LECTURE 9 EMPRICIAL EVIDENCE CAPM AND APT.ppt

LECTURE 9 : EMPRICIAL EVIDENCE : CAPM AND APT (Asset Pricing and Portfolio Theory) Contents How can the CAPM be tested by using time series approach by using cross sectional approach Fama and MacBeth (1973) approach CAPM : Time Series Tests (ERi – rf)t = ai + bi(ERm – rf)t + eit Null hypothesis : ai = 0 Assume individual returns are iid, but allow contemporaneous correlation across assets : E(eit, ejt) ≠ 0 Estimation : maximum likelihood using panel data (N assets, T time periods) Early studies in the 1970s find ai = 0, but later studies find the opposite. CAPM : Cross-Section Tests Useing the Security Market Line (SML) Hypothesis : average returns (in a cross section of stocks) depend linearly (and solely) on asset betas Problems : Individual stock returns are so volatile that one cannot reject the null hypothesis that average returns across different stocks are the same (s ≈ 30 – 80% p.a., hence cannot reject the null that average returns across different stocks are the same.) Betas are measured with errors (to overcome these problems form portfolios) CAPM : Cross-Section Tests (Cont.) Two Stage procedure 1st stage : Time series regression, estimation of betas (ERi – rf)t = ai + bi(ERm – rf)t + eit (Time series regression has to be repeated for each fund, using ‘say’ 60 observations.) 2nd stage : ERi = l0 + l1 bhati + ui Cochane (2001) Testing the CAPM Sort all stocks on the NYSE into 10 portfolios on basis of size, plus 2 bond portfolios 1st stage : Estimating the betas for each of the 12 portfolios using time series data 2nd stage : sample average returns regressed against the 12 estimated betas CAPM seems to do a reasonable ‘job’. (but if sorted by book to market value, decile returns are not explained by market beta). Results : see graph on next slide Size-Sorted Value-Weighted Decile Portfolio (NYSE – from 1947) Fama and MacBeth (1973) Fama and MacBeth Rolling Regression Much used regression methodology that involves ‘ro