at EDHEC Risk Institute, and two colleagues have written about the robustness of smart beta.
As readers of AllAboutAlpha of course know, smart beta is the wave of … the present. Lars Jaeger wrote of its promise here as early as January 2009. His reasoning was that the tumults of the year that had just ended had forced investors to reconsider certain once-cherished beliefs about active investment, and that smart beta would benefit from that reconsideration. Not to put too fine a point on it: he was right.
That same year, EDHEC partnered with FTSE on the implementation of Efficient Maximum Sharpe Ratio indices, a big step forward for the smart beta world.
In the new article Amenc, along with Ashish Lodh and Sivagaminathan Sivasubramanian, begin with the hypothesis that “systematic equity investment strategies… outperform cap-weighted benchmarks over the long run. “ But since a lot of the evidence for that hypothesis is founded upon back-testing historical returns, these authors see it as raising rather than resolving the issue of how robust such strategies are, how capable the outperformance is of surviving changed circumstances, and where it isn’t robust, how it can be made so.
The three authors look at robustness in both absolute and relative terms. Absolute robustness, the Holy Grail here, is the “absence of pronounced state and/or time dependencies.”
Causes of Lack of Robustness
There are four different risks inherent in strategy construction that might render smart beta less than robust, in either sense. Or maybe three. Actually, Amenc et al say “four” at first in listing the risks in a way that leaves one uncertain of the number. Their actual list is as follows: “factor fishing and model mining, [unrewarded strategy-] specific risks, and strong factor dependencies.” After providing that, the authors proceed to treat “factor fishing and model mining” as one possibility rather than two, so the list has three items.
The first two of those items influence in particular relative robustness. It is the third, factor dependency, which strikes “at the heart of the issue of absolute robustness.”
According to one scholarly count, there are no fewer than 314 factors that might be used in selecting a portfolio, each with a positive historical risk premium. A strategy construction could be produced through more-or-less arbitrary fishing in this large pond. To avoid that, these authors suggest that investors should only accept factors “as relevant in their investment process [if]there is a clear economic intuition as to why the exposure to this factor constitutes a systematic risk.”
Under the category of unrewarded risks as a meta-risk in strategy selection, these authors include idiosyncratic stock risks (risks specific to a particular issuer), as well as commodity, currency, and sector risks without a positive risk premium. For example, some strategies tilt toward financial sector stocks; and a tilt in that direction would have been punished quite severely during the crisis of 2008.
Parenthetically, I’d like to observe here that not long ago there was a fair amount of fuss and feathers in academic circles over the supposedly negative link between idiosyncratic risk and return. It appeared from certain statistics that investors didn’t have to be bribed to accept this risk, but counter-intuitively were paying for the opportunity to embrace it. Fortunately for modern portfolio theory, those statistics looked a lot less impressive when broken down, and the fuss over them has died away.
Factor Dependency
But back to the robustness of smart beta. Now our authors return to the core issue of absolute robustness as they see it, individual factor exposures.
Some portfolio strategies are fundamental weighted, which gives them a value tilt (which in turn would favor financial stocks, as mentioned above). Other strategies look to minimize vol and thus have a low beta tilt.
The conclusion to be drawn sounds simple, given the somewhat complex reasoning by which it had been derived. But the point is that diversification as a practice has to embrace strategic factors. Investing “in a single factor is not a robust approach in absolute terms, as the performance will vary greatly over time across different time periods.”
Lodh and Sivasubramanian are, respectively, senior quantitative analyst and quantitative analyst at ERI Scientific Beta. Amenc is its CEO.