There’s a large body of research demonstrating that, while past returns do not predict future returns, past volatility largely predicts future near-term volatility.
For example, since 1926, the correlation of returns between adjacent 60-trading-day periods is essentially zero. In contrast, the correlation rises to 0.55 for risk as measured by standard deviation. Evidence that past volatility predicts future volatility has been found not only in stocks, but in bonds, commodities and currencies.
Based on evidence that shows they improve portfolio efficiency, managed volatility strategies have been developed to adjust exposures in inverse relation to a risk estimate.
The aim of managed volatility strategies is to stabilize realized portfolio volatility through time and produce a more efficient portfolio with less downside risk.
Funds such as AQR’s Style Premia Alternative Fund (QSPRX), AQR’s Managed Futures Strategy Fund (AQMRX) and the Stone Ridge All Asset Variance Risk Premium Fund (AVRPX) have been successfully managing portfolios to a targeted volatility for a number of years now—increasing leverage and exposure when volatility is falling and decreasing them when volatility is rising. (Full disclosure: My firm, Buckingham Strategic Wealth, recommends AQR and Stone Ridge funds in constructing client portfolios.)
Using Volatility To Mitigate Risk
Anna Dreyer and Stefan Hubrich contribute to the literature on managed volatility strategies through their November 2017 paper, “Tail Risk Mitigation with Managed Volatility Strategies.”
Their study covered the period 1926 through 2016. The authors found managed volatility (MV) strategies that dynamically alter their daily equity market exposure according to the ratio of realized past long-term equity volatility to trailing 20-day realized volatility (the former represents the volatility target; the latter represents the forecast) have “consistently hugged the target volatility within a much tighter range than the benchmark itself.”
Dreyer and Stefan Hubrich write: “This observation holds even during turbulent market episodes like the mid-1970s, the 1987 crash, the tech bubble/crash in the early 2000s, and the Global Financial Crisis (GFC) in 2008-09.” More specifically, they found 90% of their realized volatility observations for MV fell into the 0.68% to 1.59% interval, while the range for the benchmark was much wider, at 0.41% to 2.04%.
Reduced volatility and less downside risk are clearly important both to risk-averse investors and to all investors who are in the withdrawal stage, when the sequence of returns can matter greatly. Reduced downside risk can also help investors prone to panicked selling, which generally increases in severe bear markets.
While MV strategies are related to time-series momentum, they don’t perform similarly in all environments. However, realized volatility and returns tend to be negatively correlated.
For example, Dreyer and Hubrich found that the correlation for 60-day windows in their sample is -0.25—volatile periods tend to be associated with “bad news” and market sell-offs. In other words, volatility tends to spike up, not down.
Note that if market returns are positive, the use of leverage allows MV strategies to capture enhanced returns when volatility is low. By reducing leverage when volatility is high (and when returns tend to be poor), MV strategies tend to avoid the worst of bear markets.
The research shows that Sharpe ratios (SRs) are lower (higher) when volatility is higher (lower). This enabled Dreyer and Hubrich to conclude that MV strategies are not only related to time-series momentum, but also to returns’ long-term mean reversion.
Sharpe Ratio Hit Rates
Dreyer and Hubrich did note that while SRs were improved in all three of the periods they examined (the full period from 1960 through 2016 and two subperiods, 1960 through 1989 and 1990 through 2016), they found what they called the “hit rate” (persistence of a higher SR over 10-year rolling periods) was not high.
Specifically, the authors found that, starting in 1929, “which has the greatest overall SR improvement, these hit rates are barely above 50%. The hit rate is higher in the later periods, but never much above 60%.” They attributed “any MV SR improvement to perhaps 1-2 crisis periods in the particular sample. Since these crisis periods do not occur often, investors can often see a decade without a SR improvement from MV.”
In other words, as with value investing or any other investment strategy, MV strategies often require great patience and discipline to reap their benefits. The authors warn: “In practice, investors pursuing pure MV strategies for SR enhancement should consider the potential for regret risk even with a long-term commitment to the strategy.”
Skewness & Kurtosis
Dreyer and Hubrich then turned their attention to the higher moments of the distribution of returns: skewness and kurtosis. Skewness measures the asymmetry of the return distribution. Zero skewness indicates a symmetric distribution, while positive skewness indicates a distribution with a greater likelihood of large positive returns than large negative ones. Investors naturally prefer positive skewness.
Kurtosis measures the fatness of the distribution’s tails. A normal distribution has a kurtosis of three, and a higher degree of kurtosis indicates more extreme returns than under a normal distribution of the same volatility. In other words, relative to a normal distribution, the tails contribute more to volatility than the center of the distribution. The authors found that MV strategies:
- Had lower values for kurtosis and more positive skewness for nearly all expanding-window periods since 1929.
- Consistently thin the tails at 20-day and 60-day holding periods. At these holding periods, the benchmark features meaningful excess kurtosis, while MV essentially produces normal tails with kurtosis close to three.
These results led Dreyer and Hubrich to conclude that “MV strategies have historically provided more return per unit of tail risk than the benchmark.” They noted this outcome is likely due to the tendency for volatility to cluster—realized volatility can remain conditionally higher or lower than the average for extended periods.
Dreyer and Hubrich next turned to examining drawdowns. They write: “MV improves the relevant ratios by 35-50% over the full sample, 15-20% over the sample starting in 1990, and essentially scores on par with the benchmark for the sample starting in 1960.”
Lower drawdowns also led to higher compound returns (recall that Sharpe ratios are based on annual average returns). Given the high level of most investors’ risk aversion, these improvements can offer investors significant benefits in terms of utility.
When viewed through the lens of the Sharpe ratio, Dreyer and Hubrich found that, while MV enhances outcomes, on average, over the buy-and-hold benchmark, it does so with great variation. In addition, enhancements are modest and period-dependent. However, they do alter the higher moments of the return distribution in a favorable way, enhancing skewness and removing fat tails. Thus, they improve utility for risk-averse investors.
They also observed that MV strategies are characterized by material turnover and trading activity due to their dynamic adjustment of exposures. The result is that MV strategies are likely to be tax inefficient, so investors should strongly prefer to hold them in tax-advantaged accounts.
Dreyer and Hubrich showed that MV strategies reduce tail risk. As my co-author Kevin Grogan and I showed in our 2014 book, “Reducing the Risk of Black Swans,” there is another way to reduce tail risk: Build portfolios with low exposure to market beta and high exposure to other factors that have delivered premiums (such as value and size, but also momentum and profitability/quality).
You need less exposure to market beta risk because the equities you hold have higher expected returns. That allows you to hold more safe bonds, which tend to perform their best in times of stress in the equity markets.
Portfolios designed in this manner are often referred to as risk parity strategies because they minimize the concentration of risk in the single factor of market beta that is found in typical 60/40 portfolios.
We are working on an updated version of the book, which will include a new section on how newly available alternatives—such as reinsurance, alternative lending and the risk variance premium—can also be used to both further reduce tail risk and improve portfolio efficiency. The book should be available in the first part of 2018.
Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.