Using Risk Factors to Understand Long/Short Equity Mutual Fund Returns: Part 1

In this article and the next two, we examine the risk factors that help explain long/short equity (LSE) mutual fund performance. We show that for most LSE mutual funds, 50%-80% of their returns can be explained using common factors such as capitalization, book-to-value ratio, dividend yield, and volatility. The explanatory strength of these factors is so strong  that in most cases  adding an option overlay using puts and calls (to mimic long and short positions) adds little or no explanatory value. We also show that while many funds have positive alpha, their alphas for the most part are not statistically significant, i.e., in reality they are not different than zero.

Factor Exposure

The best known work on using factors to explain stock returns is a series of articles by Fama and French[1]. The economic thought behind the use of factors is to use them to reduce the difficulty of identifying skilled managers. The genesis of this solution is the capital asset pricing model (CAPM). A substantial improvement is achieved with the Fama-French factors. Their model takes into account (1) market beta, and exposure to factor “portfolios” consisting of (2) size: the return on small-cap stocks less the return on large-cap stocks and (3) book-to-value ratio: the return on high book-to-market stocks less the return on low book-to-market stocks. The Fama-French model (also known as “the three-factor model”) shows that many long-only stock portfolio returns can be explained in large part by using the three-factor model. The Fama-French model can also determine if there is a statistically significant alpha—a key factor in determining skilled managers. Finally, the analysis of portfolio returns using factors is so effective it can be said that portfolio returns, regardless of country, have yielded their “secrets” to factor models.

The factor exposures we use here are the traditional Fama-French factors (size and book-to-market) as well as a dividend-yield factor (the return on high-dividend-yield stocks less the return on low-dividend-yield stocks) and a volatility factor (the return on high-volatility stocks less the return on low-volatility stocks). We include a dividend-yield factor and volatility factor (which Amenc et al.[2] also did in their “smart beta” funds study) to capture any dividend tilts or volatility tilts that maybe present.

Other authors have used factor models to understand LSE funds, but their work has been focused on LSE hedge funds, not LSE mutual funds. Chen and Passow[3] and Agarwal and Naik[4] are authors of works on LSE hedge funds that we will refer to in this set of articles.

Data

The LSE mutual funds we examine are all the primary share classes that have five years of weekly returns covering the period January2009 through December 2013.  The LSE fund source is Lipper for Investment Management (LIM). There are 45 primary share class LSE funds with the required five years of weekly returns. The Fama-French factors (including Treasury bill rates) come from French’s Web site[5], while the dividend and volatility factors are constructed using the Russell 3000 index. We use the Russell 3000 because we agree with Chen and Passow that it is the better benchmark for LSE funds.

As a side note, following Agarwal and Naik (along with Chen and Passow), we tested the Goldman Sachs Commodity Index as a factor. In addition we tested the “optionality” of LSE funds by using at-the-money and out-of-the-money calls and puts. Neither factor added value to our understanding of LSE returns. We will discuss the implications of these results in the conclusion of this series of articles.

Methodology

Typically, a form of linear regression, such as ordinary least squares (OLS), is used to test factor models. Sometimes a more complicated model such as generalized method of moments (GMM) is used when the investigator wants to account for the well-known facts that financial data more often than not exhibits a variance that changes over time as well as a certain amount of serial correlation. In these articles we choose a third regression method that takes into account both the linear and nonlinear influences a factor may have on a particular fund’s returns. This third regression method, like GMM, takes into account any potential time-varying variance and correlation properties noted above.

In our next article we will show the results of our factor analysis.