Low Volatility is Not Low Risk

September 20, 2016

14 minute read

This blog post is available as a PDF here.

Low volatility equity ETFs have seen huge inflows this year, driven by both a compelling story (risk-managed equity investing) and significant outperformance.
Critics have raised concerns that short-term performance chasers have driven up the valuations of these strategies, increasing the likelihood of a crash back to reality.
These criticisms are not supported by the data. Valuations are not a particularly useful tool for timing factor strategies. High valuations for low volatility equities do not necessarily imply poor future performance.
That being said, all factor-based strategies, including low volatility, go through regular periods of underperformance. Sometimes this underperformance is significant and so can be characterized as a crash.
Low volatility equity investing can be a useful tool for many investors. Yet, low volatility is not the same thing as low risk. Building a diversified, multi-strategy approach to risk-managed equity exposure can help to mitigate the risks inherent in low volatility investing, including the risk of a factor crash.

Low Volatility ETF Explosion

Low volatility ETFs are on fire. These strategies, which aim to offer lower volatility exposure to various portions of the equity market, brought in $13 billion through June of this year.

In our opinion, these flows are evidence of what can happen when the right story (risk-conscious equity market access) lines up with short-term performance.

Source: Data from Yahoo! Finance, Calculations by Newfound Research. Low Volatility (U.S.) is the iShares Edge MSCI Min Vol USA ETF (ticker: USMV). Low Volatility (Foreign Developed) is the iShares Edge MSCI Min Vol EAFE ETF (ticker: EFAV). Low Volatility (Emerging Markets) is the iShares Edge MSCI Min Vol Emerging Markets ETF (ticker: EEMV). Market-Cap (U.S.) is the SPDR S&P 500 ETF (ticker: SPY). Market-Cap (Foreign Developed is the iShares MSCI EAFE ETF (ticker: EFA). Market-Cap (Emerging Markets) is the iShares MSCI Emerging Markets ETF. Past performance does not guarantee future results.

Criticisms of Low Volatility and Other Smart Beta Strategies

Of course, success inevitably brings out the critics. One vocal critic has been Rob Arnott of Research Affiliates, which laid out his view in a February article titled “How Can “Smart Beta” Go Horribly Wrong?” His basic argument is:

Factors can outperform for “structural” or “situational” reasons. Structural outperformance (i.e. premium in value stocks as a result of higher business cycle risk) is sustainable. Situational outperformance – mainly from increasing valuations – is not.

Certain factor premiums – especially low beta/volatility and quality – become significantly less compelling when the impact of increasing valuations is netted out.

Performance chasing pushes valuations higher, creating continued short-term outperformance, but also exposing investors to a “reasonable probability of a smart beta crash.”

Source: Research Affiliates. Underlying data from CRSP/Compustat. Past performance does not guarantee future results.

A few quick notes before we continue. First, like many we hate the smart beta label and prefer to call these strategies factor strategies. However, we will use these terms interchangeably. Second, throughout the article we are going to restrain our discussions to what we view are proper implementations of smart beta / factor strategies. While there are many good strategies on the market, there are others with material flaws or that are simply not backed by reputable research. Third, we will talk quite a bit about low beta strategies in addition to low volatility strategies. These are similar approaches to defensive equity investing, the difference being that one measures risk with market beta and the other with volatility. The need to talk about both low volatility and low beta arises from the fact that many academics have studied low beta, while the largest ETFs use the low volatility flavor.

With this housekeeping out of the way, let’s return to the Research Affiliates article. Our views on their argument are mixed.

We agree that factors can out-/underperform for structural and/or situational reasons and that the structural reasons are where the long-term value-add resides.

We agree that performance chasing can perpetuate short-term returns in certain factors and that this behavior can lead to significant pain when the factor mean reverts.

We even agree that factor/smart-beta crashes are inevitable.

We disagree, however, with the conclusion that certain factors - like low volatility - have only outperformed due to increasing valuations and that using them now is dangerous. Any investment strategy can be dangerous if used improperly for a given client. That is nothing new. When used responsibly and thoughtfully, however, we believe that low volatility and other smart beta strategies can have significant roles in many investor portfolios.

More specifically, our disagreements lie with certain aspects of Research Affiliates’ methodology as well as their framing of the issues.

Issues with Research Affiliates’ Methodology

In terms of methodology, Research Affiliates uses value spreads to assess the impact of valuation changes on factor returns. Value spreads quantify the difference in valuations between the long (i.e. low volatility or low beta stocks) and the short (i.e. high volatility or high beta stocks) sides of a long/short factor strategy. A paper published by AQR, “Are Defensive Stocks Expensive? A Closer Look at Value Spreads,” found a “tenuous link between valuations and returns for style premia.” For example, the paper found that the correlation between value spreads and next month returns for the low beta factor has been very close to zero (-0.01 to be exact)[1].

To understand why, it’s helpful to understand the conditions that would have to be met for value spreads to predict returns.

First, value spreads would need to exhibit mean reversion. This first condition means that relatively high valuations will eventually move downward towards the long-term average and vice versa for relatively low valuations.

Second, changes in valuations would have to be driven mainly by price changes. This second condition means that rising valuations imply rising prices and outperformance while falling valuations imply declining prices and underperformance.

Do Value Spreads Mean Revert?

The first condition, value spread mean reversion, does seem to hold as evidenced by the chart below. The yellow dashed line plots the value spread for the low beta factor. When the line is above (below) zero, low beta stocks are undervalued (overvalued) relative to high beta stocks.

Image from AQR Paper: “Are Defensive Stocks Expensive? A Closer Look at Value Spreads.”

The difficulties in using this knowledge are that over/under-valuations can get more extreme before reverting and that the speed of mean reversion is unpredictable.

Are Valuation Changes Mainly Driven by Price Changes?

The short answer is not necessarily. AQR’s paper on defensive stock valuations goes into quite a bit of detail as to why this is the case. The Cliff’s Notes version is that prices are not the only factor that can drive valuation changes. For example, valuations can change as a result of fundamentals. Earnings growth will cause P/E ratios to fall if prices do not rise accordingly. Valuations in factor portfolios can also change because of portfolio rebalances. Let’s say a low volatility ETF rebalances, increasing its allocation to technology and decreasing its allocation to utilities. Holding all else equal, it would be reasonable to expect this activity to increase the valuation of the portfolio since technology stocks tend to have higher P/E ratios than utilities stocks. Yet, this valuation increase has no immediate impact on strategy returns.

Inconsistent and unpredictable timing of value spread mean reversion combined with a complicated relationship between valuation changes and price changes makes for a weak relationship between low volatility / low beta strategy valuations and returns.

To hammer this point home, AQR discusses an interesting phenomenon for the low beta factor. In 2013 and 2014, the low beta factor got cheaper as low beta stocks became less expensive relative to high beta stocks. Intuitively, we might assume that the factor underperformed, yet the opposite occurred. The low beta factor outperformed over this period despite becoming less expensive[2].

Image from AQR Paper: “Are Defensive Stocks Expensive? A Closer Look at Value Spreads.”

Issues with Research Affiliates’ Framing

Putting the technical critiques aside, we find Research Affiliates’ forecast of a smart beta crash in overvalued factors like low volatility to be a bit misleading. Not because it isn’t true, but rather because factor “crashes” are nothing new. They’ve happened regularly over time. In fact, factor crashes are completely consistent with the theory that many factor premiums exist as a form of compensation for bearing various types of risk. If risk did not exist, there would be nothing to compensate for and premiums may disappear.

Nowhere is this clearer than with the most famous factor strategy: value. Going back to the 1920s, there has been a 1 in 8 chance that the value factor underperforms by 10% or more over a 12-month period. And in some cases the underperformance has been significantly worse. For example, the value factor lost more than 40% from March 1934 to March 1935 and more than 30% between February 1999 and February 2000.

The Good News

Despite our disagreements, we believe that the Research Affiliates piece and others like it are a positive for the continued growth of factor investing. Short-term performance chasing is dangerous. While we like low volatility strategies generally, the reason for using them has to go beyond solid short-term performance.

Sustainable allocations to factor strategies requires investor’s buying into these products with eyes wide open.

When it comes to low volatility investing, we believe it’s important to recognize that volatility and risk are not synonymous. Take cash as an example. It has very low volatility, but can be an extremely risky investment for a young investor who needs growth to fund retirement.

It follows then that low volatility strategies are not necessarily low risk strategies.

At Newfound, we have a simple statement that we like to live by when advising clients: “Risk cannot be destroyed, it can only be transformed.”

With this in mind, we’d like to spend the rest of this article identifying some of the risks of low volatility strategies. Just to be clear, none of what we are about to discuss is meant to discourage use of these strategies. All investment approaches have risk. Our goal is to bring these risks to the forefront so that they can be addressed during the overall portfolio construction process.

Low Volatility X-Ray

One way we can go about trying to identify risk is with a factor model. A factor model attempts to decompose the returns of a security, asset class, or strategy. The simplest, and most famous, factor model is CAPM. CAPM uses a single factor: the market. Another famous factor model is the Fama-French three-factor model, which adds size and value to the equation.

For this exercise, we are going to follow the approach used by Research Affiliates in a 2014 paper titled “A Study of Low-Volatility Portfolio Construction Methods.”

The model uses six factors: the four factors in the Carhart four-factor model (market, size, value, momentum) plus AQR’s “betting against beta” factor and a duration factor that reflects excess returns earned by long-term U.S. Treasuries. [Note: the betting against beta factor reflects the returns from a portfolio that goes long low beta stocks and short high beta stocks].

We used the model to decompose the returns of the market-cap weighed MSCI USA Standard (Large + Mid-Cap) Index and the MSCI USA Minimum Volatility Index.

Source: Data from MSCI and AQR, Calculations by Newfound Research. Past performance does not guarantee future results. Covers the period from June 1988 to May 2016.

Unsurprisingly, the market-cap weighted MSCI USA Index is dominated by market risk. The negative exposure to the size factor results from the fact that the MSCI indices are large/mid-cap and so are biased away from small-cap exposure.

The low volatility index, on the other hand, has significantly less market exposure. Yet, it has higher exposure to the value, size, betting against beta, and duration factors. Low volatility strategies do not have less risk; they just have different risks. Realized volatility has been lower than the market-cap weighted benchmark because these different factors have historically done a good job of diversifying each other.

Risk #1: Interest Rate Exposure

The exposure analysis shows that historically low volatility strategies have positive exposure to the Duration factor. This implies that they, like bonds, have tended to benefit from falling interest rates. The loading – or exposure – of the MSCI USA Minimum Volatility Index to the Duration factor has been approximately 0.09 over the period studied. In simple terms, this means that a 1.00% excess return for 10-year Treasuries has contributed approximately 0.09% towards the return of the low volatility strategy. Likewise, a 1.00% loss for 10-year Treasuries would cause a 0.09% performance drag for low volatility equity strategies.

We think this is important for three reasons.

Rising rates may cause performance drag.
In the event that rising rates cause an equity downturn, there is the risk that the historically negative correlations between the Market factor and the Duration factor could turn positive, reducing potential for volatility reduction.
In the 1988 to 2016 period, the Duration factor returned 5.7% annually as rates fell pretty consistently during this period. With a loading of 0.09, this means that stellar bond returns caused by falling rates contributed about 50bps per year to low volatility returns. Unless rates fall far into negative territory, this tailwind for low volatility returns will likely be limited going forward.

Risk #2: Changing Exposures

The low volatility X-ray presented earlier summarizes a 28-year period. As we mentioned earlier, low volatility strategies tend to have quite a bit of turnover. As a result, the exposures presented in the X-ray can change significantly over time. Below, we plot the value exposure of the MSCI USA Minimum Volatility Index over rolling 3-year periods. As the X-ray indicates, the average exposure to value is generally positive. However, it has dipped negative recently. This implies that recent constituents are actually overvalued relative to other stocks in the universe.

Source: Data from MSCI and AQR, Calculations by Newfound Research. Past performance does not guarantee future results. Covers the period from June 1988 to May 2016.

As we’ve already discussed, this does not necessarily mean that trouble is around the corner. It does suggest, however, that the strategy, since it doesn’t account for price in its buy/sell decisions, can potentially get over-concentrated in expensive companies. This may expose the strategy to losses if a bubble develops in lower volatility stocks and then it pops before the strategy has the ability to reduce exposure.

Risk #3: Expectations

This one is simple. No factor strategy, even low volatility, is perfect. Low volatility strategies have gone through periods of significant underperformance in the past and they will again in the future. When this does happen, it’s crucial to remember that this is just part of the ride and no reason to abandon ship.

Source: Data from MSCI, Calculations by Newfound Research. Past performance does not guarantee future results. Covers the period from June 1988 to May 2016.

Risk #4: Process Risk

No process is perfect. Low volatility strategies use trailing volatility as a proxy for risk. Trailing volatility is appealing for its simplicity and timeliness. However, it has its drawbacks as well. Trailing volatility can fail to accurately forecast future volatility. In addition, it ignores additional information that may be available about a company (e.g. financial statement data).

In our opinion, one of the most powerful ways to address process risk – and many other risks specific to a given strategy – is through strategy diversification. In this context, strategy diversification means finding other managers or strategies that further the goal of risk aware equity exposure while mitigating the risks a pure low volatility approach.

A Case Study in Strategy Diversification

For an investor looking for lower risk exposure to the equity markets, the process might go something like this:

Start with a low volatility ETF like USMV (tracks the MSCI USA Minimum Volatility Index)

Recognize that volatility isn’t the only measure of company risk and so complement USMV with another strategy that measures risk through a different lens. This complement could be a quality ETF that uses fundamental measures of risk like earnings growth or a dividend growth ETF. For this example, we will use NOBL, a ProShares ETF that tracks the S&P 500 Dividend Aristocrats ETF. This strategy invests in U.S. large-cap companies with a 25+ year history of growing their dividend.

The potential risk with both USMV and NOBL is that neither uses price in their security selection or weighting methodology. This suggests that we could be exposed to stock bubbles to the extent that inflated companies find their way into either strategy. We can address this by adding a strategy that will maintain a more consistent value tilt. In the U.S. large-cap sleeve, this could mean an ETF like RSP that tracks the S&P 500 Equal Weight Index. This product will hold roughly equal amounts of each S&P 500 constituent. It tends to have a significant value tilt without roaming into deep value territory that might do more harm than good for risk averse equity investors.

Finally, we recognize diversification isn’t perfect. In markets like 2008, when most companies are crashing simultaneously, there is a limit on the amount of protection that diversification can offer. For these scenarios, we add in a bit of tactical equity. For this sleeve, we constructed a hypothetical strategy that uses momentum to allocate between equities and short-term U.S. Treasuries.

To keep things simple, let’s assume an equal allocation to each of these four strategies.

Below we present summary statistics for each of the sleeves as well as the hypothetical multi-strategy portfolio. We see that the multi-strategy portfolio is able to offer similar risk metrics to a pure low volatility approach with higher risk-adjusted returns.

More importantly, we avoid having our success or failure ride on a single approach. In normal market environments, this can reduce tracking error to the market, making it easier to manage client expectations. In times of market crisis, it increases the odds that the portfolio will be equipped to deal with the next crash regardless of what it looks like.

Performance Comparison

Source: Data from MSCI, S&P, Yahoo! Finance, Calculations by Newfound Research. Past performance does not guarantee future results. Covers the period from May 1996 to July 2016. Index proxies are used prior to ETF launch. Index returns are adjusted to account for approximate ETF management fees. You cannot invest directly in an index and unmanaged index returns do not reflect any fees, expenses, or trading costs except for the previously noted ETF management fee estimate. All returns include reinvestment of dividends and capital gains. The tactical equity strategy uses a 50-day/200-day moving average cross over to allocate between an equal weight portfolio of five factor ETFs and a short-term Treasury mutual fund. This index, as well as the multi-strategy example portfolio, are hypothetical and backtested and use the benefit of hindsight.

Annual Returns Relative to SPY

Low volatility equity ETFs provide investors with valuable building blocks when constructing risk-managed portfolios. Yet, low volatility is not low risk. Instead, these strategies typically trade a bit of market risk for exposure to other risks, including interest rate risk, client expectation risk, and process risk. One way to mitigate these risks is to complement low volatility equities with other strategies that use different means to get to the same end: risk-managed equity exposure. Potential complementary strategies may include quality, dividend growth, value, and tactical equity.

[1] The correlation of -0.01 was for a non-industry neutral version of the factor L/S portfolio. The correlation rises slightly to 0.10 for an industry neutral portfolio.

[2] The AQR paper does a wonderful job decomposing why this is the case. In this particular example, it has to do with differences between how the value spread is calculated (dollar neutral) and how the low beta long/short portfolio is constructed (beta neutral). We encourage readers to read the AQR paper for great insights into this topic.