Capital Asset Pricing

7 Think about how to apply different models in

real situations. How can the CAPM be
applied?
1
OUTCOMES
CHECK IF YOU CAN DO/UNDERSTAND THE FOLLOWING AFTER
CLASS

• Understand CAPM and multi-factor model

• Difference between the models
• The shortcomings of CAPM and its practical
application

2
 7.1 THE CAPITAL ASSET PRICING MODEL (CAPM)
BY TREYNOR, SHARPE, LINTNER AND MOSSIN IN 1960S
 CAPM formalizes our understanding about the relationship between
expected return and risk. 𝐸 𝑟𝑖 = 𝑟𝑓 + β𝑖 𝐸 𝑟𝑀 − 𝑟𝑓 Whether the
expected return can be realized is another issue. In other words, it
tells us how asset should behave.
 Assumptions
 Markets are competitive, equally profitable
 No investor is wealthy enough to individually affect prices
 All information publicly available

 All securities are publicly owned and traded

 No taxes on returns, no transaction costs

 Unlimited borrowing/lending at risk-free rate

 Investors are alike except for initial wealth, risk aversion

 Investors plan for same single-period horizon;
 They are rational, mean-variance optimizers

 Use same inputs, consider identical portfolio opportunity sets 3

 7.1 THE CAPITAL ASSET PRICING MODEL
 Hypothetical Equilibrium
 All investors choose to hold market portfolio (M)
 Market portfolio is on efficient frontier, optimal risky portfolio
(see next slide), although you may question whether it is
optimal (the textbook has a brief account on this)

Note: we won’t go through the details of the research paper, we

will only point out some important features of CAPM

4
FIGURE 7.1 EFFICIENT FRONTIER AND
CAPITAL MARKET LINE

5
 7.1 THE CAPITAL ASSET PRICING MODEL
 Hypothetical Equilibrium
 When you buy stocks, demand drives prices, lowers
expected rate of return/risk premiums
 When premiums fall, investors move some of their
funds into risk-free asset
 Equilibrium risk premium of market portfolio (should
be high enough for you/investors to hold the stocks),
should be proportional to the variance of market
portfolio and investor’s risk aversion

E(rM) – rf = Ā σ2M , Ā – average degree of risk

aversion 6
 7.1 THE CAPITAL ASSET PRICING MODEL
 Risk premium on individual assets
 Proportional to risk premium on market portfolio
 Proportional to beta coefficient of security on market
portfolio
 Will investors be reward of nonsystematic risk?
 Compensation for bearing only systematic risk!
 Risk cannot be diversified

 So risk premium of an asset (A) is proportional to its

beta
𝐸 𝑟𝑀 −𝑟𝑓 𝐸 𝑟𝐴 −𝑟𝑓
 =
1 𝛽𝐴
=> 𝐸 𝑟𝐴 = 𝑟𝑓 + β𝐴[𝐸 𝑟𝑀 − 𝑟𝑓] 7
 7.1 THE CAPITAL ASSET PRICING MODEL
 Passive Strategy is Efficient
 Mutual fund theorem*: All investors desire same
portfolio of risky assets, can be satisfied by single
mutual fund composed of that portfolio
 If passive strategy is costless and efficient, why
follow active strategy?
 If no one does security analysis, what brings about

efficiency of market portfolio?

*Nobel laureate James Tobin suggests that all investors

should hold an identically comprised portfolio combined
with some percentage of risk-free assets 8
7.1 THE CAPITAL ASSET PRICING MODEL
 The Security Market Line (SML), next graph
 Represents expected return-beta relationship of CAPM
𝐸 𝑟𝑖 = 𝑟𝑓 + β𝑖 𝐸 𝑟𝑀 − 𝑟𝑓
 Graphs individual asset risk premiums as function of asset
risk, beta is also a risk measure, remember Total variance
= β2iσ2M + σ2(ei) in Ch.6.

 Alpha
 Abnormal rate of return on security in excess of that
predicted by equilibrium model (CAPM)
 Underpriced stocks plot above the SML meaning expected
returns are greater than those indicated by the CAPM
9
FIGURE 7.2 THE SML AND A POSITIVE-
ALPHA STOCK

10
 7.1 THE CAPITAL ASSET PRICING MODEL
 Applications of CAPM
 In capital budgeting decision, SML provides “hurdle rate” for
internal projects
 Use SML as benchmark for fair return on risky asset

-e.g. beta calculated by CAPM can be used to set price limit

Investment research firms or advisers maintain a list of common
stocks with target price with buy limit. So based on CAPM, an
expected return, r, can be calculated and then the r will be used to
discount future dividends of the stock to get the fair price.

-e.g. beta of stocks help you decide the types of stocks to buy. You
may buy stocks of high beta if stock market is going up to enjoy a
bigger gain.
11
7.2 CAPM AND INDEX MODELS

• CAPM applies to expected as opposed to

actual return

• CAPM relies on the theoretical market

portfolio including all assets

• Index model replaces market portfolio with

a market index

• Index model applies to realized returns

12
7.2 CAPM AND INDEX MODELS

13

Take expectation of the equation, CAPM predicts alpha to be zero

 7.2 CAPM AND INDEX MODELS
•

Alpha is the average of the firm’s specific factor at the time, it cannot be
explained by the market index
14
 TABLE 7.1 MONTHLY RETURN
STATISTICS 01/06 - 12/10
Statistic (%) T-Bills S&P 500 Google
Average rate of return 0.184 0.239 1.125

Average excess return - 0.054 0.941

Standard deviation* 0.177 5.11 10.40

Geometric average 0.180 0.107 0.600

Cumulative total 5-year return 11.65 6.60 43.17

Gain Jan 2006-Oct 2007 9.04 27.45 70.42

Gain Nov 2007-May 2009 2.29 -38.87 -40.99

Gain June 2009-Dec 2010 0.10 36.83 42.36

* The rate on T-bills is known in advance, SD does not reflect risk.

15
FIGURE 7.3A: MONTHLY RETURNS

16
FIGURE 7.4 SCATTER DIAGRAM/SCL: GOOGLE VS.
S&P 500, 01/06-12/10

Can we change x-axis

to y-axis?

17
 TABLE 7.2 SCL FOR GOOGLE (S&P 500), 01/06-12/10
Linear Regression

Regression Statistics
R 0.5914 Correlation coeff.
R-square 0.3497 34.97% of the return variation is
explained by index changes

Adjusted R-square 0.3385

SE of regression 8.4585
Total number of observations 60

Firm’s specific risk: standard error =8.4585% = sqrt(4149.65/58)

using outputs next slide, i.e. residual SS/df

18
 TABLE 7.2 SCL FOR GOOGLE (S&P 500),
01/06-12/10

ANOVA
df SS MS F p-level
Regression 1 2231.50 2231.50 31.19 0.0000
Residual 58 4149.65 71.55
Total 59 6381.15

 2 (egoogle) 4149.65
1 2  1  0.3497
 google S &P500   (egoogle)
2 2
6381.15

 google
2
 S2& P 500 2231.5
R  Square  2   0.3497
 google S & P 500   (egoogle ) 6381.15
2 2

19

For exam. purpose, knowing the meaning of R-square above is good enough,
The details of SS, MS, df, F and p-level would be reference materials.
TABLE 7.2 SCL FOR GOOGLE (S&P 500),
01/06-12/10
Standard
Coefficients Error t-Statistic p-value
Intercept 0.8751 1.0920 0.8013 0.4262
S&P 500 1.2031 0.2154 5.5848 0.0000

Required rate of (excess) return = alpha + β x Expected excess

return of index

Regression equation:
Google (excess return) = 0.8751 + 1.2031 × S&P 500 (excess return)

20
 7.2 CAPM AND INDEX MODELS
 Predicting Betas
 Mean reversion
Betas move towards mean over time
High-beta securities tend to have lower

beta in the future and the reverse

firms may become more “typical” as they

age, causing their betas to approach 1.

To predict future betas, adjust estimates
from historical data to account for
regression towards 1.0
21
ADJUSTED BETA EXAMPLE:

X’s beta was estimated to be 1.2 based on past

60 months data
Adjusted beta =(2/3 x 1.2 )+ 1/3 = 1.13333

Predicted Beta example:

βt = 0.2 + 0.8βt-1

22
 7.3 CAPM AND THE REAL WORLD
 CAPM is false based on validity of its assumptions
 Useful predictor of expected returns
 There seem to be risk factors affecting security
returns beyond beta’s measure

Note: You can refer to the Journal article mentioned

in the text (section 7.3) for more details

23
7.4 MULTIFACTOR MODELS AND CAPM

• In index model, the return of the market

portfolio summarized the aggregate impact
of macro factors.

• However, in reality, systematic risk is due

to many sources. Worldwide factors like
interest rate, business-cycle risk, etc.

24
7.4 MULTIFACTOR MODELS AND CAPM
•

Two sources of risk:

-Returns on broad market 25
-Unanticipated changes in interest rates captured by returns
of T-Bond (TB)
7.4 MULTIFACTOR MODELS AND CAPM

Historical observations: average returns on stocks of small firms

and stocks return of high book to market value
are higher than predicted by CAPM

Size matters!

Size premium is constructed by difference in returns between

small and large firms (SMB: small minus big)

B/M premium: high B/M minus low B/M ,i.e. HML

26
•

27
TABLE 7.4 REGRESSION STATISTICS: ALTERNATIVE
SPECIFICATIONS
Regression statistics
for: 1.A Single index with S&P 500 as market proxy
1.B Single index with broad market index
(NYSE+NASDAQ+AMEX)
2. Fama French three-factor model (Broad
Market+SMB+HML)

Monthly returns January 2006 - December 2010

Single Index Specification FF 3-Factor Specification
Broad Market with Broad Market
Estimate S&P 500 Index Index

Correlation coefficient 0.59 0.61 0.70

Adjusted R-Square 0.34 0.36 0.47

Residual SD = Regression SE
(%) 8.46 8.33 7.61
Alpha = Intercept (%) 0.88 (1.09) 0.64 (1.08) 0.62 (0.99)
Market beta 1.20 (0.21) 1.16 (0.20) 1.51 (0.21)
SMB (size) beta - - -0.20 (0.44) 28
HML (book to market) beta - - -1.33 (0.37)
EXERCISES

Q1, 3, 4, 5, 6, 11

29