Asset Pricing:Asset Pricing Formula
State Price Model:
p
j
=
∑
s
=
1
S
q
s
x
s
j
p_j=\sum_{s=1}^Sq_sx_s^j
pj=s=1∑Sqsxsj
Stochastic Discount Factor:
suppose
{
π
s
}
s
=
1
S
\{\pi_s\}_{s=1}^S
{πs}s=1S is the physical probability distribution of states:
p
j
=
∑
s
=
1
S
q
s
x
s
j
=
∑
s
=
1
S
π
s
q
s
π
s
x
s
j
=
∑
s
=
1
S
π
s
m
s
x
s
j
=
E
[
m
x
j
]
m
s
≡
q
s
π
s
→
S
t
o
c
h
a
s
t
i
c
D
i
s
c
o
u
n
t
F
a
c
t
o
r
p_j=\sum_{s=1}^Sq_sx_s^j=\sum_{s=1}^S\pi_s\dfrac{q_s}{\pi_s}x_s^j=\sum_{s=1}^S\pi_sm_sx_s^j=\mathbb E[mx^j]\\m_s\equiv\dfrac{q_s}{\pi_s}\to Stochastic\ Discount\ Factor
pj=s=1∑Sqsxsj=s=1∑Sπsπsqsxsj=s=1∑Sπsmsxsj=E[mxj]ms≡πsqs→Stochastic Discount Factor
Note that:
p
j
=
E
[
m
⋅
x
j
]
=
E
[
x
j
]
E
[
m
]
+
C
o
v
(
m
,
x
j
)
r
i
s
k
−
f
r
e
e
b
o
n
d
:
p
b
=
E
[
m
]
=
1
R
f
→
R
f
=
1
E
[
m
]
,
m
=
u
′
(
c
1
)
u
′
(
c
0
)
p
j
=
E
[
x
j
]
R
f
+
C
o
v
(
m
,
x
j
)
p_j=\mathbb E[m·x_j]=\mathbb E[x_j]\mathbb E[m]+Cov(m,x_j)\\risk-free\ bond:p_b=\mathbb E[m]=\dfrac{1}{R^f}\to R^f=\dfrac{1}{\mathbb E[m]},m=\dfrac{u'(c_1)}{u'(c_0)}\\p_j=\dfrac{\mathbb E[x^j]}{R^f}+Cov(m,x^j)
pj=E[m⋅xj]=E[xj]E[m]+Cov(m,xj)risk−free bond:pb=E[m]=Rf1→Rf=E[m]1,m=u′(c0)u′(c1)pj=RfE[xj]+Cov(m,xj)
Typically,
C
o
v
(
m
,
x
j
)
<
0
Cov(m,x^j)<0
Cov(m,xj)<0.
Defining R j ≡ x j p j → E [ m R j ] = 1 R^j\equiv\dfrac{x^j}{p_j}\to \mathbb E[mR^j]=1 Rj≡pjxj→E[mRj]=1
我们有
E
[
m
R
f
]
=
1
\mathbb E[mR^f]=1
E[mRf]=1,所以:
E
[
m
⋅
(
R
j
−
R
f
)
]
=
0
E
[
m
]
(
E
[
R
j
]
−
R
f
)
+
C
o
v
(
m
,
R
j
)
=
0
E
[
R
j
]
−
R
f
=
−
C
o
v
(
m
,
R
j
)
E
[
m
]
\mathbb E[m·(R^j-R^f)]=0\\\mathbb E[m](\mathbb E[R^j]-R^f)+Cov(m,R^j)=0\\\mathbb E[R^j]-R^f=-\frac{Cov(m,R^j)}{\mathbb E[m]}
E[m⋅(Rj−Rf)]=0E[m](E[Rj]−Rf)+Cov(m,Rj)=0E[Rj]−Rf=−E[m]Cov(m,Rj)
Which implies that the Expected Excess return for a generic asset
j
j
j is determined solely by the covariance with the stochastic discount factor.
−
1
E
[
m
]
→
-\dfrac{1}{\mathbb E[m]}\to
−E[m]1→ price of market risk ,
C
o
v
(
m
,
R
j
)
→
Cov(m,R^j)\to
Cov(m,Rj)→ Quantity of risk for Asset
j
j
j
Equivalent Martingale Measure:
start with:
p
j
=
∑
s
=
1
S
q
s
x
s
j
p_j=\sum_{s=1}^Sq_sx_s^j
pj=s=1∑Sqsxsj
for a riskfree bond we have:
p
b
=
∑
s
=
1
S
q
s
=
1
1
+
r
f
p_b=\sum_{s=1}^Sq_s=\frac{1}{1+r^f}
pb=s=1∑Sqs=1+rf1
where
r
f
r^f
rf is the risk-free net return. We have:
p
j
=
∑
s
=
1
S
q
s
∑
s
=
1
S
q
s
x
s
j
∑
s
=
1
S
q
s
=
1
1
+
r
f
∑
s
=
1
S
q
s
∑
s
=
1
S
q
s
x
s
j
=
1
1
+
r
f
∑
s
=
1
S
π
^
s
x
s
j
=
1
1
+
r
f
E
Q
[
x
j
]
w
h
e
r
e
π
^
s
≡
q
s
∑
s
=
1
S
q
s
p_j=\sum_{s=1}^Sq_s\sum_{s=1}^S\frac{q_sx_s^j}{\sum_{s=1}^Sq_s}=\frac{1}{1+r^f}\sum_{s=1}^S\frac{q_s}{\sum_{s=1}^Sq_s}x_s^j=\frac{1}{1+r^f}\sum_{s=1}^S\hat\pi_sx_s^j=\frac{1}{1+r^f}\mathbb E^Q[x^j]\\where\ \hat\pi_s\equiv\frac{q_s}{\sum_{s=1}^Sq_s}
pj=s=1∑Sqss=1∑S∑s=1Sqsqsxsj=1+rf1s=1∑S∑s=1Sqsqsxsj=1+rf1s=1∑Sπ^sxsj=1+rf1EQ[xj]where π^s≡∑s=1Sqsqs
State-Price Beta Model:
stochastic discount factor:
m
∗
≡
[
q
1
∗
π
⋮
q
S
∗
π
]
m^*\equiv\left[\begin{matrix}\frac{q_1^*}{\pi}\\\vdots\\\frac{q_S^*}{\pi}\end{matrix}\right]
m∗≡⎣⎢⎢⎡πq1∗⋮πqS∗⎦⎥⎥⎤
define its return as
R
∗
=
m
∗
p
m
∗
≡
α
m
∗
,
α
>
0
R^*=\dfrac{m^*}{p_{m^*}}\equiv\alpha m^*,\alpha>0
R∗=pm∗m∗≡αm∗,α>0, we can write:
E
[
R
j
]
−
R
f
=
−
C
o
v
(
R
∗
,
R
j
)
E
[
R
∗
]
\mathbb E[R^j]-R^f=-\frac{Cov(R^*,R^j)}{\mathbb E[R^*]}
E[Rj]−Rf=−E[R∗]Cov(R∗,Rj)
define
β
j
≡
C
o
v
(
R
∗
,
R
j
)
V
a
r
(
R
∗
)
\beta_j\equiv\dfrac{Cov(R^*,R^j)}{Var(R^*)}
βj≡Var(R∗)Cov(R∗,Rj), we have:
E
[
R
j
]
−
R
f
=
−
β
j
V
a
r
(
R
∗
)
E
[
R
∗
]
\mathbb E[R^j]-R^f=-\beta_j\frac{Var(R^*)}{\mathbb E[R^*]}
E[Rj]−Rf=−βjE[R∗]Var(R∗)
and for security
x
∗
x^*
x∗, we have
β
∗
=
C
o
v
(
R
∗
,
R
∗
)
V
a
r
(
R
∗
)
=
1
\beta^*=\dfrac{Cov(R^*,R^*)}{Var(R^*)}=1
β∗=Var(R∗)Cov(R∗,R∗)=1:
E
[
R
∗
]
−
R
f
=
−
V
a
r
(
R
∗
)
E
[
R
∗
]
\mathbb E[R^*]-R^f=-\frac{Var(R^*)}{\mathbb E[R^*]}
E[R∗]−Rf=−E[R∗]Var(R∗)
so we have:
E
[
R
j
]
−
R
f
=
β
j
(
E
[
R
∗
]
−
R
f
)
\mathbb E[R^j]-R^f=\beta_j(\mathbb E[R^*]-R^f)
E[Rj]−Rf=βj(E[R∗]−Rf)