Lower Volatility Smart Beta Funds – A Safe Haven in Turbulent Times? Part 2 of a Series on Multifactor Smart Beta ETFs
In this post we take a closer look at an important building block of many multi-factor portfolios, low volatility. Low volatility funds seek to take advantage of the “low volatility anomaly” – the empirical observation that lower risk, securities outperform their higher volatility counterparts.
Smart Beta funds are hot. According to ETF.com, more than half of the 150 funds launched in 2016 implemented smart beta strategies. For the year to June 30, 2016, ETFGI’s most recent data show that assets in smart beta funds have a five-year annual compound growth rate of 31.3 percent. And, low volatility funds, up $15.1 billion in the first seven months of the year are the most popular. BlackRock’s Holly Framsted called minimum volatility funds the “fastest growing” smart beta segment. In this report we examine the sector and factor exposure of a low volatility index to see how the strategy may fare if volatility increases.
In our earlier post on multi-factor smart beta products we noted the difficulty in assessing performance, partially because we don’t have clear expectations about how these funds should behave. We noted earlier that some funds which target the exact same factors perform quite differently.
In this post we take a closer look at an important building block of many multi-factor portfolios, low volatility. Critics have cautioned against low volatility strategies citing sector concentration, rate sensitivity and potential crowding risks. Anticipated rate rises, currently, may be contributing to withdrawals from low volatility funds after a long period of inflows.
Low volatility funds seek to take advantage of the “low volatility anomaly” – the empirical observation that lower risk, (as measured by beta or volatility) securities outperform their higher volatility counterparts. Researchers have advanced several reasons1 for this anomaly – investor preference for lottery like payoffs combined with leverage or other investment constraints. Some question the existence of the anomaly at all, citing instances where the observations are not robust to changes in data frequency, horizon or some other change in methodology. Whether one believes in the low volatility anomaly or not, there is also the draw implicit in the name. Post-financial crisis, the simple promise of mitigating losses has its own appeal.
There are a number of easily investible low-volatility products, both mutual funds and ETFs on the market. In this note, we focus on the PowerShares S&P Low Volatility fund (SPLV), started in 2011, and the index it tracks, the S&P 500 Low Volatility Index2.
|Index Tracked||S&P 500 Low Volatility Index||S&P 500 Index|
|Index Inclusion||100 stocks of the S&P 500 index with lowest recent 12 month realized volatility||S&P 500 stocks|
|Weighting||Inverse Volatility||Market Capitalization|
|Ann Std Dev (ETF/Index full hist)||9.4%/11.0%||11.9%/14.3%|
|Ann Return (ETF/Index full hist)||13.0%/11.1%||11.8%/10.0%|
|Sharpe Ratio (ETF/Index full hist)||1.35/0.75||0.99/0.54|
Source: Morningstar, Powershares
ETF performance data Jun 2011-Aug 2016
Index performance data Dec 1990-Aug 2016
The S&P 500 Low Volatility Index was launched in April, 2011, with history backfilled to 1990. While experience (and providers’ disclaimers) teach us to be wary of back-tested results, a look at the longer term behavior of the indexes helps establish expectations for the strategy.
- Using the S&P 500 Index as a benchmark, the index tends to deliver on the promise of lower volatility on average, although there are some periods of volatility higher than that of the index3, where the relative standard deviation values on the plot are greater than one.
- Outperformance claims are less supported. While the index does outperform over the full period, there are extended periods of underperformance observed.
- Although not shown in the chart above, the SPLV index can be tracked with high accuracy with an R-Squared value over 95% and predictability, as measured by MPI’s proprietary measure of Predicted R-Squared of over 90% using a portfolio of sector indices.5
- The fund’s implied sector exposures change over time, in particular post-crises. Note that in 2009, prior exposure to financials is reduced, for obvious reasons. The index’s exposure to Utilities also fell following 2013’s “taper tantrum”.
- Sector concentration is clearly visible, and it confirms the intuition that the strategy is highly focused in defensive sectors, more so since 2008.
While investible sectors provide the best fit, it is also insightful to review the relationship between the index’s historical return series and the Fama/French 5 factor model – book value (high vs. low)6, market cap (small vs. large), Beta (market exposure), profitability (robust or weak) and the firm’s capital investment level (conservative or aggressive). These factors, supplemented with a stock price momentum measure and the price series of the ten year constant maturity US Treasury index, are regressed with the index’s return history in the chart below.
- Exposures have hanged since 2008, with Value (High-Low) exposure becoming negative and momentum increasing.
- Rate sensitivity also increases, as does the magnitude on the quality (profitability and the firm’s capital investment level) factors. These reinforce the picture painted by the sector analysis.
Higher frequency analysis of the ETF product, the only out-of-sample test, supports the longer term analysis of the index, with total risk and return numbers reported in the table above and similar exposures. Higher frequency analysis also observes periods such as the taper tantrum and recent weeks where the low volatility products can be in fact both as or more volatile than the market and underperform as well. Between May 20, 2013 and June 20, 2013 SPLV returned -7.1%, compared to a loss of -4.6% for SPY with the same daily volatility. More recently, SPLV lost -5.2% between July 25, 2016 and Sept 9, 2016, while SPY returned a loss of -1.9% with lower daily volatility than SPLV.
So what are the implications of this analysis for investors? Low volatility investing is an active strategy with active sector and factor tilts. While these tilts do change over time, they appear persistent over moderately long periods, changing rapidly only with regime changes. Recent experience has shown that current tilts are sensitive to interest rate shocks, and long term experience suggests that defensive sectors tend to underperform the market in a rising rate environment; this is reinforced by the negative relationship with rising rates on a factor basis. Positive exposures to momentum, which itself is subject to periodic crashes as well as recently limited value exposure supports the contention that the strategy is relatively more expensive than it has been historically. For those interested in tactical allocation, we’ll leave it at that.
On a longer term basis, low volatility tends to deliver on the only characteristic it truly targets, which is a lower standard deviation of returns than its parent index, albeit with high sector concentration and higher turnover.
As a component of a multi-factor portfolio, the strategy’s exposures to other targeted factors should be a major consideration. While significant interest rate exposure might be unique to this particular strategy, the quality and momentum factor exposures could possibly result in an unintentional “doubling down” of factors in a multi-factor portfolio, an analysis we will provide in a subsequent post.
DISCLAIMER: MPI conducts returns-based analyses and, beyond any public information, does not claim to know or imply what the actual strategy, positions or holdings of the funds discussed are, nor are we commenting on the quality or merits of the actual investment strategies. This analysis is purely returns-based and does not reflect insights into actual holdings. Deviations between our analysis and the actual holdings and/or management decisions made by funds are expected and inherent in any quantitative risk factor analysis. MPI makes no warranties or guarantees as to the accuracy of this statistical analysis, nor does it take any responsibility for investment decisions made by any parties based on this analysis.
- 1See “Betting Against Beta”, Frazzini and Pederson here or “Benchmarks as Limits to Arbitrage: Understanding the Low Volatility Anomaly”, Baker, Bradley and Wurgler here.
- 2S&P does not constrain the fund by sector/security minimums or caps, emphasizing the role of strategy implementation in the analysis.
- 3The chart measures the volatility of monthly returns as is reported in the products’ fact sheets. Results are similar on a weekly basis. On a daily basis, the volatility of the Low Volatility index only exceeds that of the S&P 500 during some periods when a shorter window such as 3 months is used.
- 4While the goodness of fit analysis (R-squared) uses Cash and ten sectors to measure fit, in the chart, we purposely exclude the cash component to make the chart more readable.
- 5Equal weighted sectors are used in the analysis shown. We the index’s historical return series using both cap-weighted and equal-weighted sectors indexes and found equal-weighted to provide the best fit, indicating the presence of a mid or small-cap bias. The cap weighted sector analysis provides similar but not identical sector exposures.
- 6Note that the Value factor may be redundant, subsumed by the two “quality” factors. We retain it as the most frequently targeted factor, whose relationship with other factors is germane to the discussion. https://famafrench.dimensional.com/questions-answers/qa-what-does-it-mean-to-say-hml-is-redundant.aspx
- 7This chart plots the R-squared between the index return series and the factor return series for each of the seven factors (the five Fama/French factors plus momentum and the constant ten-year US Treasury return factors). Each of these seven R-squared values are plotted as a stack in the chart. Hence, the total factor exposure may be higher or lower than one, depending upon the strength of the relationship between the index’s and factor’s return series.