Most factor-based, otherwise known as Smart Beta, ETF strategies are based on a single concept like value or momentum. Over the last two years, the largest flows have been to ETFs investing in low volatility stocks. The most popular being the iShares Edge MSCI Min Vol USA ETF (USMV), which as of September 30th had grown to $14.4bn USD, more than doubling over the last 12 months.
With product proliferation, there are now a number of low volatility ETFs, each with a different portfolio construction methodology as well as their own method on the best way to select stocks with low volatility. The most straight-forward method is to look at the raw volatility of the trailing returns, either on a daily or monthly basis, for anywhere from three months to one year. Some other strategies use the Beta of the stock: the covariance of the stock with market returns, scaled by the volatility of the market. In 2014, Frazzini and Pedersen published a “Betting Against Beta” (BaB) strategy, that goes long low-beta and short high-beta, leveraging and deleveraging each side to a beta of 1. Another option is tracking error, the volatility of the excess returns stock versus the market. Ang, Hodrick, Xing and Zhang’s 2006 paper “The Cross-Section of Volatility and Expected Returns” introduced the idea of idiosyncratic volatility: the excess volatility of the stock after a regression on the Fama-French factors. A last way to measure volatility would be through implied volatility, which uses options pricing to derive the expected future volatility.
We could spend a lot of time arguing the merits of each metric, but in practice the results from investing in volatility factors look very similar. They exhibit the same return profile: a portfolio of stocks with high volatility in the past gives high volatility in the future, along with strong underperformance. Stocks with low volatility continue to have low volatility. Coupled with modest outperformance, the risk-adjusted Sharpe Ratios look strong. For comparison, the following charts show the excess return, the excess volatility, and the sharpe ratios of portfolios based on Value and Volatility characteristics within the Large Stocks universe, stocks with a market cap greater than average. Stocks are selected monthly with a holding period of one year as part of the decile portfolio. While Value has stronger overall return, volatility is less volatile, giving similar Sharpe Ratios between Value and Volatility. Volatility factors also correlate very highly with one another, much higher than Value characteristics. This would indicate that Value factors capture different expressions of valuation, where all the volatility metrics seem to be based on the same market phenomenon.
A recent McKinsey report cites research that high-net worth investors top focuses are protecting principal, hedging against downside risks, minimizing volatility and generating income. After the market crash of 2008-2009, it’s easy to see how advisors and plan sponsors could be drawn to “Defensive Equity” or “Low Risk” strategies as ways to protect against future drawdowns. From the point of an advisor, low volatility strategies ETFs cover three of these, offering down-side protection with equity-like returns.
The risk of low volatility strategies is its usage within the total allocation of a portfolio. For the asset management industry, the value chain for clients is to hire advisors to establish an asset allocation for them. Once an allocation is decided upon, a manager selection process determines the best people to manage assets within those groups. Passive investing has supplanted active mandates within these allocation buckets. For example, if a consultant believes active managers have no edge in large cap stocks, just buy the Vanguard S&P 500 ETF for five basis points. Small cap stocks are less efficient and active managers have edge there, so use individual mandates in that space.
It is unclear how Smart Beta strategies fit into this. If it is viewed as a separate asset class, it is invested in based on the total expected return, volatility and diversification it adds to the total portfolio. If it is viewed as part of an equity allocation, it is judged on the excess return versus a passive benchmark, scaled by the excess volatility. In the case of a passive benchmark and an active manager, these roles are clear. Smart Beta makes this more confusing. Is it a passive allocation to an asset class, or is it a cheap source of alpha?
Volatility factors might deliver solid risk-adjusted returns for an allocator, but they are lacking in the realm of active management. Low volatility has had modestly higher performance with a lower raw volatility, but it also came with higher excess volatility. Using the same basic portfolios formed on the deciles of each factor in Large Stocks as above, the tracking error of volatility factors shows higher excess volatility than Value factors. The tracking error of the top decile of raw volatility is 9.7% versus the equally weighted universe, versus tracking errors in the 6.5%-7.8% range for value factors. With lower excess returns, Volatility factors have an information ratio about half to one quarter of that of Value and Yield.
The higher tracking error can be managed down in portfolio construction: equally weighted versus market-cap weighted, sector agnostic versus sector relative. But this still leaves a large amount of excess volatility in the portfolio. The MSCI Min Volatility USA Index, which the iShares Edge ETF is based on, is a good example. In the index construction methodology, there are several risk factor and sector constraints, but it still leaves the MSCI Min Volatility USA Index with a tracking error of 5.73% to the broader MSCI USA benchmark since it was incepted in 1988. Looking at the history of this index through the lens of active management, this gives it an Information Ratio of only 0.25.
Tracking Error and Information Ratios seem a bit clinical compared to the real-time experience the investor has. A better way to show the effect of this is the rolling 1-year excess return of the strategy versus the broader market benchmark. This tracking error difference leads to multiple periods of time where there is strong relative underperformance. In over 20% of the rolling one-year observations, the MSCI Min Vol USA Index is trailing the MSCI USA Index by over -5%. Several periods of over -10% underperformance, and one time reaching -15%.
The framing of how the investment is perceived matters to the advisor and to the discipline they will have in maintaining an investment in it. Allocations tend to be more strategic, and not subject to the relative performance of one asset class versus another. Bonds are supposed to behave differently than equities, which is why you own both. Investments within the allocation tend to be questioned on a more regular basis.
It is impossible to determine how every person is using low volatility ETFs, but asset flows should give some insight. If the flows were not reacting to near term performance on a relative basis, then it is being used in a strategic allocation. But the flows to the Volatility ETFs appear to be based a recent spike in the relative performance, and specifically on near-term performance. The orange line shows the trailing 12-month performance of the USMV ETF relative to the MSCI USA Index. Low volatility stocks have been outperforming the average stock since the beginning of 2015, with peak outperformance coming around the second quarter of 2016. This coincided with very strong flows into the product, where by the end of the second quarter there had been almost $8 billion invested over the trailing 12 months with a coincident trailing 12-month outperformance of USMV over the MSCI USA benchmark of +13%.
To his credit, Andrew Ang, who heads the Factor-Investing group at Blackrock which runs USMV, is trying to educate investors about how best to utilize low volatility investing. He wrote in a September 2016 Forbes article “Investors’ aim with low-volatility strategies shouldn’t be to outperform the market, rather to reduce risk and to measure that performance over a full market cycle.” But it seems to fall on deaf ears, as assets chase one-year relative returns.
As investors flock to the low volatility ETF based on near-term outperformance, they also sell after near-term underperformance. These types of fund flow reactions to recent performance will only increase the difference in the time-weighted and money-weighted returns of the fund. As of September 30th, the USMV had returned 14.39% since inception in October 2011, only +22 basis points over the MSCI USA Index which only returned 14.17%. But because of the flows chasing performance, the average money-weighted return of USMV is only 11.89% over that time frame, creating a gap for investors of 250bps because of return-chasing. This gap looks like it’s only going to increase. In the third quarter of 2016, MSCI USA was up +4.06% while USMV was down -1.16%, a gap of -5.22%. In reaction, USMV saw redemptions of -$877 million in the month of October.
Some of this gap is going to be created by individual investors, operating without an advisor or a structured asset allocation plan. Some of this will also be factor timers, who are shifting allocations based on when they believe a factor will generate outperformance. But some of the investors in low volatility ETFs are advisors trying to figure out how to use them in a long-term structured allocation. In my opinion, a long investment in low volatility portfolios can have a place in a portfolio if you’re thinking about it as an allocation. If you’re going to judge an investment in low volatility as a cheap active equity investment, there are better factors such as value and momentum that offer the opportunity for greater excess return given the active risk taken.
As a last thought, the shift towards factor-based “Smart Beta” ETFs makes me believe more than ever that advisors need to learn all they can about how factor investing works. Manager selection involves a long process of getting to know the people and style of their investment. Investing in an ETF seems easier, but it also comes with reduced switching costs. And while index construction methodology is published as transparent, there could be less understanding of the process given the lack of interaction with the investment manager and the complexity of some of the strategies. Lower switching costs and lower understanding of factor investing leads to less investment discipline and a continued degradation of investor returns on a money-weighted basis.
-Thanks to Ehren Stanhope and Patrick O’Shaughnessy for the feedback. Appreciate the help guys.