Investigating value and growth : what labels hide ?

Value and growth investment styles are a concept which has gained extreme popularity over the past two decades, probably due to its practical efficiency and relative simplicity. We study the mechanics of different factors' impact on excess returns in a multivariate setting. We use a panel of stock returns and accounting data from 1979 to 2007 for the companies listed on NYSE without survivor bias for clustering, regression analysis and constructing style based portfolios. Our findings suggest that value and growth labels often hide important heterogeneity of the underlying sources of risks. Many variables, conventionally used for style definitions, cannot be used jointly, because they affect returns in opposite directions. A simple truth that more variables does not necessarily mean better model nicely summaries our results. We advocate a more flexible approach to analyzing accounting-based factors of outperformance treating them separately before or instead of aggregating.