Yesterday I attended a presentation by EDHEC on their Efficient Equity Indices and Benchmarks. Efficient Indices (EI) are a family of equity indices EDHEC created that are alternatives to market capital weighted indices (MCI).
One of the most well-known family of non-MCI is Research Affiliates Fundamental Indexes (RAFI), which bases its asset weights on fundamental measures of company size instead of on market capitalization. In recent years other providers have emerged, including MSCI Barra with its MSCI Global Minimum Volatility Indices. (Note that Russell’s new Stability indices remain cap-weighted.)
Still holding all the market constituents
The short list of distinctions among MSCI’s, RAFI’s, and EDHEC’s construction approaches is as follows.
1) MSCI’s method seeks to minimize volatility
2) RAFI’s construction method is strictly an alternative to market cap weighting, without regard to risk or return
3) EDHEC’s method seeks to offer a comparable alternative to the market cap portfolio – lower risk, higher return, but still holding all the universe constituents
How EDHEC models risk and return
EDHEC uses principal components analysis (PCA) to find a sufficient number of risk factors to create a tractable covariance matrix. As a proxy for expected return – which EDHEC takes pains to point out is not the same as forecast return – it clusters stocks according to their semi-variance, using two years of weekly return data. With risk and return estimates in hand, EDHEC simply uses mean-variance optimization to find the tangency portfolio – the portfolio with the highest return per unit risk, or maximum Sharpe-ratio (MSR) portfolio.
Over the decades I have spent a good deal of time examining and comparing different measures of risk, and have long been unconvinced that downside risk measures such as semi-variance and semi-deviation are superior to symmetric ones for the purpose of forecasting risk on individual stocks. This skepticism regarding downside risk as a risk modeling input initially clouded my appreciation for the EDHEC approach when I first read about it in 2010.
Yesterday I realized my skepticism was misplaced, because EDHEC uses downside risk not for the purpose of modeling risk but instead as a proxy for “expected return.”
Think about that. Stocks that have fallen more and more frequently than others over the past two years have higher expected returns. It’s simple, clever, and I would think intuitively appealing to most agnostics. When I think of what might scare your typical investor, I think of negative recent returns.
Results
I won’t recreate the presentation or rehash the performance. You can read about the Efficient Index performance yourself on the EDHEC site, where you can find supporting research and other information. Not surprisingly, the indices load positively on Value and negatively on Size, and their backtested performance results show higher returns and lower volatility in the long run than their MCI counterparts. Very little of the performance attribution is due to sector weighting.
I wonder how much of the advantage is due to short-term reversal. Further I would be interested to know how simply applying the EDHEC PCA covariance matrix approach would work with other estimates of return, even with random numbers (ie, does the downside risk proxy for expected return even matter?)
Final Thoughts
If you’re going to adopt an index method other than cap weighting or equal weighting, you have to make some choices. EDHEC seems to have thought this through quite well. Its choices to include all assets and seek the MSR portfolio should make EI appealing to many investors who dislike MCI.
EDHEC is seeking partners to sponsor enhanced strategy portfolios, possibly for ETFs. The enhanced strategies would blend the EI method with MCI weights. It seems inevitable that one or more of the usual suspects will wind up partnering to offer such products.
Update, 6/23, 3:40p: Vijay Vaidyanathan, who presented for EDHEC, clarified a subtle point I had not mentioned about EDHEC basing its return estimate on semi-deviation: “SemiDeviation serves the vitally important role of serving to, in effect, penalize the low vol stocks so that you don’t end up with the severe concentration problem that you would have if you assumed a flat relationship (as in MinVol).”
