Content area
Full Text
ABSTRACT
The goal of this study is to compare the CAPM to the Fama-French (FF) Three Factor Model and to Carhart's extension of the FF Model with regard to (1) statistical goodness of fit, and (2) the quality of prediction. My sample consists of actively managed domestic equity mutual funds and the sample period is April 1986 to March 2006. My results indicate that each of the three regression lines explains about 71% of equity fund returns. Thus, with respect to the statistical goodness of fit, the difference between the three models is not significant. However, with respect to the quality of prediction, the FF Three Factor Model is a remarkable improvement over the CAPM, and the Carhart Model is a significant improvement over the FF Model. I do not find any evidence of harmful collinearity in my analyses.
Key words: Asset Pricing Models, Statistical Goodness of Fit, Model Specification, Multicollinearity
JEL Classification: G11, G12, C12, C13, C25.
(ProQuest: ... denotes formulae omitted.)
I. INTRODUCTION
"The goal of statistical model specification is to assist in understanding the system from which the data is extracted, to learn which regression terms are important, and to learn which model is the best in terms of prediction" and that "an adroit job of model selection does not require that one actually locate the correct model, and that the correct model may indeed never be found." (Myers, 1990).
The purpose of this study is to compare the specification of three related asset pricing models in terms of (1) the statistical goodness of fit, and (2) the qualities of prediction. I use several traditional methods of model selection and nontraditional model selection criteria which are prediction oriented. The asset pricing models that I focus on are those that specify risk in microeconomic terms, i.e. those that use the characteristics of the underlying sample of securities. Connor (1995) compares the explanatory power of the three types of multifactor models of asset returns, including a macroeconomic factor model, a fundamental factor model, and a statistical factor model. He finds that the statistical factor model and the fundamental factor model substantially outperform the macroeconomic factor model, and that the fundamental factor model outperforms the statistical factor model. I concentrate on the three...