1. The minimum volatility portfolio (MVP) exists only in theory. In practice, the MVP can only be determined historically (ex post) for a specific sample and return frequency. This means different low volatility portfolios (LVP) co-exist, all aiming to reduce and minimize future volatility (ex ante). In general, most LVPs have high average exposures to low volatile stocks2 and low beta stocks. As a result, the name of this new emerging investment style evolved over time from ‘minimum variance’, or ‘minimum volatility’ to ‘low volatility’.
2. LVPs achieve risk reduction of about 30%. Risk reduction varies between 15% to 45% across historical samples and economic regimes. Risk reduction is about 10% higher when currency risk is hedged and 10% lower when unhedged3. As equity risk is the most important risk factor for most portfolios, LVPs offer huge opportunities for significant downside risk reduction. This is possible while still maintaining full exposure to the equity risk premium in the long run.
3. LVPs profit from the oldest anomaly, but are also a relatively new phenomenon. The first academically documented alphas were found in low beta stocks as early as the 1970s. This low beta anomaly was discovered many years before the size, value and momentum effects were documented, and just a few years after the Capital Asset Pricing Model (CAPM) was developed.
4. LVPs' alpha is not the result of a magic formula, but is instead driven by persistent behavioral effects that cause markets to be inefficient. In the growing amount of literature on this subject, explanations for a structural alpha in low risk stocks are: (1) an increasing number of market participants focus on tracking error instead of total risk and from this perspective low risk stocks are ‘high risk’ and therefore unattractive. (2) Many investors are unwilling or unable to apply leverage in their portfolios. All else being equal, more balance sheet leverage leads to a higher expected equity returns, and so return-seeking investors tend to prefer high risk stocks. (3) The lottery ticket effect. A large number of risk-seeking investors buy volatile stocks to get rich quickly. (4) Attention bias. Stocks of companies which are in the news generate attention. This generally motivates investors to buy rather sell, as most investors own only a limited number of stocks and cannot easily sell a stock they do not own. (5) The winner’s curse. As a result of asymmetric information, the highest bidder often pays more for a stock than its true intrinsic value. The winner's curse applies more to highly volatile stocks than to stocks with low volatility4.
5. LVPs can be constructed with varying correlation dependence. A correlation estimate is unnecessary if stocks are sorted on total return volatility, but in practice correlations are taken into account and hybrid approaches are commonplace. The degree of correlation dependence should be managed in order to avoid the ‘maximizing errors’ problem, which tends to produce inefficient portfolios that require a lot of turnover. A literature survey shows that the different LVP approaches tend to produce similar levels of risk reduction and 30% turnover is enough to reduce risk5. Since correlations are not stable and risky low correlation stocks have low alphas we advise caution when using correlations6. All the currently available low volatility strategies successfully significantly reduce downside risk.
6. LVPs can also outperform during bull markets. A common misconception is to think that LVPs' low beta is a perfect predictor for future returns7. So if markets are expected to go up, then LVPs will underperform. In other words, the CAPM holds true. If this were the case, however, then LVPs would not contain alpha in the first place. This reasoning also implies that every investor with a bullish view on equities in general should not buy into LVPs, but just stick to high beta cyclical stocks and every investor with a more bearish view should abandon equities altogether.
7. LVPs generate huge tracking errors of 6-12% when compared to traditional market-capitalization weighted indices. But for other investment solutions which aim to reduce downside risk, such as put options, Constant Proportion Portfolio Insurance (CPPI) techniques or managed volatility products, nobody calculates the tracking errors. To stretch this argument to the extreme, consider a stock which is certain to generate 10% each year. For this stock the tracking error is equal to equity volatility of about 20%, but why would you care? In the end, absolute return per unit of risk is the objective of any strategy, including LVPs8.
8. LVPs exhibit time-varying style exposures. LVPs had a value bias in 2006-2007, but this shifted to growth in 2008-2009. On average, value stocks tend to have lower risk, but since this risk tends to increase during recessions, LVPs are sometimes tilted to growth stocks, especially in periods of economic uncertainty. One could also say that value has a time-varying beta, which rises during bad times such as recessions and declines when the outlook is positive. Over the past few years we have written several papers on this topic9.
9. LVPs tend to have somewhat higher interest rate sensitivity. Typically, when bond yields go down, low volatility stocks tend to outperform10. This feature is particularly interesting for pension funds aiming to stabilize their coverage ratios. Since falling bond yields tends to reduce the coverage ratio, LVPs can be used as an indirect hedge.
10. The alpha of LVPs is very difficult to arbitrage away, in contrast to better known alphas such as value and momentum. Not all low-volatility stocks11 have the same alpha and ‘good’ low volatility stocks can significantly outperform ‘bad’ low volatility stocks. To catch the alpha in the low volatility segment of the stock market, either the market capitalization benchmark should be completely abolished and ignored, or the Strategic Asset Allocation (SAA) framework should be adjusted to include a separate style allocation to LVPs12. In contrast to other alphas, the effect is also strong for large-cap stocks, stable across regions and has become stronger over the last few decades. I therefore believe that low volatility is a strong and significant anomaly that will continue to generate superior returns for a long time to come.
This article was written in October 2010 and updated in July 2017
1Examples are Acadian, Analytics Investors, Robeco, State Street, Unigestion.
2An Edhec research paper (November 2008) shows that minimum volatility simply loads on a low volatile factor using a return-based style analysis.
3Managing FX risk in low-volatility strategies, Robeco research paper July 2013 and ‘The impact of currency choice on minimum variance portfolios, HSBC research paper November 2015.
4Explanations for the Volatility Effect: An Overview Based on the CAPM Assumptions, Blitz, Falkenstein and Van Vliet, Journal of Portfolio Management, Spring 2014.
5Low-volatility needs little trading, Pim van Vliet, Journal of Portfolio Management, 2018.
6Be cautious with correlations in low-volatility strategies, Robeco research paper June 2015.
7Low-volatility investing: Expect the unexpected, Robeco research paper, October 2014.
8Benchmarking low-volatility strategies, Blitz and Van Vliet, Journal of Indexing 2011.
9Enhancing a low-volatility strategy is particularly helpful when generic low-volatility is expensive, Robeco Research Paper, June 2012. The value of low volatility, David Blitz, Journal of Portfolio Management, 2016. Is there still value in low-volatility? Robeco research paper September 2016.
10Interest rate risk in low-volatility strategies, Robeco research paper, June 2014.
11The beauty and the beast of low-volatility investing, Robeco research paper, February 2015.
12Black argues in ‘Beta and Return’ that investors who normally hold both equities and bonds can shift to a portfolio of similar total risk and higher expected return by emphasizing low-beta stocks. Journal of Portfolio Management, Fall 1993.