Semester

Fall

Date of Graduation

2009

Document Type

Dissertation

Degree Type

PhD

College

Chambers College of Business and Economics

Department

Finance

Committee Chair

Ashok B. Abbott.

Abstract

This paper presents the results of time-series tests of the Capital Asset Pricing Model (CAPM) and the Fama-French 3-factor (FF3) model in the estimation of equity capital from a perspective of corporate investment decision-making. This paper addresses Fama-French (1992, "The Cross-section of Expected Stock Returns," Journal of Finance) who find that the CAPM beta has little explanatory power in cross-sectional tests of the Fama-French portfolios sorted on market capitalization and book-to-market ratios.;The main tests of this paper are (1) the equivalence of the predicted stock risk premia of the CAPM and the FF3 model, or how significantly the two model-predicted stock risk premia are correlated with the realized returns and (2) the inter-temporal and cross-sectional shift of the market beta regimes. This paper tests the CAPM and the FF3 model on daily returns of a wide range of individual stocks and Fama-French portfolios. All statistical tests are conducted on an individual stock and portfolio level.;The regression test results of the factor loadings underline the similarities and dissimilarities between the CAPM and the FF3 model applied on individual stocks and the FF portfolios. The most consistent finding throughout the tests in this paper is that the market index is the most significant risk factor loading among three risk factors (the market index, SMB, and HML) throughout the test period and across individual stocks and the Fama-French portfolios.;For individual stocks, the market index is the most significant factor loading alone or in combination with SMB and HML. The statistical explanatory power of the CAPM beta of individual stocks has inter-temporally increased, particularly for micro-cap stocks. The test of individual stocks does not reject the null hypothesis that the predicted risk permia of both models are equivalent. The overall time-series test results on individual stocks and portfolios contrast with Fama-French (1992).;On the other hand, the FF3-predicted risk premia for the FF25 portfolios largely outperform the CAPM predictions. The most intriguing finding on the risk factor loadings is that the market index rarely is significant by itself for the Fama-French 25 (FF25) portfolios within the FF3 model.;The turnabout is a stark contrast with the evidence that the market beta is more significant by itself for the majority of individual stocks in the FF3 model than in combination with two other factors.;The CAPM and the FF3 model in general are statistically equivalent in explaining individual stock returns. However, the two models are diametrically dissimilar in explaining the FF25 portfolios. The CAPM best explains large/growth (LG) portfolio returns, but is poor in explaining small/value (SV) portfolios. The high explanatory power of the CAPM for all LG portfolios is consistent, and so is the poor explanatory power for SV portfolios. The distinctive explanatory power of the CAPM between LG portfolios and SV portfolios is clear and unambiguous, and is consistent with the CAPM or the market model structure. On the other hand, the FF3 model best explains small/value (SV) portfolios, a direct opposite of the CAPM. The FF3 model is most poor in explaining large/value (LV), not LG portfolios. In the FF3 model, the "value" portfolios are at both extremes in explanatory power. Unlike the CAPM, most growth portfolios are located in the middle of the rank-order in the explanatory power of the FF3 model. The virtual diametrical dissimilarities in explanatory power between the CAPM and the FF3 model for the FF25 portfolio and the statistical equivalence of the two models for individual stocks underline the unique return factor structures of the FF portfolios and the two risk factors (SMB and HML). The superiority of the CAPM in explaining large-cap growth stocks and the inferiority of the FF3 model in explaining large-value growth may undermine the application value of the FF3 model for the estimation of the cost of equity capital.;The results of test of out-of-sample forecast using 5-year rolling regressions on monthly stock returns confirm all major findings of the test using daily data and conditioning information. (Abstract shortened by UMI.).

Download

Share

COinS