Diversifying factor exposures in Systematic Equity Portfolios

Quantitative equity managers seek to diversify among various shared traits that drive stock returns, such as valuation and momentum. The number of these traits, known as factors or signals, that managers consider ranges widely; our research suggests surprisingly few may be needed.

There are few financial truisms as widely embraced as the free lunch in diversification—that, all else being equal, adding sources of return reduces portfolio performance swings, and thereby improves risk-adjusted return over time.

Many investors today are interested in data-driven approaches employed by systematic equity managers. These managers seek to diversify exposure to equity factors. Equity factors are common traits that drive stock behavior, or underlying signals of stock direction based on company fundamentals or investor sentiment. Such factors are the building blocks of quantitative stock selection models built and utilized by systematic equity managers, such as ourselves, to identify stocks for their portfolios. Equity factors are categorized into larger buckets, or model composites, such as valuation and momentum.

But how many individual factors are really needed? Quantitative equity research is ever evolving and new factorsare continuously identified. Most systematic equity managers build portfolios by utilizing typically between 10 and 20 factors, while a few consider more than 70 factors.

While it may be intuitively appealing to pile on ever more factors, we wanted to analyze whether doing so would result in improved performance. Diversification is only valuable up to a series of points, known as the efficient frontier. After that, adding another source of return (adding an additional factor) can potentially reduce performance, especially after accounting for the direct and indirect costs of increased trading activity.

In our analysis, we specifically looked at commonly used return factors—factors which are drivers of expected returns over a future time period. The factors we chose for this study are what we consider to be “distinct,” that is less than perfectly correlated to each other. Our main criteria for selection in the study was choosing factors that enhance overall performance, with risk adjusted performance as a secondary criterion.

Key Findings

Our research suggests that:

- The optimal number of factors to employ is surprisingly low. This is because many factors are actually quite similar, and the investment benefit of using additional factors that are similar for all stocks is insignificant or even negative (Figure 1).

- There is, however, a significant investment benefit in incorporating additional factors in a contextual approach. This means incorporating additional factors to the specific stocks for which they are more relevant.


A Few Are Plenty

A systematic approach to equity investing analyzes companies broadly. Rather than primarily focusing on industry trends and individual companies as fundamental analysts do, systematic managers look at equity factors, with each one providing another lens through which managers gain insights into a stock’s risk and potential return. New factors can offer newand interesting insights, but also add complexity to the process.

Our first exercise in this study tested the historical efficacy of adding factors to an investment process. We looked at the marginal benefit of adding successive factors to the management of a U.S. stock portfolio, based on the monthly excess returns (or alpha), the risk of 10 common factors, and the correlations among them, since 1995. We first added the factor that most improved portfolio performance versus the S&P 500 Index, then the one with the second-best impact, andso on, up to 10. Total alpha nearly peaked after just five factors, far fewer than what we believe most managers use today (see Figure 1).

Why should this be? More often than realized, quantitative research identifies new factors that are merely shades or formulations of well-known factors. Just as coral, burgundy and maroon are shades of red, there are many shades of valuation and momentum. The dimensionality of factors employed, the breadth and depth of the factors, is usually far less than the number of factors may seem to indicate.

For the full version of this paper, please click here or use download function on the upper right of this page.