Supplemental material for Asset Pricing and Excess Returns over the Market Return Seung C. Ahn Arizona State University Alex R. Horenstein University of Miami This documents contains an additional figure and five tables whose results are discussed in the paper but are not part of it.
Alternative Figure 3: Distributions of s from Regressions With and The market betas are estimated using 33 portfolio returns over the one-month treasury bill rate (RF-excess returns) from January 97 to December 3. The valueweighted portfolio returns, the one-month treasury bill rate, the five factors of Fama and French (5), and the 33 test portfolio returns are from Kenneth French s webpage. The 33 portfolios are 5 Size and Book to Market, 5 Size and Investment, 5 Size and Operating Profitability, 5 Book to Market and Investment, 5 Book to Market and Operating Profitability, 5 Investment and Operating Profitability, 3 Industry, Residual Variance, Variance, et Share Issues, Market Beta, Accruals, Long-term Reversal, Short-term Reversal, Momentum, Dividend Yield, Cash Flow to Price, Earnings to Price, Size, Book to Market, Investment, and Operating Profitability. The five VW-PC factors are obtained from the VW-excess returns on these 33 portfolios. The market betas (the betas of the RF-excess return on the VW portfolio and the raw return on the VW portfolio ) are estimated by three different factor models: the basic (which uses the MKT factor only), the five-factor model of Fama and French (5), and the augmented with the five VW-PC factors. (a) (b) Fama-French Five-Factor model (c) Augmented with five VW-PCs.8.6.8.6.8.6.4.4.4.....4.6.8..4.6.8..4.6.8
Alternative Table : Summary Statistics from s using Raw Returns and o-intercept The market betas (the betas of the MKT factor) are estimated from three different models without intercept: the (one factor model with the MKT factor), the five-factor model of Fama and French (5, FF5), and the augmented with the five VW-PC factors ( plus VW-PC5). The data used for this table are the same as those used for Table in the main body of the paper. For each model, the response variables are raw returns on individual stocks or the 8 tests portfolios. The five VW-PC factors are extracted from the 5 base portfolios as in Table in the main body of the paper. Four different sample periods are considered. Panel (a) reports the rejection frequencies of the hypothesis that the market beta equals one ( ) for an individual asset i at a 5% significance level. The VW, i t-statistics are computed with the White heteroskedasticity robust OLS standard errors. Panels (b) and (c) report the magnitude of the dispersions (cross-sectional standard deviations) of the estimated market betas and the root mean square errors that are computed by b ˆ. VW i VW, i ( ˆ b ) and i VW, i VW ( ˆ ), respectively, where i VW, i T FF5 Panel (a) Rejection frequency of the hypothesis that market beta equal (at 5% level) plus VW-PC5 Individual Stock Returns 97-98 4 3 37% 4% % Individual Stock Returns 98-99 856 3 35% 8% 8% Individual Stock Returns 99-877 3 5% 7% % Individual Stock Returns 3-3 6 3 4% 3% % Test Portfolio Returns 97-98 8 3 49% 3% 5% Test Portfolio Returns 98-99 8 3 58% 54% 39% Test Portfolio Returns 99-8 3 69% 37% 3% Test Portfolio Returns 3-3 8 3 67% 48% 5% Panel (b) Dispersion (standard deviation) of the estimated market betas Individual Stock Returns 97-98 4 3.39.3.3 Individual Stock Returns 98-99 856 3.38.37.37 Individual Stock Returns 99-877 3.58.53 7 Individual Stock Returns 3-3 6 3.56.5 7 Test Portfolio Returns 97-98 8 3..9 Test Portfolio Returns 98-99 8 3..8 Test Portfolio Returns 99-8 3 9. Test Portfolio Returns 3-3 8 3 5 Panel (c) Root Mean Square Error Individual Stock Returns 97-98 4 3 5.33.3 Individual Stock Returns 98-99 856 3.38.37.37 Individual Stock Returns 99-877 3.53 7 Individual Stock Returns 3-3 6 3.5 7 Test Portfolio Returns 97-98 8 3.9.9 Test Portfolio Returns 98-99 8 3.5.9 Test Portfolio Returns 99-8 3.3.7. Test Portfolio Returns 3-3 8 3 7 3
Alternative Table 3 using EW-excess returns: Frequencies of Significant Pricing Errors (at a 5% Significance Level) This table reports the rejection frequencies of the hypothesis of no pricing error ( i ) for an individual asset i at a 5% significance level. The pricing errors of individual assets are estimated for five different models: the, the five-factor model of Fama and French (FF5), the model of one single EW-PC factor (EW-PC), the model of five EW-PC factors (EW-PC5), and, finally, the augmented with the five EW-PC factors ( ). Each model is estimated using the EW-excess returns as response variables. The standard errors of the estimated pricing errors are obtained using the White heteroskedasticity robust OLS variance matrices. The hypothesis of no pricing error was tested for each asset by the usual t-statistic. All of the results are obtained using the same data that are used for Table. T FF5 EW-PC EW-PC5 Individual Stocks Returns 97-98 4 3 4.7% 3.5% 8.8%.8%.3% Individual Stocks Returns 98-99 856 3.7% 7.6%.%.%.3% Individual Stocks Returns 99-877 3 5.9%.6% 6.6% 4.%.3% Individual Stocks Returns 3-3 6 3 6.8% 9.3% 7.7% 8.% 9.3% Test Portfolios Returns 97-98 8 3 3.8% 5.% 7.% % % Test Portfolios Returns 98-99 8 3 8.3% 9.4% 6.%.8%.% Test Portfolios Returns 99-8 3.8% 5.6%.7% 3.%.% Test Portfolios Returns 3-3 8 3 5.% 3% 8.3% 8.3% 5.% Alternative Table 4 using EW-excess returns: Annualized Average Absolute Pricing Errors This table reports the estimated annualized average absolute pricing error of each of the five different models considered in Table 4. All of the results are obtained using the same data that are used for Table. T FF5 EW-PC EW-PC5 Individual Stocks Returns 97-98 4 3.7.89.74.8.79 Individual Stocks Returns 98-99 856 3.4.3.9.93.94 Individual Stocks Returns 99-877 3.98.9.99.7 Individual Stocks Returns 3-3 6 3.85.96.88.89.9 Test Portfolios Returns 97-98 8 3.5.3.5.. Test Portfolios Returns 98-99 8 3.45.9.3.. Test Portfolios Returns 99-8 3.36.45.33.3.8 Test Portfolios Returns 3-3 8 3.8.35.7.4.8 4
Alternative Table 5 using EW-excess returns: Correlations between Average and Predicted Expected Returns I This table reports the correlations between (ex post) average returns and predicted expected returns on individual assets. The expected returns are predicted by the same models that are used for Table 4. For each model, the predicted expected return on asset i is obtained by r ˆ f, where is the mean return on the EW portfolio, ˆi EW i f is the vector of mean factors, and ˆi is the estimated factor beta for asset i. All of the results are obtained using the same data that are used for Table. r EW T FF5 EW-PC EW-PC5 Individual Stocks Returns 97-98 4 3 7 -.4.95 -.3.39 Individual Stocks Returns 98-99 856 3.3.94 59 59 96 Individual Stocks Returns 99-877 3 -.7.34 -.34 -.55 Individual Stocks Returns 3-3 6 3.33.77.336 6.33 Test Portfolios Returns 97-98 8 3.93.3.9 7 84 Test Portfolios Returns 98-99 8 3 -.35 98 3 4 49 Test Portfolios Returns 99-8 3-53 85-9.33.39 Test Portfolios Returns 3-3 8 3.357.34 43 9.58 Alternative Table 6 using EW-excess returns: Correlations between Average and Predicted Expected Returns II This table reports the correlations between average (ex post) returns and predicted expected returns on individual assets. The expected returns are predicted by the same five models that are used for Table 4. For each model, the predicted expected return on asset i is = r ˆ ˆ, where is the mean return on the EW portfolio, is the estimated factor beta, and ˆ f same data that are used for Table. 5 i EW i f is the estimated risk premium vector by the Fama-MacBeth (973) cross-sectional regression. All of the results are obtained using the T FF5 EW-PC EW-PC5 Individual Stocks Returns 97-98 4 3 7 4.95 89 95 Individual Stocks Returns 98-99 856 3.3.34 59 5 Individual Stocks Returns 99-877 3 6.34 9 4 Individual Stocks Returns 3-3 6 3.33 33.336 4 Test Portfolios Returns 97-98 8 3.93 98.9.59 35 Test Portfolios Returns 98-99 8 3.35 75 3.744.746 Test Portfolios Returns 99-8 3 53.343 9.395.56 Test Portfolios Returns 3-3 8 3.357.585 43 37 39 r EW ˆi