Ex2 Estimating and Testing the Capital Asset Pricing Model Preamble In this exercise, I use monthly data and ordinary least squares (OLS) method to estimate the ‘alphas’ ‘ betas’ and ‘least square residuals’ of two stocks (equities)—Mobil and Tandy separately. According to the results from TSM, I can calculate the 95% confidence intervals for β, use the t-test or p-values to test the significance of α and t-test to test the null hypotheses that β=1 against the alternative hypothesis β>1, and find the coefficient of determination and the standard deviation of the residuals to measure the goodness of fit of regression, in the other word in this exercise, the proportion of the risk attributable to the market and the individual risk of the stocks. To check the stability of the models over the full 10-year period of the sample, I use the method of forecast adequacy test (Chow stability test 2). The strict form of the CAPM assumes that markets are efficient. This assumption can be tested by including macro variables in the regression—the rate of inflation, the growth in industrial production, and changes in the real oil price (the Arbitrage Pricing Model). I use F-test to test the joint significance of these macro variables. πΈ(πππ‘ )−πππ‘ Finally, the values of πΈ(πππ‘ ), πΈ(πππ‘ ) and πππ‘ is used to calculate the CAPM equation π½π = , πΈ(πππ‘ )−πππ‘ which is compared with the OLS estimate of π½π for the same time period. Stock 1(Mobil) M M 0 0 . . 1 . R K E O B I L T 0 . 4 0 . 3 0 . 2 0 . 2 5 0 0 A . 0 1 5 0 - 0 . - 0 1 5 0 . 1 0 - 0 . - - 0 . - 1 5 0 . 2 - 0 - 0 0 . 0 1 / 1 0 9 1 7 / 1 S . 1 0 6 9 1 7 c / a 1 0 7 9 t t 1 7 e / 1 r 0 8 9 P 1 7 l / 1 o 0 9 9 t 1 , 8 M / 1 0 0 9 1 8 A / 1 0 1 9 R K 1 8 E / 1 0 2 9 1 8 T / 1 0 3 9 v s 1 8 . / 1 0 M 4 9 1 8 / 1 0 5 O 9 B 1 8 I / 1 6 9 8 7 L 2 0 . 0 . 0 5 - 0 . 0 5 - 0 - 0 0 5 . 1 0 - 0 . - - . 1 0 . 0 1 . 2 3 0 0 . 2 5 1 5 . 2 2 5 . 3 - 0 . 2 - 0 . 1 0 0 . 1 0 . 2 0 . 3 0 . 4 1 / 1 0 9 1 7 / 8 1 0 9 1 7 / 9 1 0 9 1 8 / 0 1 0 9 1 8 / 1 1 0 9 1 8 / 2 1 0 9 1 8 / 3 1 0 9 1 8 / 4 1 0 9 1 8 / 5 1 0 9 1 8 / 6 1 9 8 7 It can be seen from the two plots in the first line that the trend of the stock return followed that of market return. In the special time October 1987 which witnessed the market crash, the return dropped sharply. The scatter diagram of returns disregards the time ordering but represents the relationship between market and stock apparently. The rate of change in stock return was bigger than that in market return. Estimate Std. Err. t Ratio p-Value R-Squared = 0.3653 Intercept 0.00339 0.00914 0.37 0.712 Residual SD = 0.0686 MARKET 0.67838 0.11742 5.777 0 Chow Stability Test: ChiSq(2) = 0.6869{0.709} Using ordinary least squares in TSM, I get the estimation of πΌΜ=0.00339 and π½Μ =0.67838 which means that if the market return is equal to zero, the stock return will be 0.00339 and when the market return grows every one unit, the stock return will increase by 0.67838 units accordingly.95% confidence intervals of β is π½Μ -1.96ππ½Μ ≤β≤π½Μ +1.96ππ½Μ . ππ½Μ is in practice unknown, it is replaced by the estimate, S.E.(π½Μ )=0.11742 in this model. According to my calculation, 95% confidence intervals of β is (0.448 , 0.909). To test the null hypothesis, α=0, I use the t-test | Μ πΌ 0.00339 |=| Μ) π .π.(πΌ |≈0.37 (t Ratio), which is less than the critical value (1.96) 0.00914 of 5% significance level from normal distribution (2-tailed test). I have no reason to reject the null hypothesis, so α is equal to zero. I can draw the same conclusion from testing p-value of α, which is 0.712 more than 0.05. Another null hypothesis is β=1 against the alternative hypothesis β>1. On the basis of ttest, Μ −π½∗ 0.67838−1 π½ Μ )= 0.11742 ≈-2.74<1.64 π .π.(π½ (1-tailed test), I cannot reject the null hypothesis that β is equal to one. π 2 is equal to 36.53% which means that 36.53% of the risk is attributable to the market. The standard deviation of the OLS residual is 0.0686 which measures the individual risk of the stock. Through running Chow stability test, I get the result that ChiSq(2)=0.6869{0.709}. P-value is equal to 0.709 which is more than 0.05. I do not reject the null hypothesis. On the other word, the model is stable over the full 10-year period of the sample. I change the Capital Asset Pricing Model (CAPM) instead of Arbitrage Pricing Model (APM) by including macro variables in the regression. I use an F-test to decide whether the null hypothesis that π½2 =π½3 =π½4 =0 is reasonable. According to the result that p-value is equal to 0.697 which is bigger than 0.05, I cannot reject the null hypothesis. So the error term π’ππ‘ depends solely on individual firm effects, and is not predictable from macroeconomic variables. Wald Test of Zero Restrictions on: GIND RINF ROIL F (3,115) = 0.4795 {0.697} πΈ(π )−π The CAMP equation is π½π =πΈ(π ππ‘ )−πππ‘ . Through the TSM, I get the sample means, and ππ‘ Μ Μ Μ Μ Μ −π πππ Μ Μ Μ Μ Μ ππ‘ ππ‘ 0.0164667−0.0080517 calculate π½π ≈Μ Μ Μ Μ Μ −π = ≈0.74848.The OLS estimate of π½Μπ for the same time period is Μ Μ Μ Μ Μ 0.0192833−0.00808517 π ππ‘ ππ‘ 0.67838, which is near to the CAMP estimate of π½π . ***Summary Statistics for MARKET*** Mean = 0.0192833 ***Summary Statistics for MOBIL*** Mean = 0.0164667 ***Summary Statistics for RKFREE*** Mean = 0.00808517 Stock 2(Tandy) T M 0 0 . . 1 . R K E A N D Y T 0 . 5 0 . 4 0 . 3 0 . 2 0 . 2 5 0 0 A . 0 1 5 0 - 0 . 0 1 5 0 - - 0 . - - 0 0 1 1 - 0 . 1 - 0 . 2 - 0 . 3 5 0 . - . . 2 2 5 0 . 3 0 0 1 / 1 0 9 1 7 / 1 0 6 S 0 . 1 0 9 c 1 7 / 1 a t 0 7 t 9 1 7 e / r 1 0 8 9 P 1 7 l o / 1 0 9 t 9 1 8 , M / 1 0 0 9 1 8 A / 1 0 1 R 9 K 1 8 E / 1 0 2 9 1 8 T / 1 v 0 3 9 s 1 8 . / 1 0 4 9 T A 1 8 / 1 0 5 N 9 1 8 D / 1 6 9 8 1 / 1 0 9 1 7 / 8 1 0 9 1 7 / 9 1 0 9 1 8 / 0 1 0 9 1 8 / 1 1 0 9 1 8 / 2 1 0 9 1 8 / 3 1 0 9 1 8 / 4 1 0 9 1 8 / 5 1 0 9 1 8 / 6 1 9 8 7 Y 5 . 1 0 . 0 5 - 0 . 0 5 - 0 - 0 0 - 0 . - 0 . - . 1 0 1 5 . 2 2 5 . 3 - 0 . 3 - 0 . 2 - 0 . 1 0 0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 It is apparent from the information given the market return and the stock return changed around the value of zero. The trend of stock return usually followed that of market return, but the stock return of Tandy had a larger fluctuation. In October 1987, the market experienced a crash, the market return dropped down quickly, and so do the stock return of Tandy. The scatter plot shows the relationship between market return and the stock return. 7 Estimate Std. Err. t Ratio p-Value R-Squared = 0.3056 Intercept 0.03084 0.01591 1.939 0.057 Residual SD= 0.1194 MARKET 1.03233 0.20433 5.052 0 Chow Stability Test: ChiSq(2) = 4.4871 {0.106} The estimation of α and β is equal to 0.03084 and 1.03233 separately, which means that if the market return grows every one unit, the stock return will increase by 1.03233 units, and if the market return is zero, the stock return of Tandy will be 0.03084. In this model, S.E.(π½Μ ) is equal to0.20433, so 95% confidence intervals of β is (0.632,1.433). On the basis of t-test, | Μ πΌ 0.03084 |=| Μ) π .π.(πΌ 0.01591 |≈1.939<1.96 (2-tailed test), I cannot reject the null hypothesis that α is equal to zero. P-value test can also be used in this hypothesis test. The value is 0.057, more than 0.05. I get the same conclusion from this result that α is equal to zero. To test the whether β is equal to one or it is bigger than one, I calculate the t-value of it, Μ −π½∗ 1.03233−1 π½ Μ )= 0.20433 ≈0.158<1.64 π .π.(π½ (1-tailed test), I have no reason to reject the null hypothesis π½=1. The goodness of fit of regression (π 2) is 30.56%. The probability of risk attributable to the market is 30.56%. The individual risk of the stock is measured by the standard deviation of the OLS residual which is 0.1194. I use the Chow stability test to check whether the model is stable over the full 10-year period of the sample. P-value is equal to 0.106 bigger than 0.05, therefore the model is stable. To test the assumption that markets are efficient, macro variables (growth, inflation and oil price) is included. Though the Wald test of zero restrictions on GIND, RINF and ROIL, I get the p-value (0.397) which is more than 0.05. The assumption is acceptable. Wald Test of Zero Restrictions on: GIND RINF ROIL F (3,115) = 0.9964 {0.397} Μ Μ Μ Μ Μ −π Μ Μ Μ Μ Μ π 0.05075−0.0080517 ππ ππ‘ In the model, I calculate the CAMP estimate of π½π ≈Μ Μ Μ Μ Μ −π = ≈3.80910. However the Μ Μ Μ Μ Μ 0.0192833−0.00808517 π ππ‘ ππ‘ OLS estimate of π½Μπ for the same time period is 1.03233. There is a big difference between these two values. OLS regression is estimated on the basis of a quadratic weighting approach that conflict the assumption of risk aversion, and the data do not show a normal distribution, because the probability distribution of market returns are with “fat” tails. Due to these factors, beta is sensitive to market fluctuations, which may make OLS inappropriate for beta estimation. ***Summary Statistics for MARKET*** Mean = 0.0192833 ***Summary Statistics for RKFREE*** Mean = 0.00808517 ***Summary Statistics for TANDY*** Mean = 0.05075 Conclusion In the Capital Asset Pricing Model (CAPM), πππ‘ =πΌπ +π½π πππ‘ +π’ππ‘ , πΌπ and π½π are constant coefficients. Usually, πΌπ is equal to zero. Seeing from the tests done separately in the two stocks (Mobil and Tandy), I both have no reason to reject the null hypothesis that α is equal to zero. π½π is the basic indicator of the stock’s risk and return characteristics. CAPM says that it is impossible for an asset to have a higher than average return, or π½π >1, without being relatively risky, and/or relatively highly correlated with the market. In the CAPM of stock 1 and stock 2, the null hypothesis of π½π =1 is not refutable. Through the project, I am familiar with the Capital Asset Pricing Model (CAPM) and ordinary least squares (OLS) method. I know how to do some estimates and tests, and check the stability of the models. I can analyse the stocks of the companies with the conclusion drawn from TSM.