Investors need to consider the interaction effects between different factors when constructing multi-factor portfolios, writes Paul Amery 

At a glance

• The case for holding a variety of factors is based on the cyclical character of returns for different profiles.
• Multi-factor portfolios can be bottom-up or top-down.
• The top-down approach arguably provides greater transparency.
• In practice, institutional investors may be using both approaches.

Many equity investors are taking the short step from single factor to multi-factor portfolios. But combining factors is not as straightforward as combining asset classes. Complex questions of intra-factor correlations, the desired intensity of factor exposure, costs and governance come into play.

According to index provider FTSE Russell, asset owners and large investors worldwide are increasingly interested in holding multiple factors in their equity portfolios. In a 2016 ‘smart beta’ survey of 250 large institutions with a collective $2trn (€1.9trn) under management, the firm found that the adoption levels of multi-factor strategies had doubled year-on-year to 37% of those participating.

The case for holding a variety of equity factors appears clear. The return profiles of different factors are cyclical. For example, value and small-cap stocks may do well in the early stages of an economic recovery, while momentum stocks join the party as growth accelerates.

As growth slows, quality and low-beta stocks may come to the fore, while low beta and momentum tend to be the best-performing factors in the depths of a crisis. A diversified approach to factor investing therefore makes sense. “Having factors in your portfolio is a good thing in general but there’s a tremendous power in putting them together,” says Ronen Israel, principal at AQR Capital Management.

At the same time, many investors appear uncertain about the best way of combining factors. Respondents to FTSE Russell’s 2016 smart beta survey said that their number one implementation concern was how to select the right factor strategy or combination of strategies for their portfolio.

“Initially, the factor debate centred on which factors are justified and how investors should get exposure to them,” says Felix Goltz, head of applied research at Edhec-Risk Institute. “But the debate has now moved on to more technical questions about portfolio construction. If you want to get exposure to five or six factors, how should you do it?”

Some investors allocate to factors by constructing separate portfolios for each factor and combining them. Alternatively, investors can average factor scores for each equity in a starting universe and then work out a composite factor weighting.

Such blending or averaging approaches, however, ignore the correlations between factors, which are meaningful. For example, the value and momentum factors tend to be negatively correlated, an intuitive result, given that the first strategy buys low and sells high, while momentum does the opposite.

According to calculations by State Street Global Advisors (SSGA), several other factors have exhibited dependencies on each other over the past two decades. Since 1993, based upon a broad index of global equities, value and quality, quality and size, and momentum and size have all been negatively correlated, while value and size, and quality and (low) volatility have tended to be positively correlated.

These statistical relationships suggest there may be better ways of putting factors together than simply averaging across multiple single-factor portfolios. “There are benefits in the interaction effect between factors. We can exploit these by building multi-factor portfolios from the bottom up,” says Ana Harris, European head of equity portfolio strategy at SSGA.

“In a top-down approach you consider factors individually and then blend them. In a bottom-up approach you consider all of the factors simultaneously,” says Peter Gunthorp,

managing director of research and analytics at FTSE Russell. “We think the latter approach is a far more efficient way of obtaining a given level of exposure for a given level of diversification.”

But even with bottom-up construction approaches, there can be important methodological differences. “An intersection approach to multi-factor portfolio construction requires sorting stocks by the first factor, picking the top x% of stocks, sorting that new universe by the second factor, picking the top x% of stocks, and so on,” adds Gunthorp.

“It’s a relatively hard approach to doing things as you’re either in the final portfolio or out of it. An approach involving sequential tilts is gentler – you have the intersection portfolio, plus a lot of the stocks that score highly on either of the two factors, this time with smaller weights.”

Intra factor correlations

AQR’s Israel stresses the benefits of using tilts in a bottom-up approach. “Given a choice, we’d rather put factors together in an integrated way. Investors benefit from both reduced turnover and lower transaction costs.”  

“In addition, imagine a stock just misses the cut-off point for both the value and momentum factor portfolios. A blended or intersection approach would not hold a position in that stock. But, in terms of your desired combination of factors, that stock may be the best in the universe, which an integrated, tilt approach would capture.”

Not all industry practitioners accept that bottom-up factor combination approaches are superior to those constructed from the top down. “When you select so-called factor ‘champions’ in a bottom-up approach, you create a concentrated portfolio with high turnover,” says EDHEC’s Felix Goltz. “But there’s a more conceptual problem as well. The academic studies on which factor investing relies all used a top-down approach. Fama and French looked at the value and size effects across the whole stock market, for example.  

“The factor relationships that were originally defined using a top-down approach may break down if we start to make too precise distinctions about factor scores at the individual stock level,” Goltz continues. “The problem about bottom-up approaches is that you may end up overexploiting the information you have in stock-level factor scores. It’s better to stick to a broad-brush approach.”

Another market participant points out that factor returns are partially dependent on each other as a benefit, not a problem. “We prefer the top-down approach to combining factors,” says Stan Verhoeven, portfolio manager at NN Investment Partners. “In our view the bottom-up approach doesn’t take individual factor risks properly into account, and it tends to ignore the potential diversification benefits. Negative correlations between factors are not a problem in our view,” he adds.

“There are sweet spots when value and momentum are aligned – when something undervalued is recovering and picks up momentum, for example. But when a value score is negative and momentum is positive, it may be a sign of the market’s overvaluation. The alignment of factors reflects our conviction level. When they are all aligned, we can have more confidence in the overall market.”

Vitaly Kalesnik, head of equity research at Research Affiliates, stresses that more complex factor construction approaches also place an extra burden on those responsible for fund governance. “In theory, smart beta should be low-cost and transparent, lowering governance costs,” says Kalesnik. The top-down factor combination approach provides better transparency.

“More sophisticated investors may prefer a more concentrated, bottom-up factor combination approach. In effect, they are happy to take the black run on the ski slope. The risk for a less sophisticated investor when following this approach is to sell out after a few years of sub-par performance and to do so at the wrong time.”

In practice, institutional investors may be using both approaches to combining factors. “The blended portfolio approach is more transparent. It’s easy to understand what the holdings are within each factor sleeve and to assess performance,” says Steve Artingstall, senior investment managers at RPMI Railpen.

“But you do get a dilution of factor intensity. The bottom-up approach gives you a more efficient portfolio with higher levels of factor exposure, but it’s a bit of a black box. And you probably need to use a single provider, whereas, with the portfolio approach, you can use the best-of-breed provider for each factor.”

“Up to now we’ve taken the former approach but we’re starting to allocate more to bottom-up multi-factor strategies,” says Artingstall.

And according to Dimitris Melas, global head of equity research at MSCI, many investors are now using bottom-up multi-factor approaches to produce customised solutions. “A lot of institutions use the bottom-up approach as a toolkit. They may wish to vary the parameters and introduce other constraints, such as a tilt towards environmental, social and governance factors,” he says.