Capital Asset Pricing Model (CAPM)

Chapter 2: Capital Asset Pricing Model

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 1 / 27

...
Agenda

1 Introduction

2 Capital Asset Pricing Model (CAPM)

3 Economic interpretations associated with CAPM

4 Hedging with CAPM

5 Parameter Estimation for CAPM

6 Real Applications

7 References

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 2 / 27

...
Introduction

In common sense, a price is the amount you need to pay for some
goods/services
In the context of portfolio selection, a price can be regarded as the return
required for bearing a risk
Capital Asset Pricing Model (CAPM) states that

E(Ra ) = r + βa (E(RM ) − r)
r is the risk-free rate

E(Ra ) is the expected return of an individual capital asset a

E(RM ) is the expected return of the market portfolio M

E(RM ) − r is market premium: return needed to compensate the risk of M
E(Ra )−r
βa = E(RM )−r
is sensitivity of expected excess return (asset over market)
Cov(Ra ,RM )
(will show) βa = Var(RM )
can be regarded as a measure of risk

high risk ⇔ large βa ⇒ large E(Ra ) ⇔ high return

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 3 / 27

...
Capital Asset Pricing Model (CAPM)

Capital Asset Pricing Model (CAPM)

William F. Sharpe (1934∼): Nobel Economics 1990

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 4 / 27

...
Agenda

1 Introduction

2 Capital Asset Pricing Model (CAPM)

3 Economic interpretations associated with CAPM

4 Hedging with CAPM

5 Parameter Estimation for CAPM

6 Real Applications

7 References

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 5 / 27

...
Capital Asset Pricing Model (CAPM)
Market Setting:
Return of n assets: R = (R1 , . . . , Rn )
Expected return: µ = E(R R)
R) = Σ = (σij ) = (Cov(Ri , Rj ))
Variance of return: Var(R
Σ −1 (µ
µ−r1
1) cµµ −rc1µ |γ|
Tangency portfolio M : w M = c1µ −rc11
, µM = c1µ −rc11
, σM = c1µ −rc11

γ 2 = cµµ − 2rc1µ + r2 c11 , c11 = 1 T Σ −11 , c1µ = 1 T Σ −1µ , cµµ = µ T Σ −1µ

Theorem 1 (CAPM)
For any asset a with return Ra , the expected return µa = E(Ra ) can be
expressed as

µa = r + βa (µM − r) ,

r is the risk-free rate

µM = E(RM ) is the expected return of the market portfolio
Cov(Ra ,RM )
βa = Var(RM )

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 6 / 27

...
Capital Asset Pricing Model (CAPM)
Σ −1 (µ
µ−r1
1) cµµ −rc1µ |γ|
Tangency portfolio M : w M = c1µ −rc11
, µM = c1µ −rc11
, σM = c1µ −rc11

γ 2 = cµµ − 2rc1µ + r2 c11 , c11 = 1 T Σ −11 , c1µ = 1 T Σ −1µ , cµµ = µ T Σ −1µ

Cov(Ra ,RM )
Thm 1 (CAPM): µa = r + βa (µM − r), where βa = Var(RM )

Proof: Method 1: direct computation of covariance.

If a market is complete, any asset can be replicated by existing assets
⇒ asset a can be expressed as w a = (wa1 , wa2 , . . . , wan )
⇒ return of asset a is Ra = w Ta R . (R
R is return vector of assets in the market)
⇒ expected return of asset a is µa = w Ta µ , where µ = E(R R)
Computation of variance and covariance:
2 |γ|2 cµµ − 2rc1µ + r2 c11 µM − r
Var(RM ) = σM = 2
= =
(c1µ − rc11 ) (c1µ − rc11 )2 c1µ − rc11
Cov(Ra , RM ) = wT
Cov(w T T
a R , w M R ) = w a Σw M
Σ −1 (µ
µ − r1
1) w T (µ
µ − r1
1) µa − r
= wT
aΣ = a = wT
(w a 1 = 1)
c1µ − rc11 c1µ − rc11 c1µ − rc11
µa − r
= Var(RM )
µM − r
The result follows by straightforward algebra
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 7 / 27
...
Capital Asset Pricing Model (CAPM)
Σ −1 (µ
µ−r1
1) cµµ −rc1µ |γ|
Tangency portfolio M : w M = c1µ −rc11
, µM = c1µ −rc11
, σM = c1µ −rc11

γ 2 = cµµ − 2rc1µ + r2 c11 , c11 = 1 T Σ −11 , c1µ = 1 T Σ −1µ , cµµ = µ T Σ −1µ

Cov(Ra ,RM ) σaM
Thm 1 (CAPM): µa = r + βa (µM − r), where βa = Var(RM )
=: σM2

Proof: Method 2: tangency of portfolios formed by i and M .

Consider a portfolio with weight w on asset a & (1 − w) on tangency portfolio M
1 µw = wµa + (1 − w)µM
2 2 = w 2 Var(R ) + 2w(1 − w)Cov(R , R ) + (1 − w)2 Var(R )
σw a a M M

The curve {(σw , µw ); w ∈ R} should be tangent to the efficient frontier line

At w = 0, the curve takes value (σ0 , µ0 ) = (σM , µM ) on the frontier line
If the curve crosses frontier, then efficiency of the frontier is contradicted.
Tangency ⇒ slopes of curve and frontier agree
µM −r
Slope of frontier is σM
(the frontier line passes (0, r) and (σM , µM ))
dµw dµw
 dσ
w
Slope of curve at (σM , µM ) is dσw w=0
= dw dw
, where by 1 & 2
w=0
dµw
dw
= µa − µM
2
dσw
2σw dσ
dw
w
= dw
= 2wσa2 2
+ 2(1 − 2w)σaM − 2(1 − w)σM
µM −r µa −µM
dµw dµw
 dσw
Simplifying σM
= dσ = dw dw
= 2 )/σ
(σaM −σM
⇒ Done
w w=0 w=0 M

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 8 / 27

...
Agenda

1 Introduction

2 Capital Asset Pricing Model (CAPM)

3 Economic interpretations associated with CAPM

4 Hedging with CAPM

5 Parameter Estimation for CAPM

6 Real Applications

7 References

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 9 / 27

...
1. Capital Market Line and Security Market Line
Efficient frontier (with risk-free asset): µ∗ = |γ|σ∗ + r , where
p
|γ| = cµµ − 2rc1µ + r2 c11 , c11 = 1 T Σ −11 , c1µ = 1 T Σ −1µ , cµµ = µ T Σ −1µ

Cov(Ra ,RM ) σaM

Thm 1 (CAPM): µa = r + βa (µM − r), where βa = Var(RM )
=: σM2

Definition 1 (Capital Market Line & Security Market Line)

Capital Market Line: µ against σ (same as efficient frontier)

µ = r + |γ|σ

Security Market Line: µ against β

µ = r + (µM − r)β

µ L µ
CM L
M b M SM b

r
r a a
σ 0 1 β
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 10 / 27
...
1. Capital Market Line and Security Market Line
Capital Market Line (µ against σ): µ = r + |γ|σ
Security Market Line (µ against β): µ = r + (µM − r)β
p
|γ| = cµµ − 2rc1µ + r2 c11 = (µM − r)/σM
c11 = 1 T Σ −11 , c1µ = 1 T Σ −1µ , cµµ = µ T Σ −1µ

µ L µ
CM L
M b M SM b

r
r a a
σ 1 0 β
Capital Market Line (CML):
for each level of risk (in σ), CML gives the best achievable expected return µ
µM −r
slope of CML= σM
is called market price of risk
represents the extra return required for each additional unit of risk (in σ)
Security Market Line (SML):
for each level of risk (in β), SML gives the expected return µ of any asset
slope of SML= µM − r is called market risk premium
represents the extra return required for each additional unit of risk (in β)
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 11 / 27
...
2. Assumptions in CAPM

1 Capital markets are perfect

No taxes or transaction costs
All traders have costless access to all available information
Any single participant cannot influence market prices
2 All investors are mean-variance optimizers.
3 All investors have the same estimates of E(Ri ), σi , & σij , for all
securities i, j ∈ {1, . . . , n} in the market .
4 All investors have the common risk free rate of return r .
5 Individual assets are infinitely divisible ⇒ can hold any units

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 12 / 27

...
3. Separation Theorem
One-fund theorem: When risk-free asset is available, the efficient frontier is a linear
combination of the tangency portfolio M and the risk-free asset.

Corollary 2 (Separation Theorem)

The optimal combination of risky assets for an investor can be determined
without any knowledge of the investor’s preferences toward risk and return.
One’s choice of risky portfolio and risk preferences are separated.

Proof.
1 From CAPM Assumptions, all investors µ
B’s
arrive at the same mean-variance optimizer & tangency portfolio M .
A’s
invest in the same risk-free asset & same optimal risky portfolio M .
M
2 One’s risk preference is determined by his indifference curves: r r
each investor has his own set of curves. (e.g. two people A & B) σ
indifference curves are convex (increasing return required) due to risk aversion
points on the same line represent (σ, µ)-points that are of equal preference
higher curves (from the origin) are preferred (higher µ & lower σ)
final choice: the highest indifference curve touching efficient frontier ⇒ still same M
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 13 / 27
...
4. Market Equilibrium
CAPM model
E(Ra ) = r + βa (E(RM ) − r)
involves expected values, representing a state of equilibrium
Properties
1 Each security must have a non-zero proportion in the composition
of the tangency portfolio.
Proof: Zero proportion ⇒ no demand ⇒ price drops
⇒ expected return rises ⇒ included in tangency portfolio
2 The tangency portfolio is the market portfolio.
Market Portfolio: The weight on security i is

(m) pi ni market capitalization of security i

wi = Pn = ,
i=1
p i ni total market capitalization

where (pi , ni ) is (price per share, number of shares) of security i

Proof: Under CAPM, everyone holds the same risky (tangency) portfolio M.
⇒ The weights on the securities are the same for everyone
But the sum of everyone’s holdings is the total market capitalization
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 14 / 27
...
4. Market Equilibrium
The tangency portfolio is the market portfolio.
Market Portfolio: The weight on security i is [(pi , ni ) is (price per share, number of shares)]
p ni market capitalization of security i
wi
(m)
= Pn i = ,
pn total market capitalization
i=1 i i

Example 1
Suppose that in a market with n securities and N investors,
the tangency portfolio is w M = (w1 , . . . , wn ),
values of tangency portfolio holded by the investors are v = (v1 , . . . , vN ).
The Market Portfolio is indeed the tangency portfolio:
Value of security i holded by investor j is wi vj
PN
Total value of security i in the market is j=1
wi vj (= pi ni )
Pn PN  Pn

Total maret capitalization is i=1 j=1
wi vj = i=1
pi ni

Market portfolio’s weight on security i is

PN PN
p ni wv
j=1 i j
wi v
j=1 j
(m)
wi = Pn i = Pn PN = P
P  = wi
i=1
pi ni wv
n
w
N
vj
i=1 j=1 i j i=1 i j=1
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 15 / 27
...
5. Performance Measures

The following indexes/ratios evaluate the performance of an asset with

expected return µ
standard deviation σ
beta β

Definition 2 (Sharpe Ratio)

Sharpe Ratio= µ−r σ
µM −r
Sharpe ratio for the market portfolio σM is the Market price of risk

Definition 3 (Treynor Ratio)

Treynor Ratio= µ−r β
µM −r
Treynor ratio for the market portfolio βM = µM − r is the risk premium

The higher the ratio, the more valuable the portfolio

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 16 / 27

...
6. Systematic and Idiosyncratic risk
For any asset a and market portfolio M , the CAPM model

E(Ra ) = r + βa (E(RM ) − r)

involves expected returns, representing a state of equilibrium

In terms of   returns, we may assume a model

expected
Ra = r + βa (RM − r) + a ,
where a is an error term satisfying
E(a ) = 0
Cov(RM , a ) = 0
Var(a ) ≥ 0
σa2 = Var(Ra ) = βa2 σM
2
+ Var(a ) = systematic risk + idiosyncratic risk
systematic risk:
risk inherent to the entire market
also known as “undiversifiable risk”
idiosyncratic risk
risk inherent to an individual asset
also known as “specific risk” and “unsystematic risk”
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 17 / 27
...
Agenda

1 Introduction

2 Capital Asset Pricing Model (CAPM)

3 Economic interpretations associated with CAPM

4 Hedging with CAPM

5 Parameter Estimation for CAPM

6 Real Applications

7 References

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 18 / 27

...
Hedging with CAPM
Theorem 3 (β-hedge)
Consider hedging a portfolio Pt using a future contract of market portfolio Mt :
The future contract trades MT for M0 (1 + r) at time T with no initial cost

1)Borrow $ M0 and buy market portfolio; & 2) Hold future contract; give the same payoff

PT −P0 MT −M0
CAPM model: E(Rp ) − r = βp (E(RM ) − r), where Rp = P0
and RM = M0

The hedge of the portfolio requires shorting N = βp P0 /M0 future contracts.

Proof.
Values t=0 t=T
CAPM ⇒ PTP−P − r ≈ βp MTM−M


Pt P0 0
0
0
0
−r ,
βp P0
⇒ PT ≈ P0 (1 + r) + M 0
[MT − M0 (1 + r)]
Ft F0 = 0 FT = MT − M0 (1 + r)
β p P0
To reduce randomness (purpose of hedging) ⇒ hold − M0
future (short position)
Hedged Portfolio t=0 t=T
βp P0
Pt − M 0
Ft P0 ≈ P0 (1 + r)
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 19 / 27
...
Agenda

1 Introduction

2 Capital Asset Pricing Model (CAPM)

3 Economic interpretations associated with CAPM

4 Hedging with CAPM

5 Parameter Estimation for CAPM

6 Real Applications

7 References

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 20 / 27

...
Parameter Estimation for CAPM
CAPM: E(Ra ) = r + βa (E(RM ) − r) =⇒ Ra − r = βa (RM − r) + a

Estimation of βa by regression:
1 Data: for t = 1, . . . , T
Yt = Ra,t − rt : excess return of asset a at time t
Xt = RM,t − rt : excess return of a proxy of M at time t
Obtaining the tangency portfolio M requires the computation
Σ −1 (µ
µ − r11)
wM = ,
(c1µ − rc11 )
which is infeasible as Σ ∈ Rn×n is difficult to be estimated accurately
In practice a market index (e.g. HSI, SP500) serves as a proxy of M
rt : risk free rate at time t
2 Fit a regression in R by lm(Y∼X):
   
Y1 1 X1  
Y = X β +  , where Y =  ...  , X =  ...
   ..  , β = αa .
.  βa
Yn 1 Xn
α̂a is Jensen index, measuring excess return over CAPM forecast.
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 21 / 27
...
Agenda

1 Introduction

2 Capital Asset Pricing Model (CAPM)

3 Economic interpretations associated with CAPM

4 Hedging with CAPM

5 Parameter Estimation for CAPM

6 Real Applications

7 References

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 22 / 27

...
CAPM: R Example

CAPM ⇒ Ra − r = βa (RM − r) + a ⇒ Regress Yt = Ra,t − rt against Xt = RM,t − rt

Notes on variable names: Z[,1] = Market index level, Z[,2] = Share Price

Example 2 (A) Download available monthly security prices)

rm(list=ls(all=T))
library("tseries")
library("quantmod") # bug fix for tseries

d0="2000-02-01"; d1="2022-07-01"; Q="AdjClose" # Parameters

n <- 22 * 12 + 6 # Sample size from 22.5 years
Z <- matrix(0, n, 2) # Matrix to hold data
Z[,1] <- get.hist.quote("ˆHSI", d0, d1, Q, compression="m")
Z[,2] <- get.hist.quote("1038.HK", d0, d1, Q, compression="m")
Z # Check data

# [,1] [,2]
# [1,] 17169.44 4.294067
# [2,] 17406.54 5.765742
# [3,] 15519.30 4.959346

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 23 / 27

...
CAPM: R Example
CAPM ⇒ Ra − r = βa (RM − r) + a ⇒ Regress Yt = Ra,t − rt against Xt = RM,t − rt
Notes on variable names: rf = risk free rate, R = [RM Ra ]n×2
Example 2 (B) Regression)
rf <- 0.02/100/12 # Assumed risk-free rate per month
R <- diff(as.matrix(Z))/as.matrix(Z[-n,]) # Compute returns
Y<- R[,2]-rf # Define response
X<- R[,1]-rf # Define predictor
model <- lm(Y∼X) # Regression
summary(model) # Check outputs:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.010012 0.003409 2.937 0.0036 **
X 0.313223 0.058658 5.340 1.99e-07 ***
mu.M<- mean(R[,1]) # µ̂M = 0.00231
mu.a<- mean(R[,2]) # µ̂a = 0.0107
s.a<- sd(R[,2]) # σ̂a = 0.0587
Fitted CAPM model: E(Ra ) − r = 0.0034 + 0.313(E(RM ) − r); (r = 0.02/100/12)

Jensen index: α̂a = 0.01

CAPM β: β̂a = 0.313
µ̂a −r
µ−r


Estimated Sharpe Ratio σ
= σ̂a
= 0.18289
µ̂a −r
µ−r


Estimated Treynor Ratio β
= = 0.034256
β̂a
β-hedge: Short β̂a P0 /M0 =model$ coef[2]*Z[1,2]/Z[1,1]=7.831e-05 futures
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 24 / 27
...
Summary
Cov(Ra ,RM ) σaM
CAPM: µa = r + βa (µM − r), where βa = Var(RM )
=: σM2

In terms of returns: Ra − r = αa + βa (RM − r) + a

Decompose Var(Ra ) as systematic [Var (βa (RM − r))] & idiosyncratic [Var(a )] risks.
Estimate αa and βa by regressing {Ra,t − rt } against {RM,t − rt }
Graphs in portfolio theory
Capital Market Line (CML): µ = r + |γ|σ (µ against σ); for optimal portfolios
Security Market Line (SML): µ = β(µM − r) + r (µ against β); for all portfolios
Pricing risk:
Jensen index = α̂a
µa −r
Sharpe ratio of security a = σa
µa −r
Treynor ratio of security a = βa
Market price of risk |γ| = (µM − r)/σM
Risk premium = µM − r
Separation Theorem: Choice of risky portfolio and risk preference are separated.
The tangency portfolio is the market portfolio
β-hedge: Hedge a portfolio P by shorting N = β0 P0 /M0 future contracts of M .
RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 25 / 27
...
Agenda

1 Introduction

2 Capital Asset Pricing Model (CAPM)

3 Economic interpretations associated with CAPM

4 Hedging with CAPM

5 Parameter Estimation for CAPM

6 Real Applications

7 References

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 26 / 27

...
References

Lai, Tze Leung, and Xing, Haipeng. Statistical Models and Methods for Financial Markets.
Springer New York, 2008.

Levy, Haim. Ch.5, The Capital Asset Pricing Model. The Capital Asset Pricing Model in the
21st Century : Analytical, Empirical, and Behavioral Perspectives. New York: Cambridge
University Press, 2012. Print.

Luenberger, D.G. Investment Science, Oxford University Press, 1998

RMSC 4003 (CUHK) Ch2: Capital Asset Pricing Model 27 / 27