Your account has not been activated, please check your inbox for the activation email.

Unenviable task – searching for alpha and managing risk

Amanda White




In the wake of the global financial crisis, investors have the unenviable task of searching for alpha and managing portfolio risk. Many investors, including Sunsuper, AustralianSuper and Hostplus, are now increasing their risk budgets allocation for alternative assets, rather than within equities.

Andrew Fisher, portfolio manager at Sunsuper, says alternative assets is where the fund is allocating its incremental active risk budget, and this is partly due to the view that it is not a great environment for active returns in equities.

Similarly, Innes McKeand, head of equities at AustralianSuper, says most of the fund’s risk budget is used on infrastructure, property and credit.

Other funds, like Cbus and VicSuper, have taken it a step further and actually trimmed their allocation to equities.


Cbus ‘actively de-risking’

“We’ve been aggressively de-risking over the past 18 months,” says Brett Chatfield, investment manager at Cbus. “All new money has been going to build up cash and then opportunities that have arisen from that.”

Within equities, Cbus has been using low volatility within global equities, and VicSuper has also used low volatility strategies to help manage portfolio volatility.

Meanwhile First State Super has hired Eben Van Wyk, portfolio manager of systematic beta, to manage factor exposures internally within the fund, and manage built-up biases, or tilt the portfolio to factors that make sense in various conditions.

“When we fully insource the management of the core fund, we can do more ad hoc or more customised portfolio construction and we can bring in multiple factor risk tilts, including low volatility, Van Wyk says.

“I’m just a bit uncomfortable maybe with only having one factor, which is low volatility, because that tends to potentially expose you to the risks that are not reflected within stock prices – systematic risks, like overcrowding potentially in low volatility, or liquidity risks.”

Van Wyk highlighted the behavioural anomaly within low volatility, and that high-risk stocks are overvalued because of behavioural factors. He said he thinks low volatility might be overpriced because of the demand by investors.

But Ruben Feldman, director of business development at STOXX Ltd, says it is important to distinguish between low risk or low volatility and minimum variance portfolio construction.

“What I would distinguish very strongly is that constructing a minimum variance portfolio doesn’t necessarily involve being exposed to the low risk factor at all; it’s not a single factor portfolio management strategy. Instead, it gives you the portfolio that has the lowest risk. Minimum variance looks at correlations between the stocks, it also can look at skewness.

“There’s a whole slew of information that goes in there. The minimum risk portfolio goes in and out of factors. Obviously there’s a strong tendency to be highly exposed to the low volatility factor, but if you look at momentum or value, active allocations vary through time. All those risk factors are strategically allocated to, in a systematic way. Some factors get very risky just before they blow up, minimum variance actually picks that up and diversifies away from those factors.”


Minimum variance strategies

This is because minimum variance does not only look at individual stocks, but also looks at correlations and at the risk of the portfolio. In case the portfolio contains factor exposures that become risky, those exposures are reduced.

By way of example, Feldman says well-built minimum variance strategies picked up that financial stocks were moving as a group just before the global financial crisis and diversified away from that. In contrast, low volatility strategies actually “piled up on financials” because they were lower risk.

“I think that correlation is a huge component of risk,” he says. “Obviously the underlying asset volatility is important, as is the estimation timeframe and how long your risk timeframe is. But definitely correlations are extremely important and that’s what differentiates a minimum variance portfolio to a low risk factor allocation.”

It could be said that smart beta, as a marketing term, has been almost overdone, but at its core, managing factor tilts is an important part of risk management, and alpha generation, in equities portfolios.

In recent times, minimum variance portfolios have been seen as delivering the best of all worlds, including low risk, low drawdowns and strong returns.

At the roundtable, Feldman spoke about a large US pension fund client that had recently used minimum variance to solve “multiple problems”.

“They’ve essentially reduced their fixed income allocation and replaced active managers and market cap weighted equity allocation to create a minimum variance allocation. They have thus increased their overall equity allocation, but risk has been reduced by slowly shifting their core equity allocation to minimum variance, as opposed to market cap weighting,” he says.

“Through that they’ve been able to reduce the cost because of less active management, increase total expected return on the entire fund because there’s a higher equity allocation with the same amount of risk, resulting from the shift to a minimum variance allocation. The pension fund is relatively large with trading capacity, but obviously the use of an index makes it very easy for them to implement in-house. They don’t even need to externalise the implementation either, so they have more transparency and more control.”


Caution re different environments

But portfolio manager of Australian equities at BT Financial Group, Sonia Bluzmanis, expressed caution about the different environments, or markets, in which minimum variance might work.

“Generally the more efficient a market is – so if you look at the US or global blue chips – minimum variance adds less value,” Feldman says.

China A shares, by example, presented an opportunity to add value in minimum variance by reducing market cap volatility – by between 30 and 40 per cent – and producing slightly higher performance. But there are also liquidity and rebalancing issues to consider.

“We have two versions of minimum variance; an unconstrained one, and then a constrained one. The latter constrains the minimum variance to be similar in terms of risk allocation to most risk factors, geographies and industries to be similar to the benchmark, so that the tracking error is limited. But in the end the only times when minimum variance, or at least historically, has underperformed the markets, is when they have been on a sharp up. And so to underperform when you’re only making 10 per cent instead of 20 per cent is not as bad as underperforming under any other circumstances,” Feldman says.

Investors agreed that getting access to the equity risk premium at a lower risk is a bonus. In addition, the market cap benchmark was not held in high regard.

“Market cap is the worst, most inefficient benchmark,” says Van Wyk.

“It could just be that a systematic or a quantitative model is not very good at predicting alpha, whereas fundamental managers are probably the best place you can get your alpha from, from benchmark unaware, concentrated, highly skilled, fundamental managers. There’s your alpha part and at the core you might want to say, “Okay, I want a passive equity risk premium.” Instead of a market cap tracker I can get a low volatility portfolio and it’s going to be lower risk.”

“I think the logic behind why market cap dominates is fairly easy to see, it’s the portfolio you don’t have to rebalance, you buy and hold it and there’s no transaction cost really,” Sunsuper’s Fisher says.

Dmitry Capel, head of equities at Hostplus, said it was difficult for a fund like Hostplus to change its approach, after having so much success with its strategy of high-conviction equities strategies and high allocation to alternatives.

“As an alternative to market cap weighted indices, it’s definitely worth exploring and considering,” he says. “Our equities are relatively high conviction, a lot of them are unconstrained portfolios that have historically done very well for us. It’s hard to deviate – I mean every investor and every fund is guided by its own history I suppose, but how do you make an argument for a fund like ours that’s done really well historically getting exposure to proper, active strategies? That’s the challenge I think.

“Some managers have been quite successful but how much real alpha are they adding in addition to what can be readily explained by factors that are available to us, that’s some of the work that we’re looking to do, to inform our views about should we continue to support active managers, and what can be replicated.”


Strong argument

But Feldman believes that the performance, and risk reduction of minimum variance portfolios is a strong argument.

“I think over the past 100 years minimum variance has significantly outperformed and delivered a risk that’s half to two thirds,” he says. “I also think that even if you have alpha generators, that doesn’t necessarily mean that you have a good portfolio. You could have lots of stocks that provide alpha but if they all tank at the same time, you may have people getting out of the portfolio at the wrong time and never getting the alpha that is being proposed and so I think that minimum variance actually provides that.”

Fisher added that Sunsuper is looking at minimum variance as an alternative benchmark, but he had queries about whether correlations were a better indication than prices to determine the future.

“The real question is ‘is risk easier to estimate than returns’?” Feldman says. “Definitely I would say yes. The estimation of risk empirically over the past 100 years or so – where we have equity markets data available – has been pretty stable, whereas returns estimation is next to impossible. You’re not assuming past risk is equal to future risk, there’s a whole estimation in there and you’re trying to get predicted risk, you’re not just estimating past risk, if that makes sense.”

But Fisher was concerned about the tracking error and where, or how, minimum variance would fit into the equities portfolio. Sunsuper splits its portfolio into three portions – a passive portfolio with limited tracking error, systematic factor based area with tracking error between 1 and 2 per cent, and unconstrained with tracking error of between 4 to 8 per cent.

“I mean so it could be that a minimum variance goes into the unconstrained part of the portfolio but it seems unlikely to me that it would go that path,” he says.

And as Van Wyk points out, it is important to consider not just the low volatility factor, but efficient portfolio construction.

“I know from people I speak to, some just prefer simple, arbitrary rules in portfolio construction, whereas I’ve used optimisers for a very long time, so I’m more comfortable with it,” he says.


‘Train and constrain’ optimisers

Feldman says it is important to “train and constrain” optimisers to deliver a sensible result.

“I think you can get weird looking portfolios, unless you put enough constraints around them to force it to have enough diversification and to be sensible. You might end up with a portfolio with 20 per cent in one stock because it is producing something that will mathematically give you the best result. But it has to pass a common sense sniff test,” Feldman says.

Van Wyck identifies efficient portfolio construction as the defining criteria of minimum variance.

“I think the difference between minimum variance and this whole smart beta movement is the smart beta movement, or scientific movement, or systematic factors, whatever they want to call it, is basically just the underlying building blocks of quantitative fund management,” he says. “But minimum variance includes efficient portfolio construction including the use of an optimiser, and a premium should be paid for portfolio construction.”

Click here to download a PDF of the roundtable