Low-volatility portfolios seem, empirically, to outperform high-volatility portfolios, and there are plenty of theories to explain why. But Lynn Strongin Dodds finds practitioners tinkering with the pure expression of this insight because lowering volatility risk results in other risks popping up in its place

The idea that a low-volatility equities portfolio might perform better than standard modern portfolio theory predicts is not new – indeed, it is almost as old as modern portfolio theory itself. And products based on the concept didn’t just appear yesterday – Wells Fargo launched a strategy called Stagecoach as far back as 1972. But the popularity of this strategy waxes and wanes with the vagaries of the stock market, and it is no surprise given the euro-zone crisis and uncertain economic outlook that it is once again back in vogue.

And ‘vogue’ can be the operative word. Typically a minimum variance portfolio consists of holdings from the unglamorous side of the investment tracks – utilities, telecommunications, healthcare and consumer staples. They are defensive stocks that weather the storms and offer the much sought-after downside protection in bear markets. These strategies lose their appeal when prices start to climb and these defensive stocks are unceremoniously dumped in favour of edgier companies better placed to ride the bull-market wave. The more exciting stocks also fit better into a fund manager’s compensation game plan: some are rewarded for performance which is typically found in companies with higher risk profiles; others are benchmarked against market indices and the easiest way to beat those indices is to hang on to the coat-tails of high-beta stocks as those markets rally.

In fact, over short periods, minimum variance will dramatically lag a euphoric market.  For example, in the year ending March 2010 the MSCI World index jumped by 52% while US-based minimum variance managers fell behind their benchmarks by around 15–25%. Proponents of the strategy point to numerous empirical and industry studies that suggest they will keep pace with the traditional market-cap benchmark if the rise is more gradual and deliver superior risk-adjusted returns over a full cycle. According to Pim van Vliet, senior portfolio manager at Robeco Conservative Equities, academic research shows the returns are up to one percentage point better than market cap weighted benchmarks with roughly a third of the volatility.

Institutional investors need to adopt a long-term horizon. For example, the FTSE 100 Minimum Variance index significantly outshone the FTSE 100 in the three months to 30 June 2012, with a 1.7% return versus –2.3%. But over a 12-month period, the relative performance was even more pronounced with the minimum variance index returning 6.6% versus –2.7% for the market cap-weighted index. In the 12 years since its inception, the alternative index has outperformed more than threefold, providing a 160% return versus 49% for the FTSE 100.

The same holds for the US on an even longer time scale. Studies show that the least-volatile quintile of the 1,000 biggest stocks in the country returned 10.2% annually from 1968 to 2010, while the most-volatile quintile gained only 6.6%, according to Brendan Bradley, a director of managed volatility strategies at US based Acadian Asset Management, who earlier this year published a study on the returns of low-volatility stocks with Malcolm Baker of Harvard University and Jeffrey Wurgler of New York University in the Financial Analysts Journal. The US stock market overall returned 9.6% during the same period.

These results challenge the long-held traditional Capital Asset Pricing Model (CAPM) which argues that an asset’s expected return is directly proportional to its beta or systematic risk – or, in other words, riskier securities should be rewarded with higher expected returns and vice versa.

This ‘anomaly’ is partly driven by investor behaviour, according to research from Acadian Asset Management. First is the ‘lottery effect’, whereby investors will take a punt with a risk of near certain loss if the potential payoff – however unlikely – is sufficiently large. They also have an irrational support for high-risk stocks from a ‘representativeness’ bias, which is the tendency for individuals to jump to a simplistic and often incorrect conclusion from just a few observations. Last but not least, growth stocks can elicit a sense of overconfidence: when valuing stocks fund managers can veer on the side of optimism and stick with the false precision of their estimates. There is also a structural driver that can be tied to the asset management industry’s dependence on the tried-and-tested benchmarks – despite the performance history of low-volatility stocks.

Despite the track record, as with all strategies, there is no free lunch and minimum variance does have its share of disclaimers. The biggest complaints, according to François Millet, product line manager for indexing at Lyxor Asset Management, are a high tracking error (around 8%), risk concentrations in a select sub-set of stocks and “a less robust framework”. He adds: “There are other low-volatility strategies such as risk parity that produce better comparable risk-adjusted returns at a lower tracking error.”

Michael Fraikin, director portfolio management at Invesco, also notes that the rush to buy the lowest beta stocks has made minimum variance a moderately expensive proposition. “By the time most investors realise they have bought something overcrowded another 12–18 months may have passed, which is why it is so important to take an active approach to minimum variance,” he suggests.

Research from AllianceBernstein reveals that, while most periods since the early 1970s have seen low-beta stocks trading at a discount to the broader index, in this past year the price-to-book valuations of the lowest-beta stocks were about 22% higher than those of a universe of global large-cap stocks. The premium was even higher in 2008 and 2009.

Fraikin also warns that the future may not simply be a replay of the past. “There may be several studies that show minimum variance outperforming the market cap benchmark over a 20 year periods or more, but the past could be kinder than the future,” he says. “Investors may find new surprises and they may not benefit from risk reduction coupled with higher returns going forward.”

Given these criticisms it is no surprise, perhaps, that most fund managers or index providers do not solely follow the unadulterated quants-based approach to minimum variance.

“It is easy to get into mathematical arguments but if you try to blindly apply quantitative criteria then there is no stability,” says Bertrand Delarue, global head of institutional product engineering at BNP Paribas Investment Partners, which applies equal weights to its minimum variance portfolios. “It is important to look at different techniques of building a portfolio such fixing maximum weights or adding fundamental criteria.”

Alexei Jourovski, managing director, head of equities at Unigestion – a pioneer that launched its first minimum variance fund on the Swiss market 15 years ago and has since extended its strategy to European, global, Japanese, US and emerging markets – concedes that it is important to avoid ending up with a concentration in a small number of illiquid stocks.

“There are two ways around this,” he suggests. “You can put in constraints, or you can do what we have done, which is use a broader spectrum of risk indicators, including not only correlation and volatility, but also fundamentals, debt spreads, news-flow and liquidity.”

Peter Gunthorp, managing director in research analytics at FTSE Group, agrees. “This is why we have a maximum upper stock weight limit as well as industry and country constraints,” he says. “We also have a diversification target that allows us to target a minimum number of stocks.”

FTSE recently broadened its minimum-variance range by adding eight new benchmarks to provide a wider geographical reach. The indices, which cover Asia, Europe and the US, follow on from the launch earlier this year of a version of the FTSE 100 index of UK-listed large-capitalisation stocks.

MSCI also places caps on top of its global minimum variance range. “There are different ways to construct a low-volatility portfolio,” says Dimitris Melas, global head of new product research at MSCI. “We use optimisation and impose constraints on the weights of stocks, sectors and countries in the MSCI Minimum Volatility indices. These constraints ensure that the indices are investable while retaining the main characteristics of unconstrained minimum variance strategies.”

STOXX, on the other hand, is a newcomer to the game, having recently launched its family of minimum variance indices in conjunction with Axioma, a risk management firm. It comes in two flavours – constrained and unconstrained. In the former, the exposure to Axioma’s fundamental factors is limited to one quarter of the standard deviation of the underlying index’s exposure, with the exception of size and risk, which are not used as constraints. Caps have been placed on full investability, component weight, diversification, turnover, currencies and industry exposure. They are rebalanced quarterly in line with the respective underlying index. The unconstrained variations are not subject to any factor, currency or industry exposure requirements, and are rebalanced on a monthly basis.

Susanne Willumsen, fund manager of Lazard Asset Management’s global controlled volatility fund, which made its debut earlier in the year, also believes that evolving cross-correlations between stocks are important factors to be incorporated when building portfolios.

“We find positions that offset one another’s risk,” she explains. “For example, in the energy and transport sectors we may have two companies exposed to oil prices but they will behave differently. The airline stock would be negatively affected by rising oil prices but that would be counterbalanced by an energy company which would benefit.”

Looking ahead, the debates and discussions over the best ways to construct a minimum variance portfolio will continue. There will also be questions as to whether they lose their effect as interest grows. The general view is that in order for the low volatility anomaly to be arbitraged away, the benchmark-focused nature of the industry would have to radically alter and this is unlikely to happen. Although strategies such as minimum variance are currently fashionable, they still account for a fraction of the roughly $10trn that is currently managed in benchmark-sensitive and passive portfolios.