When IPE first spoke with Ian Heslop about the post-crisis refinements that Old Mutual Asset Managers (OMAM) had made to its quantitative equity models, it was June of 2011. The sun was shining - literally, and (for quants) metaphorically, too.

Years of increasing pairwise stock correlation that plateaued through 2008-2010 seemed finally to be turning: since October 2010 correlations had been falling. “Pairwise stock correlations are a good predictor of whether quant strategies will or will not work,” Heslop said. “It’s a stockpicking environment, with more stabilised return profiles that are much more correlated to earnings changes.”

Summer has given way to a quickly darkening autumn. Macro and stock correlations are back with a vengeance. But OMAM is coping well: despite shedding 0.6% in September, its long/short global equity strategy finished Q3 up 6.4% and is up 14% year-to-date. Long-only iterations are “broadly in-line” with this, says Heslop. This is partly thanks to market structure - while stock correlations are up, the macro drivers of that correlation have been steady, resulting in stable returns to, and correlations between, certain factors. But OMAM’s refinements have been instrumental in exploiting these moves.

Quants face a dilemma. The crisis revealed just how similar their risk (if not their stock) positions were - but that’s the nature of highly-diversified portfolios. To put it bluntly, there are limited ways of squeezing alpha from markets.

“As a quant, you run the numbers,” says Heslop. “A model that has value, momentum and a little bit of growth explains 80% of the returns. These very predictable themes capture inherent behavioural biases, and investor behaviour doesn’t really change.”
OMAM’s models are restrict themselves to five main factor signals: valuation; market dynamics (momentum); sustainable growth; analyst sentiment (exploiting the ‘staleness’ of market information); and fundamentals (balance sheet and management quality).

“Where we have moved on, is the way we capture these in our process,” says Heslop. “Strategies are much more heterogeneous now because they use that same fundamental understanding of why stock prices move in increasingly different ways. Pre-2007, the whole quants space was based on long-term average factor returns. That’s a good starting-point, but it doesn’t reflect the impact of a changing macro environment on factor returns. Thematic returns are cyclical - and recognising this was a big step.”

While factor correlations are more stable than asset-level correlations, the fact that they are cyclical requires a certain dynamism in factor-weighting. However, you don’t want to throw the factor-stability baby out with the static-weighting bathwater.

“We did an awful lot of work on this ideal weighting scheme - and the long-term average return is very difficult to beat,” says Heslop. “You have to balance more responsiveness against the danger of periodic factor-concentration - and falling into the same behavioural biases that you’re trying to exploit.”

A weighting scheme focused on the short-term would have loaded up on momentum during the summer of 2008 and avoided the crushing losses to value in Q4. But being caught severely underweight value in Q2 2009 would have been equally crushing. Similarly, a quant strategy weighted heavily to ‘naïve’ value will get hit in a flight to quality, and it will also pick up value traps: momentum is there to exploit flights (to anything), but it also puts the brakes on longs in those value traps.

This is the point of combining these factors - they generally show low correlation with one another. Re-weight them too aggressively and you lose some of that diversification benefit; weight them statically, and you will be hit on the few occasions when they dramatically fail to diversify. The solution, for OMAM, has been to make each factor as robust in its own right as it is in combination with others: that means improving downside risk profiles with more granularity in understanding the components of some, and ‘purification’ of unintended exposures from others. Value needs to be more than ‘buying cheap stocks’; momentum needs to be more than ‘something that includes an anti-value bias’.

“If you are going to blend momentum and value in a more dynamic way you need to deal with, to follow this example, the default risk of the value-trap and the heightened perception of default risk in a flight to quality,” Heslop explains.

The essence of value - market neglect - gets swamped during market stress, when investors perceive default-risk as the key driver of returns. But the important thing is to recognise that the result is not a simple style rotation out of value and into momentum or growth, but rather a rotation from non-discerning value into ‘quality’ (management and balance sheet strength, which can coincide with value).

“Our new dynamic value signal holds that value and quality are two sides of the same coin,” says Heslop. “How you balance them depends on prevailing risk appetite. By focusing on quality explicitly, we don’t have to rely on momentum to keep us out of anti-quality in the value factor.”

That has paid-off handsomely in Q3 2011 - a punishing time for value and a rewarding time for quality, a combination which has hammered ‘naïve’ value strategies.

The move from mere ‘growth’ to ‘sustainable growth’ is an example of factor-purification. Naïve growth doesn’t have the consistency that quants like, simply because some growth stocks generate disappointing earnings, leading to capital losses as investors re-price for that disappointment. That counterbalances the true growth stocks and results in a very weak signal (unless there is a major rotation into non-discerning growth going on, such as the tech boom).

“Growth managers can run successful concentrated portfolios, but if you look at thousands of stocks you only get meaningful signals if you isolate ‘sustainability’ in that factor,” says Heslop - which essentially means assessing historical earnings against proprietary criteria for quality and stability. The result is not a recipe for an outperforming 10-stock growth fund, but it is a significant refinement when paired with the law of large numbers.

In the market dynamics factor, stock and sector-level momentum are now much more clearly defined, and OMAM has been innovative in ‘orthogonalising’ it against other factors - particularly volatility. “Momentum strategies are always problematic because they can be highly correlated with things you don’t want them to be,” as Heslop says. This stands to reason: serial correlation can occur in any part of the market - if growth stocks are trending, momentum will be highly correlated with growth. And because momentum is clearly related to prices drifting away from fair value, it is no surprise often to find it overlapping with volatility. “Part of the reason momentum periodically falls out of bed is that it is very long-vol - and we can’t predict whether we want to be long or short-vol at any particular time. In addition, stocks that are really liked by momentum and stocks that are really disliked by momentum both have high volatility - so volatility introduces a lot of noise into the signal.”

Now that OMAM strips volatility out of momentum, again, the resulting signal is purer - but weaker. “It’s weaker, obviously,” Heslop concedes. “But given the downside risk inherent in momentum strategies with the volatility present, you end up with higher returns from lower risk by taking out the volatility.”

OMAM’s efforts to make its factor signals less ‘fuzzy’ has had two apparently contradictory, but actually complementary, results. “Now, each individual strategy is made very stable in its returns, and can sit as a fundamental investment strategy in its own right,” as Heslop puts it. Before, momentum was in there to diversify certain aspects of value. Today, value has been improved so that it no longer requires that outside source of balance. But momentum has also become a better diversifier against value. Why? Because, stripped of volatility, it delivers a superior Sharpe ratio and can therefore accommodate a larger risk budget. The purer the factors, the more robust their correlations through time. In environments like today’s, these refinements come into their own.