Blending Quantitative and Traditional Equity Analysis: An Overview

A colleague once pondered, "If everyone has the algorithms, will they not get the same answers?" This question reflects an important observation and is one of the reasons for this conference-Blending Quantitative and Traditional Equity Analysis-at this time. Consider three advances in quantitative modeling that have emerged recently: nonlinear estimation, conditional forecasting, and improved holdout sampling. Neural nets, classification trees, and pattern recognition are nonlinear estimation procedures. Chaos theory highlights nonstationarity, in which conditional forecasting works. Data mining (i.e., searching a data base repeatedly until you uncover some pattern from the past) focuses attention on holdout sampling. The increasing use of the new quantitative techniques introduces the dilemma posed in my colleague's question: As nonlinear and conditional forecasting algorithms are used more and more, will not everyone obtain the same answers from the same data samples? Because of the dilemma brought about by the quantitative advances, traditional investment analysis assumes a monumental role and the need for blending the two approaches becomes clear. Profitable research requires three additions that traditional analysis can provide quantitative analysis: (1) isolation of homogeneous groups of companies (value versus growth, large versus small, utilities versus high-technology firms), (2) isolation of different market environments (1988 for value, 1989 for growth), and (3) assessment of whether empirical findings coincide with both theory and "Street smarts." This last effect is especially important when statistical significance cannot be achieved because of either small sample size or data mining.