Four factors are evaluated for forecasting ability in which components are found to have predictive ability in out-of-sample return forecasts for US markets. Four other financial variables are also evaluated with conflicting results from in-sample and out-of-sample analysis.
Building on Fama and French’s work from the 1990s that explained the excess returns of small-cap and value equities, the authors assess the forecasting ability of those two factors (high minus low and small minus big) along with two other factors related to past performance—long-term reversal and momentum. In addition to decomposing each factor into size and value components, they evaluate four other financial variables—default spread, term spread, T-bill rates, and S&P 500 Index dividend yield—for their forecasting ability and potential to serve as a proxy for any of the first four factors.
How Is This Research Useful to Practitioners?
Equity asset managers could potentially benefit from higher returns after incorporating the findings of this work into their allocation decisions, and the detailed statistics may serve as an added bonus to quantitative investors. The authors present in-sample and out-of-sample predictive ability for each of the four factors and four variables on US equities over horizons from 1 to 36 months. In the in-sample analysis, they find predictive ability the strongest (at various horizons) for the value component of the small minus big portfolio, the big component of momentum, the small component of long-term reversal, and the entirety of value. In the out-of-sample analysis, the predictive ability of these factors improves.
The reverse is true for the financial variables. In the in-sample analysis, the short-term interest rate and dividend yield emerge as strong forecasters of returns, but none (except the term spread at greater than a 19-month horizon) show statistical significance in the out-of-sample analysis. They also find that the financial variables can proxy for certain size and value components of each factor, with the exception of the default spread, which seems to be a significant proxy for size, value, and momentum factors.
How Did the Authors Conduct This Research?
The authors use monthly return data from July 1963 to October 2009 taken from French’s website for each of the four-factor portfolios: (1) value, which is high book minus low book values; (2) size, which is small minus big; (3) long-term reversal, which is low price returns minus high prior returns; and (4) momentum, which is high prior returns minus low prior returns. The four financial variables include term spread, which is 10-year minus 1-year US Treasuries, and default spread, which is 10-year Baa corporate rates minus 10-year US Treasury, both of which were obtained from the FRED database. The other two variables are one-month T-bill rates and the S&P 500 dividend yield.
The authors evaluate the forecasting ability of the four factors and accompanying financial variables using an autoregressive distributed lag (ARDL) model. They test the in-sample predictive ability of each factor on the CRSP value-weighted portfolio return for horizons from 1 to 36 months, noting significance at the 10% level based on the Wald test and bootstrapped critical values. Out-of-sample tests use only information available at time T, beginning with a minimum two-thirds of the 371 available monthly observations. As is noted, and apparently common in other studies, some of the relationships from the in-sample and out-of-sample studies disappear or conflict.
Although perhaps not presented in the most succinct manner, the authors’ findings are a worthy contribution in the search for enhanced returns.