The Consumption Capital Asset

Pricing Model

Instructor: Dmytro Hryshko

1 / 30
Readings

Danthine, J. P., Donaldson, J. B. (2005), Intermediate

Financial Theory, 2nd Edition, Elsevier Academic
Press. Chapter 9.

Mehra and Prescott (2003), The Equity Premium in

Retrospect, Handbook of the Economics of Finance,
Elsevier. Chapter 14.

2 / 30
The Representative Agent Hypothesis and its Notion of
Equilibrium
An infinitely-lived Representative Agent
–Helps avoid terminal period problem
–Equivalence with finite lives if operative bequest
motive
–In complete markets, representative agent is the one
whose utility function is a weighted average of the
utilities of all the economy’s participants

No Trade Equilibrium
–Positive net supply: the representative agent willingly
holds total supply
–Zero net supply: at the prevailing price, supply =
demand =0
3 / 30
An Exchange (Endowment) Economy. Lucas 1978

Recursive trading: many periods (one period in

CAPM); investment decisions are made one period at a
time, taking due account of their impact on the future
state of the world
One perfectly divisible share
Dividend = economy’s total output, Yt
Output arises exogenously and stochastically (fruit
tree) and is perishable
Stationary stochastic process for Yt

4 / 30
Transition probability matrix (TPM)

E.g., the 3-state TPM

 
π11 π12 π13
T =  π21 π22 π23 
π31 π32 π33

with πij -th element denoting Prob(Yt+1 = Y j |Yt = Y i ), and

Y j is output in state j.
Lucas fruit tree
Rational expectations economy: knowledge of the
economy’s structure and the stochastic process

5 / 30
The problem
∞
!
X
max E δ t U (c˜t )
{zt+1 }
t=0
s.t. ct + pt zt+1 ≤ zt Yt + pt zt
zt ≤ 1, all t
F.O.C:
U 0 (ct )pt = δEt U 0 (c̃t+1 )(p̃t+1 + Ỹt+1 ) or
X
U 0 (ct (Y i ))pt (Y i ) = δ πij U 0 (ct+1 (Y j )) pt+1 (Y j ) + Yt+1 (Y j )
 
| {z }
j
utility loss | {z }
from expected utility gain at
purchasing 1 t + 1 from liquidating the
unit of sec. proceeeds and consuming
at t

6 / 30
Definition of an equilibrium. U 0 (ct )pt = δEt U 0 (c̃t+1 )(p̃t+1 + Ỹt+1 )

For the entire economy to be in equilibrium, it must,

therefore, be true that:
1 zt = zt+1 = zt+2 = . . . = 1—the representative agent
owns the entire security
2 ct = Yt —ownership of the entire security entitles the
agent to all the economy’s output
3 FOC is satisfied—the agents’ holdings of the security
are optimal given the prevailing prices
(1)–(3) imply

U 0 (Yt )pt = δEt U 0 (Ỹt+1 )(p̃t+1 + Ỹt+1 ) (*)

7 / 30
Example
Assume utility is U (c) = log c, δ = 0.96, the 3-state economy,
(Y 1 , Y 2 , Y 3 ) = (3/2, 1, 1/2). The TPM is
 
1/2 1/4 1/4
T =  1/4 1/2 1/4 
1/4 1/4 1/2
Denote p(Y i ) = p(i), then  

2 1 1 1 1 1 1
 
p(1) = 0.96( + + ) + 0.96  p(1) + p(2) + p(3)

3
|{z} 2
| 4 4}
{z |3 4 {z 2 }
=U 0 [c(Y 1 )] π1j U 0 [c(j)]Y (j) = j π1j U 0 [c(j)]p(j)
P P
= j
 
1 1 1
1p(2) = 0.96 + 0.96 p(1) + p(2) + p(3)
6 2 2
 
1 1
2p(3) = 0.96 + 0.96 p(1) + p(2) + 1p(3)
6 4
Solving the system yields p(1) = 36, p(2) = 24, p(3) = 12. 8 / 30
Equilibrium pricing
U 0 (Yt )pt = δEt U 0 (Ỹt+1 )(p̃t+1 + Ỹt+1 )

Et [U 0 (Yt+1 )pt+1 ] = Et Et+1 δU 0 (Yt+2 )Yt+2 + Et+1 δU 0 (Yt+2 )pt+2

 
| {z }
0 0
=Et δU (Yt+2 )Yt+2 + Et δU (Yt+2 )pt+2
by law of iterated expectations

Et δU 0 (Yt+1 )pt+1 = Et δ 2 U 0 (Yt+2 )Yt+2 + Et δ 2 U 0 (Yt+2 )pt+2

U 0 (Yt )pt = Et δU 0 (Yt+1 )Yt+1 + Et δ 2 U 0 (Yt+2 )Yt+2 + Et δ 2 U 0 (Yt+2 )pt+2

..
.
∞
X
0
U (Yt )pt = Et δ j U 0 (Ỹt+j )Ỹt+j + Et lim δ k U 0 (Yt+k )pt+k
j=1 | k→∞ {z }
must be=0
9 / 30
Equilibrium pricing

∞ 0
jU(Ỹt+j )
X
pt = Et δ Ỹt+j
j=1
U 0 (Yt )

asset price is the sum of all expected discounted future

dividends (on the RHS, summing starts at t + 1)

discounting is at the intertemporal marginal rate of

substitution (IMRS) of the representative agent!

10 / 30
Interpreting the Exchange Equilibrium
Define the period t to t + 1 return for security j as
pjt+1 + Yjt+1
1 + rjt+1 =
pjt
Using equation (*),
 0 
U (c̃t+1 )
1 = δEt (1 + r̃jt+1 )
U 0 (ct )
Let qtb denote the t-period price of a one-period riskless discount
bond paying 1 unit of consumption at t + 1 in every state.
 0 
U (c̃t+1 )
qtb = δEt · 1
U 0 (ct )
Since the riskfree rate is defined from qtb (1 + rf t+1 ) = 1,
 0 
1 U (c̃t+1 )
= δEt
1 + rf t+1 U 0 (ct )
Under risk-neutrality, U (ct ) is linear, rf t+1 is constant. Link
between discount factor and risk-free rate in a risk neutral
world. 11 / 30
Interpreting the Exchange Equilibrium
U 0 (c̃t+1 )
 
1 = δEt (1 + r̃jt+1 )
U 0 (ct )

U 0 (c̃t+1 )
   0 
U (c̃t+1 )
1 = δEt Et [1 + r̃jt+1 ] + δcovt , r̃jt+1
U 0 (ct ) U 0 (ct )
| {z }
=1/(1+rf t+1 )

(since cov(X, Y ) = E[XY ] − E[X]E[Y ]. Define Et [1 + r̃jt+1 ] = 1 + rjt+1 .)

U 0 (c̃t+1 )
 
1 + rjt+1
= 1 − δcovt , r̃jt+1
1 + rf t+1 U 0 (ct )

U 0 (c̃t+1 )
 
rjt+1 − rf t+1 = −δ(1 + rf t+1 )covt , r̃jt+1 .
U 0 (ct )
12 / 30
CAPM and CCAPM
U 0 (c̃t+1 )
 
CCAPM: rjt+1 − rf t+1 = −covt , r̃jt+1 δ(1 + rf t+1 )
U 0 (ct )
cov(r̃m , r˜j )
CAPM: rjt+1 − rf t+1 = (rmt+1 − rf t+1 ),
var(r̃m )
| {z }
=βj

CCAPM: asset j pays high on average, relative to the

riskfree rate, if the covariance between U 0 (c̃t+1 ) and the
return on asset j is negative and large—i.e. consumption is
low (= U 0 (c̃t+1 ) is high) when the return is low. Such an
asset will have low price pt , and high expected return
CAPM: asset j pays high on average if asset pays low when
market pays low (i.e. asset is risky).
CCAPM: Key to an asset’s value is its covariation with the
marginal utility of consumption! Consumption-smoothing
perspective 13 / 30
Towards a CAPM equation in the CCAPM framework
Let U (ct ) = act − 2b c2t , a, b > 0, a − bct > 0 for all ct .

 
a − bc̃t+1
rjt+1 − rf t+1 = −δ(1 + rf t+1 )covt , r̃jt+1
a − bct

−b
rjt+1 − rf t+1 = −δ(1 + rf t+1 ) covt (c̃t+1 , r̃jt+1 )
a − bct

δb(1 + rf t+1 )
rjt+1 − rf t+1 = covt (c̃t+1 , r̃jt+1 )
a − bct
| {z }
>0

Asset j has a high return on average if it pays low when

consumption is low, i.e. when covt (c̃t+1 , r̃jt+1 ) > 0.
14 / 30
The Formal Consumption CAPM

δb(1 + rf t+1 )
rct+1 − rf t+1 = covt (c̃t+1 , r̃ct+1 )
a − bct

“c” denotes portfolio most correlated with consumption

rjt+1 − rf t+1 covt (c̃t+1 , r̃jt+1 ) /vart (c̃t+1 ) βj,ct

= =
rct+1 − rf t+1 covt (c̃t+1 , r̃ct+1 ) /vart (c̃t+1 ) βc,ct

rjt+1 − rf t+1 = βj,ct (rct+1 − rf t+1 ) if βc,ct = 1

Compare this to the CAPM equation:

rjt+1 − rf t+1 = βj,m (rmt+1 − rf t+1 )
15 / 30
Testing the Consumption CAPM: The Equity Premium
Puzzle

mean, % s.d., %

r 6.98 16.54
rf 0.80 5.67
r − rf 6.18 16.67

Source: Data from Mehra and Prescott (1985): 1889–1978.

A reasonably parameterized CCAPM is unable to explain

r − rf (=the equity premium), the difference in ex-post
return on a diversified portfolio of U.S. stocks (S&P 500),
and the return on one-year T-bills.

16 / 30
Equity premium, annual time series
Ch 14: The Equity Premium in Retrospect 8

CD

a'
nU
.
Cr

Year-end

Fig 1 Realized equity risk premium per year, 1926-2000 Source: Ibbotson (2001).
17 / 30
 Year-end

Fig 1 Realized equity risk premium per year, 1926-2000 Source: Ibbotson (2001).
Equity premium, 20-year average

18

0
a

P
E
w
.2
a
Ec

20-Year Period Ending

Fig 2 Equity risk premium over 20-year periods, 1926-2000 Source: Ibbotson (2001).

18 / 30
Ch 14: The Equity Premium in Retrospect 895

Table 2
Equity premium in different countriesa

Country % real return on % real return % equity premium

a market index (mean) on a relatively riskless (mean)
security (mean)

UK 1947-1999 57 11 46
Japan 1970-1999 47 14 33
Germany 1978-1997 98 32 66
France 1973-1998 90 27 63

a Source: UK from Siegel (1998), the rest are from Campbell (2001).

Table 3
Terminal value of $1 invested in stocks and bondsa
19 / 30
urce: UK from Siegel (1998), the rest are from Campbell (2001).

Table 3
Terminal value of $1 invested in stocks and bondsa

Investment period Stocks T-bills

Real Nominal Real Nominal

1802-1997 $558,945 $7,470,000 $276 $3,679

1926-2000 $266 47 $2,586 52 $1 71 $16 56

a Source: Ibbotson (2001) and Siegel (1998).

he annual return on the British stock market was 5 7 % over the post-w 20 / 30
The equity premium puzzle

c1−γ
U (c) =
1−γ
Assume xt+1 ≡ ct+1ct
∼ iid lognormal(1.0183, 0.03572 ), and
so log xt+1 ∼ iidN (µx , σx2 ). Then,

1
1.0183 = exp(µx + σx2 )
2
0.03572 = exp(2µx + σx2 )[exp(σx2 − 1)]

Solving, µx = 0.01752, σx2 = 0.00123.

21 / 30
The equity premium puzzle
Assume
pt = vYt
The stock price at t is proportional to the dividend paid at
t.
Verify that it qualifies for a solution:

U 0 (c̃t+1 )
 
vYt = δEt (v Ỹt+1 + Ỹt+1 ) 0
U (ct )
 
Ỹt+1
x̃−γ
 
v = δEt 
(v + 1)

t+1 
Yt
|{z}
=x̃t+1 : Lucas tree
1−γ
δE[x̃ ]
v= indeed a constant
1 − E[x̃1−γ ]
22 / 30
The equity premium puzzle

pt+1 + Yt+1 v + 1 Yt+1 v+1

Rt+1 = 1 + rt+1 = = = xt+1
pt+1 v Yt v
v+1 E[x̃]
Et R̃t+1 = Et x̃t+1 =
v δE[x̃1−γ ]

−1
U 0 (c̃t+1 )

1 1 1
Rf t+1 = b = δEt 0 =
qt U (ct ) δ E[x̃−γ ]

Et R̃t+1 E[x̃]E[x̃−γ ] exp(µx + 0.5σx2 ) exp(−γµx + 0.5γ 2 σx2 )

= =
Rf t+1 E[x̃1−γ ] exp((1 − γ)µx + 0.5(1 − γ)2 σx2 )
= exp(γσx2 )
log(E[R]) − log(Rf ) ≈ r − rf = γσx2
23 / 30
The equity premium puzzle

r − rf = γσx2

r − rf 0.0698 − 0.008
γ= 2
= ≈ 50
σx 0.03572

If a more realistic γ = 2 is assumed, r − rf ≈ 0.2%.

24 / 30
Is γ = 50 realistic?
1−γ
Assume current wealth 50, 000, utility function u(W ) = W1−γ
and a chance to play the gamble (0,1/2; 50000,1/2). What is
the certainty equivalent (CE) of the gamble? (Π is the
maximum willing to pay to avoid playing the gamble
(–25000,1/2; 25000,1/2) with the current wealth of 75,000.)

γ CE Π
0 25,000 0
1 20,711 4,289
3 13,246 11,754
5 8,566 16,434
10 3,991 21,009
20 1,858 23,142
30 1,209 23,791
50 712 24,288
25 / 30
The risk-free rate puzzle
Similarly, if γ = 2,
1 1
δ= = ≈ 1.02
Rf t+1 E[x̃−γ ] 1.008 exp(−γµx + 0.5γ 2 σx2 )

The difficulty in explaining the low rate of return on the

riskfree asset is called the risk-free rate puzzle.
1 1
Rf t+1 = −γ
=
δE[x̃ ] δ exp(−γµx + 0.5γ 2 σx2 )

log(Rf t+1 ) = − log(δ) + γµx − 0.5γ 2 σx2

| {z } |{z} | {z }
impatience term consumption demand for
smoothing precautionary
under certainty savings

If γ = 3 and δ = 0.99, rf t+1 ≈ 6%.

26 / 30
Testing the Consumption CAPM: Hansen-Jagannathan
Bounds
 
 U 0 (c̃ ) 
t+1
1 = Et δ 0 (1 + r̃jt+1 )
 
 U (ct ) 
| {z }
≡m̃t+1
h i
1 = Et m̃t+1 R̃t+1

Should hold unconditionally:

h i
1 = E m̃t+1 R̃t+1
m̃ is called the equilibrium pricing kernel, or the stochastic
discount factor.
27 / 30
Hansen-Jagannathan bounds
h i
E m̃R̃ = 1
 

E m̃(R̃i − R̃j ) = 0
 
| {z }
≡R̃i−j

E[m̃]E[R̃i−j ] + cov(m̃, R̃i−j ) = 0

E[R̃i−j ] σm
= −ρ(m̃, R̃i−j )
σRi−j E[m̃]
|E[R̃i−j ]| σm σm
= |ρ(m̃, R̃i−j )| ≤
σRi−j E[m̃] E[m̃]
σm |E(R̃i−j )|
≥
E[m̃] σRi−j
ρ(m̃, R̃i−j ) is the correlation between the pricing kernel and
R̃i−j . 28 / 30
Hansen-Jagannathan bounds

Set i to m(arket) and j to f (riskfree asset) to obtain

σm |E(r̃m − rf )| 0.062
≥ = ≈ 0.37
E[m̃] σrm −rf 0.167

Assuming δ = 0.99 and γ = 2

E[m̃] = δx̃−γ = δ exp(−γµx + 0.5γ 2 σx2 ) ≈ 0.96

Therefore, the lowest value for σm ≈ 0.36.

Using the chosen parameters, however, σm = 0.002, an

order of magnitude lower!

29 / 30
Solutions?
Habit formation—makes IMRS, the pricing kernel in
the CCAPM, more volatile
Heterogeneity—if markets are incomplete, income risk
is persistent, and the variance of consumption growth
is countercyclical (higher in recessions than in normal
times), the model-implied equity premium will be
higher
—“. . . the risk premium is highest in a recession since equities
are a poor hedge against the potential loss of employment. . . even
though per capita consumption growth is poorly correlated with
stocks returns, investors require a hefty premium to hold stocks
over short-term bonds because stocks perform poorly in
recessions, when an investor is more likely to be laid off.” Mehra
& Prescott (2003).
Limited stock-market participation—the relevant
stochastic discount factor is IMRS of stockholders
whose consumption growth is more volatile 30 / 30