THE CAPITAL-ASSET PRICING MODEL:THE CASE OF SOUTH AFRICA By TL Reddy and RJ Thomson Announcements of the death of beta seem premature -Fischer Black Presented by Taryn Leigh Reddy, Deloitte 10 March 2010
AGENDA 1. Overview of the CAPM - The standard CAPM - The zero-beta version of the CAPM - Why test the CAPM? 2. Aim 3. Data description 4. Method 5. Preliminary observations 6. Empirical tests 7. Caveats and conclusion
OVERVIEW OF THE CAPITAL-ASSET PRICING MODEL (CAPM) Standard CAPM Sharpe and Lintner R R R R i F i M F Zero-beta version Black Why test the CAPM? If betas do not suffice to explain expected returns, the market portfolio is not efficient, and the CAPM is dead in its tracks. -Fama and French
AGENDA 1. Introduction 2. Aim 3. Data description 4. Method 5. Preliminary observations 6. Empirical tests 7. Caveats and conclusion
THE STUDY AIMED TO ANSWER A NUMBER OF QUESTIONS RELATED TO THE CAPM... Is the CAPM valid in the South African market? In particular, does the CAPM explain excess return? Is the relation between return and beta linear? Individual sectoral indices versus portfolios of sectoral indices Short term effects are relatively unimportant Are those sectors with higher systematic risk associated with higher expected return? Focus of actuarial activities will be on major sectors of the equity markets rather than on individual equities
AGENDA 1. Introduction 2. Aim 3. Data description - Sectoral indices - The market portfolio - The risk-free index 4. Method 5. Preliminary observations 6. Empirical tests 7. Caveats and conclusion
DATA DESCRIPTION Quarterly total return indices (30 June 1995 to 30 June 2009) Market portfolio Risk-free index Ten sectoral indices with the highest market capitalisations: - Basic materials (Bm) - Chemicals (Ch) - Consumer goods (Cg) - Consumer services (Cs) - Financials (Fi) - Food & drug retailers (Fd) - Health care (Hc) - Industrials (In) - Mining (Mi) and - Telecommunications (Te) The FTSE/JSE ALSI was selected as a proxy for the market portfolio The STEFI Composite Index and the Ginsberg, Malan & Carson Money -market Index
AGENDA 1. Introduction 2. Aim 3. Data description 4. Method - Individual sectoral indices - Portfolios of sectoral indices 5. Preliminary observations 6. Empirical tests 7. Caveats and conclusion
METHOD: INDIVIDUAL SECTORAL INDICES Period-by-period tests Force of return The return on that index over a unit interval (t-1,t) as: R it T it ln T it, 1 Ten one-year periods, each ending on 30 June from 2000 to 2009 Y(q) denotes the qth quarter of calendar year Y [Y] denotes the one-year period comprising the sequence of quarters: Y 1 (3), Y 1 (4), Y (1), Y (2) for Y 2000,,2009 For each sectoral index i for each period [Y], beta was estimated as: where: ˆ iy [ ] ˆ ˆ im[ Y] MM[ Y ] The excess return was determined using: The error term was found as follows: r Y (2) r R R i[ Y ] iu Fu uy1 (3) ˆ r i[ Y ] i[ Y ] i[ Y ] M[ Y ]
METHOD:INDIVIDUAL SECTORAL INDICES All periods combined Prior betas In-period betas ˆ i ˆ i[2000] i ˆ ˆ im MM r i 1 10 2009 Y =2000 r i[ Y ] r i 1 10 2009 Y =2000 r i[ Y ]
METHOD:PORTFOLIOS OF SECTORAL INDICES Period-by-period tests : For each period [Y] the sectoral indices were divided into four groups I 1[Y], I 2[Y], I 3[Y] and I 4[Y] The excess return on portfolio I p[y] during period [Y] was: r py [ ] ii py [ ] iy [ ] where k p is the number of sectoral indices included in that portfolio k p r The beta of each portfolio was calculated as follows: ˆ py [ ] ii py [ ] k p ˆ iy [ ] The error term was found as: r ˆ p[ Y ] p[ Y ] p[ Y ] M[ Y ] r
METHOD:PORTFOLIOS OF SECTORAL INDICES All periods combined Prior betas In-period betas The sectoral indices were divided into four groups I 1[2000], I 2[2000], I 3[2000] and I 4[2000] according to their beta estimates (i) p ˆ ˆ (i) pm (i) MM For portfolio I p[2000] (p=1,2,3,4), the prior beta was estimated as: ˆ (p) p ˆ p[2000] ˆ (i) pm 1 k p ii p ˆ (i) im For portfolio I p[2000] the excess annual return was determined as: where: r (p) p r (p) py [ ] 1 10 ii 2009 Y =2000 p[ 2000] k p r r iy [ ] (p) p[ Y ] The excess annual return for portfolio I p was determined as: r (i) p ii k p p r i
AGENDA 1. Introduction 2. Aim 3. Data description 4. Method 5. Preliminary observations - Sectoral indices - Portfolios of sectoral indices 6. Empirical tests 7. Caveats and conclusion
PRELIMINARY OBSERVATIONS Individual sectoral indices: Period-by-period Excess returns versus beta: [2005] Excess returns versus beta: [2006] 0,8 0,8 0,6 0,6 0,4 0,4 ri [2005] 0,2 0,0 < -0,2 0,0 0,5 1,0 1,5 2,0 ri [2006] 0,2 0,0-0,2 0,0 0,5 1,0 1,5 2,0-0,4-0,4-0,6-0,6-0,8-0,8 i [2005] i [2006] Excess returns versus beta: [2007] 0,8 0,6 0,4 ri [2007] 0,2 0,0-0,2 0,0 0,5 1,0 1,5 2,0-0,4 Evidence of linearity -0,6-0,8 i [2007]
PRELIMINARY OBSERVATIONS Individual sectoral indices: All periods combined Excess return versus beta: sectoral indices, all periods combined, prior betas 0,2 0,1 ri 0,0 0,0 0,5 1,0 1,5-0,1-0,2 i The effects of outliers and wide scattering are widely diluted and the relationship appears more linear
PRELIMINARY OBSERVATIONS Individual sectoral indices: All periods combined Excess return versus beta: sectoral indices, all periods combined, in-period betas 0,2 0,1 The scatter here is wider for lower betas and the relationship appears less linear ri 0,0 0,0 0,5 1,0 1,5-0,1-0,2 i
PRELIMINARY OBSERVATIONS Portfolios of sectoral indices: All periods combined Excess return versus beta: portfolios, all periods combined, prior betas 0,2 Excess return versus beta: portfolios, all periods combined, in-period betas 0,2 0,1 0,1 rp 0,0 0,0 0,5 1,0 1,5 rp 0,0 0,0 0,5 1,0 1,5-0,1-0,1-0,2-0,2 p p The securities market line is flatter than that of the standard CAPM, it would be consistent with the zero-beta version A general reduction in in-period betas relative to prior betas
AGENDA 1. Introduction 2. Aim 3. Data description 4. Method 5. Preliminary observations 6. Empirical tests - Tests of the mean of error terms - Sectoral indices - Portfolios of sectoral indices - Summary 7. Caveats and conclusion
HISTOGRAM OF SAMPLE DISTRIBUTION OF ERROR TERMS Histogram of the sample distribution of iy [ ] 45 40 35 30 25 20 15 10 5 0 The scatter here is wider for lower betas and the relationship appears less linear -0.8-0.6-0.4-0.2 0 0.2 0.4 0.6 0.8 More epsilon The distribution apparently deviates from the normal and its mean appears to be greater than 0
HYPOTHESIS: ε i[y] is normally distributed with a mean of 0 The emphasis must be placed on inperiod tests rather than on tests using prior betas Parametric test using prior betas Test statistic: S S 2009 Y 2000 S ε V [ Y ] 1 [ Y ] [ Y ] [ Y ] [ Y ] Parametric test using inperiod betas Test statistic: Shanken s regression statistic QC ˆ 1 Te Σ e follows a Hotelling s T 2 with N-2 = n and T-2 = m degrees of freedom Non-parametric test Test statistic: The observed values can be grouped as ε i[y] > 0 and ε i[y] > median o j e G e j j 2 On the assumption that the error term is normally distributed, the hypothesis is rejected On the assumption that the error term is normally distributed, the hypothesis could not be rejected We accept that the mean of ε i[y] is zero, we cannot reject the CAPM
WHY WERE THE TESTS PERFORMED IN THIS STUDY BETTER? Direct tests of the securities market line Test the CAPM using the following: E Ri R E R R where: F i M F i im MM Tests of the return relative to beta The following is tested : i 0 1 E R The null hypothesis for a given period t may be expressed as: r it 0 1 it it where : r R R it it Ft i Implies that ex-post estimates are unbiased estimates of ex-ante expectations (implication of the REH) It does not permit the testing of the zero-beta version Reduces the effect of the REH Permits testing of zero-beta version These tests do not use ex-post expected values of the return
INDIVIDUAL SECTORAL INDICES Test of the explanatory power of the CAPM Test for nonlinearity Period-by-period test r ˆ i[ Y ] 0 1 i[ Y ] i[ Y ] Period-by-period test r ˆ ˆ 2 i[ Y ] 0 1 i[ Y ] 2 i[ Y ] i[ Y ] All periods combined Prior betas r ˆ i 0 1 i i In-period betas r i 0 1 i i What do we want to see? Standard version 0 0 1 0 Zero-beta version 0 0 1 0 All periods combined Prior betas ˆ ˆ In period betas What do we want to see? Standard version 2 0 2 r i 0 1 i 2 i i 2 r i 0 1 i 2 i i Zero-beta version 2 0
The values of R 2 for most of the periods are low A TEST OF THE EXPLANATORY POWER OF THE CAPM Summary of regression analysis of excess returns on sectoral indices R 2 Period Estimate Value t-value p-value 0 [2000] 0,012 0,017 0,048 0,963 1 0,108 0,317 0,759 [2001] 0,593 0 0,674 3,275 0,011 1 0,656 3,411 0,009 [2002] 0,503 0 0,690 3,068 0,015 1 0,615 2,845 0,022 [2003] 0,260 0 0,390 1,450 0,185 1 0,207 0,705 0,501 [2004] 0,019 0 0,278 2,049 0,075 1 0,062 0,395 0,703 [2005] 0,150 0 0,365 5,591 0,001 1 0,091 1,186 0,270 [2006] 0,160 0 0,173 1,487 0,175 1 0,170 1,236 0,252 [2007] 0,018 0 0,312 2,573 0,033 1 0,048 0,378 0,716 [2008] 0,003 0 0,193 0,664 0,526 1 0,050 0,163 0,874 [2009] 0,114 0 0,424 1,516 0,168 1 0,252 1,016 0,339 all: prior 0,001 0 0,079 1,426 0,192 betas 1 0,003 0,064 0,950 all: in- 0,088 0 0,035 0,732 0,485 period 0,051 0,406 0,406 betas 1 We cannot reject the hypothesis either for the standard or zero-beta version of the CAPM for all periods combined
TEST FOR NONLINEARITY Summary of regression analysis of nonlinearity for sectoral indices Period t-value p-value [2000] 0,622 3,119 3,358 0,012 [2001] 0,825 1,267 3,045 0,019 [2002] 0,528 0,500 0,614 0,559 [2003] 0,579 1 570 2,942 0,022 [2004] 0,098 0,343 0,782 0,460 [2005] 0,333 0,348 1,387 0,208 [2006] 0,164 0,090 0,164 0,875 [2007] 0,065 0,227 0,593 0,572 [2008] 0,579 0,039 0,042 0,967 all: prior betas 0,064 0,156 0,687 0,514 all: in-period betas The value of R 2 increases for each period ˆ 2 R 2 0,113 0,108 0,447 0,669
PORTFOLIOS OF SECTORAL INDICES Test of the explanatory power of the CAPM Test for nonlinearity Period-by-period test r ˆ p[ Y ] 0 1 p[ Y ] p[ Y ] Period-by-period test r ˆ ˆ 2 p[ Y ] 0 1 p[ Y ] 2 p[ Y ] p[ Y ] All periods combined Prior betas p r ( p) ˆ ( ) p 0 1 p p In-period betas i r ( i) ( ) p 0 1 p p What do we want to see? Standard version 0 0 1 0 Zero-beta version 0 0 1 0 All periods combined Prior betas ˆ ˆ 2 r ( ( ( p 0 1 p 2 p p In period betas 2 r ( i) ( i) ( i) p 0 1 p 2 p p What do we want to see? Standard version 2 0 Zero-beta version 2 0
The values of R 2 for most of the periods are low TEST OF THE EXPLANATORY POWER OF THE CAPM Summary of regression analysis of excess return on portfolios R 2 Period Estimate Value t-value p-value 0 [2000] 0,022 0,123 0,541 0,643 1 0,047 0,210 0,853 [2001] 0,532 0 0,684 1,541 0,263 1 0,633 1,506 0,271 [2002] 0,112 0 0,770 3,340 0,079 1 0,692 3,046 0,093 [2003] 0,811 0 0,574 4,492 0,046 1 0,413 2,928 0,100 [2004] 0,151 0 0,319 2,062 0,175 1 0,109 0,596 0,612 [2005] 0,431 0 0,378 5,047 0,037 1 0,113 1,232 0,343 [2006] 0,932 0 0,162 5,741 0,029 1 0,180 5,226 0,035 [2007] 0,263 0 0,401 2,650 0,118 1 0,136 0,844 0,488 [2008] 0,105 0 0,393 0,854 0,483 1 0,240 0,485 0,676 [2009] 0,384 0 0,376 1,871 0,202 1 0,203 1,116 0,380 all: prior 0,417 0 0,022 0,474 0,682 betas 1 0,054 1,196 0,354 all: in- 0,415 0 0,026 0,601 0,609 period 0,064 1,190 0,356 betas 1 We cannot reject the hypothesis either for the standard or zero-beta version of the CAPM
TEST FOR NONLINEARITY Summary of regression analysis of nonlinearity for portfolios 2 Period t-value p-value [2000] 0,998 1,556 22,410 0,028 [2001] 0,999 2,539 21,570 0,030 [2002] 0,922 1,088 1,132 0,461 [2003] 0,854 0,448 0,547 0,681 [2004] 0,228 0,397 0,316 0,805 [2005] 0,779 0,480 1,255 0,428 [2006] 0,962 0,174 0,900 0,533 [2007] 0,908 1,281 2,640 0,230 [2008] 0,823 3,383 2,014 0,293 [2009] 0,384 0,028 0,028 0,982 all: prior betas 0,446 0,071 0,230 0,856 all: in-period betas The value of R 2 increases for each period ˆ R 2 0,489 0,165 0,380 0,769 We cannot reject the hypothesis that the relation between return and beta is linear
SUMMARY OF TESTS ON SECTORAL INDICES AND PORTFOLIOS OF SECTORAL INDICES Ex-post versus ex-ante Parametric and nonparametric tests use ex-post estimations of expected market return, these tests do not All periods test For all periods combined it was not possible to reject either version of the CAPM Summary Period-by-period tests The CAPM could be rejected for some periods, but not for all Therefore On the basis of the tests using in-period betas, the CAPM cannot be rejected
AGENDA 1. Introduction 2. Aim 3. Data description 4. Method 5. Preliminary observations 6. Empirical tests 7. Caveats and conclusions - Caveats - Further research - Conclusion
CAVEATS The future will not necessarily be the same as the past Possibly insufficient data Caveats Many simplifying assumptions Market portfolio is unobservable in practice
FURTHER RESEARCH Actuarial applications will generally require at least the inclusion of bonds in the market portfolio Inclusion of bonds Real terms vs nominal terms Most tests of the CAPM are applied in nominal terms Exogenous variables tend to become irrelevant after the first few projection periods Exogenous variables
CONCLUSION While on the assumption that the residuals are normally distributed, the CAPM could be rejected for certain periods While the zero-beta version better supports the relatively flat securities market line, the standard version is not generally rejected The study showed that... The use of the model for long-term actuarial modelling in the South African market can be reasonably justified Announcements of the death of beta seem premature -Fischer Black
CONTACT DETAILS Name Taryn Leigh Reddy Company name Deloitte Telephone number + 27 83 610-2724 + 27 11 209-8640 Email address tareddy@deloitte.co.za