Table 9 Split Sample Cross Sectional Regression Results
Cross sectional regression results for tests of the importance of the market portfolio and predicted returns based on lagged macroeconomic indicators on S&P 500 sector portfolio returns are summarized for the period of October 14, 1994 to March 10, 2000 and from March 11, 2000 to January 10, 2003. The OLS regression coefficients are expressed as percentage per week. For each crosssectional regression at time t, The regressors are a constant term, the betas from a contemporaneous time series regression on the return on the S&P 500 index at time t1 and a forecasted return estimated with lagged macroeconomic factors through period t1. Rolling CAPM betas and predicted returns are estimated using the most recent 50 weeks of data. Bayesian betas and predicted returns are estimated in the time varying parameter model described in section 2. The number of cross sectional regressions is 280 in the first sample and 149 in the second sample. FamaMacBeth standard errors are underneath the coefficients in parentheses.
0
1 (market)
2 (macro)
0.0177
0.3364

(0.1460)
(0.1797)

Sample I: October 14, 1994 to March 10, 2000 Panel A. Rolling CAPM Betas and Rolling Parameter Predicted Returns
0.1048
0.3813
0.2320
(0.1556)
(0.2057)
(0.1281)
0
1 (market)
2 (macro)
0.0947
0.4419

(0.2018)
(0.2452)

0.0219
0.3167
0.3769
(0.2121)
(0.2770)
(0.3428)
Panel B. Bayesian CAPM Betas and Bayesian Predicted Returns
0
(market)
1
(macro)
2
0.0383 (0.2751)
0.2359 (0.3570)
 
0.0267
0.4202
0.2697
(0.2952)
(0.3757)
(0.1460)
Sample II: March 11, 2000 to January 10, 2003 Panel C. Rolling CAPM Betas and Rolling Predicted Returns
0
1 (market)
2 (macro)
0.0383
0.2814

(0.2985)
(0.3607)

0.1125
0.5648
0.8515
(0.3029)
(0.3881)
(0.4062)
Panel D. Bayesian CAPM Betas and Bayesian Predicted Returns
39