Growth and Value Stock Indices under the EDA Microscope

Abstract

In this paper, the author uses geometrical and topological aspects of Exploratory Data Analysis (EDA) to examine Standard and Poor’s (S&P), MSCI’s and Thomson Reuters’ (TRI) ways of determining which stocks are growth and which are value. The results of the analysis are that two of the firms - S&P and TRI - do a very good job of determining what S&P calls “pure growth” and “pure value” stocks. This is found to be true to a large extent because their respective point cloud data tends to fall into linear clusters and hence is amenable to their respective linear separation techniques. A lack of linearity in MSCI’s point cloud data is more often than not foiling their linear methodologies so “the jury is still out” on MSCI’s ability to effectively separate growth and values stocks. The techniques used to arrive at these conclusions include the use of ggobi (a “grand tour” data visualization system), isomaps (a nonlinear data reduction tool), model-based clustering and multiresolution bootstrap resampling. The paper also shows that the “core” designation that has been put forward by such firms as Lipper and Morningstar and individuals such as Ron Surz is a true classification, at least if one uses TRI’s fundamental factors to separate growth and value. As to correctly identifying the value effect, i.e., those times when the Fama-French value benchmark outperforms its growth benchmark, only TRI growth and value indices consistently demonstrate the value effect