You are on page 1of 17

# A Critical Evaluation of the use of CAPM and Beta If there is one lesson that the investment market has

learned over the last 4 years, probably more than any other time in their recent history is that financial models used to reduce risk in forecasting share performances is risky business.

When Markowitz (Markowitz, 1952) first suggested through his portfolio theory that investors could select a portfolio of companies that would return a high likelihood of expected return for a known amount of risk, it spawned an enormous interest from academia and the investor market as a whole.

Capital Asset Pricing Model (CAPM) emerged from the work of William Sharpe (Sharpe, 1964) who finally published his paper introducing the risk free rate as an additional element to portfolio theory in 1964 from a paper written in 1962. When Lintner (Lintner , 1965) and Mossin (Mossin ,1966) independently arrived at similar conclusions to Sharpe it helped establish CAPM’s acceptance in economic theory circles as a major breakthrough to managing market risk through a diversified portfolio, incredibly earning Markowitz and Sharpe a Nobel Memorial Prize in Economic Sciences in 1990.

CAPm is a model to calculate the expected return on a capital asset in relation to a related market, a risk free rate and the Beta (β) (Beta is an individual relationship of the security market line (SML) risk and expected return of the asset).

To demonstrate the use of this model and to comment on the statistical information extracted from the raw data, I have taken five years of monthly share values of HSBC PLC and the FTSE All share index. To ensure that I follow the CAPm requirement for a geometric mean I have

Table of Share prices and Calculated Natural Logs .Converted the arithmetical share values to geometric by the use of natural logs (figure 1). Figure 1.

0909356250 Sum Statistic -.3936374000 .1976959400 -.Descriptives Descriptive Statistics N Statistic RF RM 59 59 Range Statistic .1441181700 Maximum Statistic .2350537950 Minimum Statistic -.4950981500 -.1274840621 .1959414600 .

002 59 1 .143 . (2-tailed) Sum of Squares and Cross-products Covariance N RM Pearson Correlation Sig.Descriptive Statistics N Statistic RF RM Valid N (listwise) Descriptive Statistics Mean Statistic -. Error .149 .633** .613 .1441181700 Maximum Statistic .01 level (2-tailed).506 Std.0065947846935 Std.311 Statistic .224 -.006 59 .006 .2350537950 Minimum Statistic -.0100262307260 .232 Kurtosis Std. Error .002 59 .311 .0909356250 Sum Statistic -.008391494068 -.0506555024063 Variance Statistic .1274840621 Correlations Correlations RF RF Pearson Correlation Sig.002160746815 Std. Error . .000 .3936374000 . (2-tailed) Sum of Squares and Cross-products Covariance N **.003 Skewness Statistic .1976959400 -.003 59 Regression Descriptive Statistics Mean RF RM -.633 ** RM 1 .143 .008391494068 -.613 59 59 59 Range Statistic .4950981500 -. Deviation .0506555024063 N 59 59 Correlations RF RM . Deviation Statistic .0770129395083 .344 .000 .0770129395083 . Correlation is significant at the 0.1959414600 .680 .002160746815 Std.

400 a. RM b.633 a R Square . 59 59 Sig.000 .000 Durbin-Watson 2.0601590807227 a.390 Estimate . .156 .Pearson Correlation RF RM 1.000a F Change 38.168 Sig.344 df 1 57 58 Mean Square .006 . F Change . . Dependent Variable: RF Model Summaryb Change Statistics R Square Change . b.805 6. Dependent Variable: RF ANOVAb Model 1 Regression Residual Total a.008 .424 . Predictors: (Constant). Error . Enter a. Dependent Variable: RF Sum of Squares .050 Sig. Predictors: (Constant).633 1.138 . Predictors: (Constant).138 . RM b. All requested variables entered.050 df1 1 df2 57 Sig.000 59 59 .962 Std. .633 . Error of the Model 1 R .000 .004 F 38.206 .400 Adjusted R Square .066 Coefficientsa Standardized Unstandardized Coefficients Model 1 (Constant) RM B -.633 Coefficients Beta t -. Dependent Variable: RF Model Summaryb Std.000 . (1-tailed) RF RM N RF RM Variables Entered/Removed Model 1 b Variables Entered RM a Variables Removed Method .000 . RM b.

078 Mean -. Dependent Variable: RF RM RM RM 1.0596382130771 1. Predicted Value Std.991 N 59 59 59 59 Scatter Graph and Trend line .000 .081159777939 .024 Collinearity Diagnosticsa Variance Proportions Model 1 Dimension 1 2 a.633 .008391494068 .838 3.802 -2.957 Condition Index 1.274 .022 .144943192601 -. Dependent Variable: RF Eigenvalue 1.000 1.52 Residuals Statisticsa Minimum Predicted Value Residual Std.48 .0000000000000 .Variables Entered/Removed Model 1 b Variables Entered RM a Variables Removed Method .0487265471040 .044 (Constant) . Deviation .1328763663769 -2.000 .043 .650 a. Dependent Variable: RF Coefficients a 95.000 Std.000 Coefficient Correlationsa Model 1 Correlations Covariances a.633 Zero-order Correlations Partial Part Collinearity Statistics Tolerance VIF .48 .000 1.209 Maximum . Enter a.000 . Residual -.009 1.0% Confidence Interval for B Lower Bound -.52 RM .1851410716772 1.633 1. Dependent Variable: RF Upper Bound .

00152 Results .00033)0. βi = The Beta which has been calculated using SPSS for the HSBC share value to be 0.Rf) βi Where Ri = the return on the asset.002.033% RFR (Monthly) using Monthly RFR = ( 1 + .Rf )βi to calculate the expected rate of return (ERR) of 0.00033 + (-0. however due to the low Rm we are using (Ri .Rf) βi 0.962. Ri = Rf + (Rm . Rf = The risk free rate.0040)(1/12) – 1 = 0.962 = -0.00033.00191.4162(Ret HSBC) + 0.40 % which equates to . which I have decided will be the 1yr 5% Tr 12 RED which is currently .002 – 0.12.0013 R² = 0.4003 CAPM formula takes the form of Ri = Rf + (Rm . however Thompson analytics was 1.RMarket = 0. Rm = the average return on the market (Geometric Mean) which has been calculated to be -0.

negative in the market and positive in HSBC suggesting that there are more bad days occurring in the market than of the company.The CAPm calculation and statistical analysis resulted in a mixed bag of results. For the expected return on HSBC I have used beta x Equity Risk Premium. and barely turn the heads of long term investors who would normally be attracted by this company. However there is also skewness in both distributions.0023 is a negative value to the risk free rate. and that there is a good argument not to do anything other than purchase gilts. the other rates were converted to geometric . and that there is more chance that good and bad days occur within the company share price than that in the general FTSE market.25 providing a expected rate of return for January at -0. suggesting that the market has a more evenly probability of distribution than that of the company. due to the very low return on the market of -0.92 (compounded) = -1. There is also a high correlation to the market and HSBC share prices. suggesting that changes within the market affect the company shares. perhaps the dividend is amazing.9) The analysis returned a beta for HSBC of 0. indicating that investment in the FTSE market overall is barely worth the risk.16 (Annual non compounded rate = -1.962 indicating that this is a lower risk portfolio company but very close to the market combined with a negative ERR would not excite short term investors at all. the Risk Premium calculates at -0.002 the Market premium of -0. and vice vesa. Roll's critique mean-variance efficient Assumption The risk free rate is Arithmetical. Higher kurtosis of the tails in the HSBC shares than that of the market.

the problems arising in the CAPM.com Hedge Funds Bank Cost efficiency CAPM Issues Asset-Liability Management Derivatives and Financial Planning Benchmark Issue in CAPM Jeffry Merril Liando (2007) Capital Asset Pricing Model (CAPM) has been widely used for finding a suitable required rate of return of a share or portfolio.Share PageGet a free website at Webs. This essay aims to discuss this issue related to the good theoretical assumptions.com Moments worth paying for Roll over to find yours now! Connect via Meebo Keziana. However. the . the assumption of using a major share market index as the benchmark becomes an issue. the distortion in its application.

This is such an excellent portfolio theory so that Sharpe (1964)[2]. Lintner (1965)[3] and Mossin (1966)[4] further modelled it to include the risk free rate. If there is a benchmark error.practice of seeking alpha in responding the distortion and the impact of distortion for the New Zealand context. Rs = Rf + Beta ( Rm – Rf ). A capital market line that used to linearly predict the market return can be drawn by lining up a point of a portfolio with only risk free rate to the other point that touches the efficient frontier. The issue of benchmark index can be seen at this point. The tangency portfolio explains that investors separate their decisions in investing and financing the investment as suggested by Tobin (1958)[5]. The benchmark index is then set artificially to be a manifestation of a whole market reflecting the acceptable portfolio chosen by investors within the efficient frontier. To see the relationship between share and market returns. a risk free asset with a zero risk return can be combined with risky assets added to the efficient portfolio and investors can then go beyond the frontier by borrowing and lending at risk free rate. Markowitz (1952)[1] suggests that rational investors would choose minimum risk and maximum return in diversification and with any combination of weight the optimal portfolio lies on the efficient frontier. it can be formulated as follows: E(Rs) – Rf = Beta ( E(Rm) – Rf ). CAPM and efficient portfolio CAPM assumes that a major stock index can be used as a benchmark to determine risk premium and beta for calculating the required rate of return of a stock. risk free rate and beta. CAPM cannot estimate the correct beta and risk premium properly thus cannot calculated the expected rate of share return correctly. The formulae can be shown as follows. The line is plotted by the expected rate of return of a share with its beta to the market return. known as a tangency portfolio. Benchmark error Benchmark error using CAPM for evaluating portfolio performance according to Ross (1980. in which a share’s rate of return can be predicted given the market return. The problem is not due to statistical variation but rather to the .1981)[6] can be seen in two ways when the market index produces an incorrect beta for the share and when it produces incorrect estimation for the market premium optimised to the risk free rate (see figure 1 and 2). In a sense of linear prediction of an individual share return. CAPM is then modelled with a security market line. As the efficient frontier only includes the portfolio of risky assets. The benchmark index assumed in the CAPM is promoted as the market proxy of the efficient portfolio of risky assets. or by noticing the future expectation as this: E(Rs) = Rf + Beta ( E(Rm) – Rf ).

Market proxy. Benchmark errors are also considered in the context of global investment. share or portfolio performance is sensitive to the choice of benchmark for the market index. Figure-1. Incorrect market premium A further study by Green (1986)[7] shows that benchmark errors are continuous behaviour and different for different indexes. The beta of domestic equity index is lower than the world equity index and much larger for the diversified global stock and bond portfolio. beta inconsistency. Market proxy and efficient frontier Using the true market proxy of the value weighted portfolios of all US shares. We may now say that as beta assumed equals to one. AMEX and NASDAQ from 1928 to 2003 and to be compared to the returns predicted with CAPM. and lower for any benchmark that produces a lower premium. Fama and French (2004)[12] tested CAPM by plotting the annualised monthly return and beta of every stock in NYSE. Ross (1978)[10] also added that market proxy is not ex ante mean-variance efficient and individual preference in portfolio selection may be judged with a different market index and then will be penalised by share’s beta according to the different market index. thus. beta and risk premium problems The root of benchmark error was based fundamentally on Roll’s critique (1977)[9] that found that market index is efficient per se. not for the individual shares or portfolios.cause that the market index is not a good predictor of mean/variance efficient portfolio. . the expected rate of return should be higher as we choose any benchmark that produces a higher market risk premium. The result is telling us further problem other than benchmark error and market proxy problem. global index or a diversified global stock and bond portfolio. Figure-3. Roll and Ross (1994)[11] found that the market proxy may be located within 22 bps below the efficient frontier (Figure-3). that is. Incorrect beta Figure-2. Reilly and Akhtar (1995)[8] found that there is a variation of beta when using a domestic index.

Beta inconsistency. cash flow. high risk free rate. there are some market characteristics need to be considered (Bowden.133-168)[19]. high B/M. as follows. low B/M. p. high dividend yield. an alternative indexing has been suggested by Arnott (2005)[18] in fundamental indexation to solve the distortion in value weighted index with some alternative weighting constraints. not beta.Rf + Beta ( Rm – Rf). The evidence shows that expected return does not compensate beta variation unrelated with size and B/M for both small stocks and big stocks. B/M and beta as concluding that size and B/M.Figure-3. We may say that if CAPM is fine a good benchmark index should contain both small stocks and big stocks. narrow true market proxy. However. gross dividend and total employment. reward the expected return. gross sales. Fama and French (2006)[14] tested whether the value premium exists in CAPM pricing. 2005. Again using the true market proxy and the three-factor CAPM. Moreover. Recently. one can say that even a passive portfolio can beat the benchmark. Figure-3 shows an inconsistency of beta in CAPM. They concluded that value shares with high dividend yield. The inconsistency of beta was identified a decade before by Fama and French (1992)[13] as suggested a three-factor model by adding size and book to market ratio (B/M) to CAPM. The result is rejecting CAPM pricing for portfolio based on size. . low holding period return and low prospect dominance. Low beta shares that are predicted to have low returns are in fact too high whereas high beta shares that are predicted to have high returns are in fact too low. large foreign capitalisation. such as book value. New Zealand context In order to see the impact of benchmark distortion in the New Zealand context. as seemingly the line rotates. revenues. Fama and French (2006)[17] discussed about a portable alpha as a way to add a portfolio consisting risk free rate and index funds with an additional hedge position that generates alpha. Bowden (2000)[16] further argued that alpha relates to market timing and cannot be observed by conventional performance measures and suggested ordered mean difference (OMO) as an alternative measure. Alpha Market proxy distortion opens an opportunity to have a better portfolio performance than the market itself which was earlier suggested by Jensen (1969)[15] with alpha as a performance indicator: Alpha = Rs . low P/E tend to have bigger expected return than growth shares with low dividend yield. high P/E. which is a function of a running mean of the difference between asset/portfolio and benchmark returns that is ordered by values of the benchmark.

5% of foreign capitalisation. the linear CAPM prediction using such proxy could be far from the reality of historical returns. it is likely that the correct benchmark used would be the gross index. A promotion to use the gross index to international investors may attract them to choose NZ equity. High risk free rate may increase the cost of capital with a condition that there is no false market risk premium estimation in the benchmark. Cavalier. a rich Waikato Fonterra farmer may see a comparable investment choice between investing in Fonterra shares and NZX-50 shares. Value shares may have a higher expected rate of return. In fact. Fonterra only issues capital notes and none of its capitalisation in the share market. thus a higher cost of capital. However. If the CAPM holds. As Fama and French (2004.44) suggested. That is why investors may have the opportunity of seeking alpha and beat the index with some shares like Michael Hill. not the capital index which is widely used in the other market. High dividend yield and low P/E make NZ shares considered as value shares. if benchmark error deviates to a lower market return then the market risk premium would be narrow and the security market line may rotate. Conclusion CAPM assumes a major share market index as the best market proxy for efficient portfolio to predict the required rate of return. For example. Under CAPM prediction.A consensus for share market proxy in NZ is the NZX-50 index that can be seen in every business page and news and the NZX-All for the true market proxy. “But we also . there is a significant proportion of investment taking in the form of farming shares as the grass root of the NZ economy as a whole. of course the market risk premium would be negative and the expected share return would be below the risk free rate. The NZX-50 benchmark is dominated by larger companies with low growth prospect. Low holding period return for NZ shares may be compensated by high risk free rate so that at some period the return of NZ shares is in fact is lower than government bond return. Despite the good theory background of mean (return) – variance (risk) efficient portfolio. if NZ benchmark consists mainly of such shares. If tested with CAPM. thus she needs a broader efficient frontier for the true market proxy other than just NZX-All. there would be a big deviation from the predicted return of such shares. than growth shares according to the CAPM test by Fama and French (1992). then the return of individual shares may be predicted below or above the original security market line. Since beta domestic equity market after influenced by beta foreign equity market may change. Fisher and Paykel. etc. Dorchester. p. However. The top ten capitalisations in the NZX-50 hold around 37.

Benchmark Portfolio Inefficiency and Deviations from the Security Market Line.warn students that despite its seductive simplicity. [3] Lintner. James (1958). Journal of Portfolio Management. Summer 1980. Jan. (1981). p7-14. 25. Portfolio selection. (1986). Vol. [5] Tobin. Richard C. The Review of Economic Studies. Vol. Winter 1981. (1964). (1980). Journal of Finance. 34 Issue 4. (1966). Econometrica. Vol. p17-22. John. Sep 1964. Vol. Richard. p13-37. Richard. Oct 1966. [7] Green. Jun 1986. p77-91. Harry. (1952). p768-783. Journal of Finance. . [2] Sharpe. 7 Issue 1. Vol. The valuation of risk assets and the selection of risky investments in stock portfolios and capital budgets. p425-442. 2. Vol. Review of Economics & Statistics. Liquidity preference as behaviour towards risk. [4] Mossin. p295-312. 7 Issue 2. Capital asset prices: a theory of market equilibrium under conditions of risk. No. 19 Issue 3. Vol. References [1] Markowitz. 41 Issue 2. [6] Roll. Roll. p65-86. Journal of Finance. Feb 1965. Equlibrium in a capital asset market. Vol. Performance evaluation and benchmark errors (II). Mar 1952. 47 Issue 1. (1965). Journal of Portfolio Management. the CAPM’s empirical problems probably invalidate its use in applications”. William F. 6 Issue 4. Performance evaluation and benchmark errors (I).

Journal of Economic Perspectives. 61 Issue 5. Jason Hsu and Philip Moore. [11] Roll. Mar/Apr 2005. Journal of Finance. Fundamental Indexation. Richard. (1978). Oct 2006. p83-99. and the evaluation of investment portfolios. Tilted Portfolios. Vol. Jun 1992. Hedge Funds. 61 Issue 2. (1969). Vol. Vol. Issue 3. Journal of Finance. Frank K and Rashid A Akhtar. (2005).[8] Reilly. The Current Status of the Capital Asset Pricing Model (CAPM). Risk. (2005). Kiwicap: an introduction to New Zealand capital markets. Winter 2007. [15] Jensen. [10] Ross. (2nd ed. Chicago GSB Magazine. (1992). Stephen A. Vol. Apr 1969. Journal of Business. 42 Issue 2. The Value Premium and the CAPM. Vol. (1977). Wellington: Kiwicap Education. Eugene F and Kenneth R French. Michael C. [18] Arnott. Journal of Portfolio Management. 4 Issue 2. On the cross-sectional relation between expected returns and betas. [16] Bowden. p167-247. Eugene F and Kenneth R French. Richard and Stephen A Ross. Mar 1977. By: Journal of Finance. p427-465. [9] Roll. Robert D. Roger and Jennifer Zhu. 49 Issue 1. p129-176. the pricing of capital assets. 18 Issue 3. 22 Issue 1. [12] Fama. Eugene F and Kenneth R French. Vol. (1994). Journal of Finance.). A critique of the asset pricing theory's tests: part I: on past and potential testability of the theory. p25-46. The benchmark error problem with global capital markets. . The capital asset pricing model: theory and evidence. The Cross-Section of Expected Stock Returns. (1995). Financial Analysts Journal. (2006). 47 Issue 2. Vol. Summer 2004. (2006). Mar 1994. Journal of Financial Economics. Vol. [13] Fama. [17] Fama. p2163-2185. 33. Eugene F and Kenneth R French. Vol. Fall 1995. Jun 1978. p33-50. and Portable Alpha. (2004). p101-121. p885-901. [14] Fama.

. Wellington: Kiwicap Education.[19] Bowden. Kiwicap: an introduction to New Zealand capital markets.). (2005). (2nd ed. Roger and Jennifer Zhu.