Henric von der Groeben och Martin Torell

Deriving Explanatory Factors for Stock Returns
Using Multi-Factor Cross-Sectional Regression and Mimicking Portfolios


Abstract:

Fundamental and macroeconomic factors seem to influence stock returns.
In this paper we focus on finding the most important explanatory
factors for returns on the U.S. stock market between 1989 and 2002. We
have used two methods for this analysis: a multi-factor
cross-sectional regression model and a model for mimicking
portfolios. The cross-sectional regression measures the variation of
factors across different stocks at a single point in time.  For the
mimicking portfolios approach, companies are sorted into two
portfolios based on the value of a factor, and the spread is
evaluated.  The two methods, though conceptually very different,
support the overall results of one another. In order to evaluate if
factors have a varying influence during different market conditions,
we have divided the sample-period into three sub-periods.
Additionally, we have tested our results out-of-sample.  Our main
conclusion is that market capitalization (Size), cash flow to market
price (CF/P), and book value of equity to market price (B/P) are the
most useful factors for explaining and predicting stock returns.