C-CAPM模型 – 潘登同学的Quant笔记
C-CAPM与CAPM的区别
CAPM | C-CAPM | |
---|---|---|
perference | M-V efficiency | Expected Utility |
Decision | portfolio Optimization | Decision under uncertainty |
Equilibrium | Partical(captial market) | General(whole economy) |
Asset pricing | CAPM | C-CAPM |
期望效用理论
理性偏好
一个理性的偏好需要满足两个特点
- Compete(完备性): ∀ x , y ∈ X , x ≿ y o r y ≿ x \forall x,y \in \Chi, x \succsim y \quad or \quad y \succsim x ∀x,y∈X,x≿yory≿x
- transitivity(传递性): x ≿ y , y ≿ z ⇒ x ≿ z x \succsim y, y \succsim z \Rightarrow x \succsim z x≿y,y≿z⇒x≿z
对于均方有效准则来说,就不满足完备性,因为在均值方差都不同的时候(A组合的均值小于B组合,但A组合的方差小于B组合),均方有效准则就不能给出谁更优的判断了;
效用函数
如果一个理性偏好还满足连续性,就能用一个连续的效用函数来描述这种偏好
- 连续性: 有一系列 { ( x n , y n } n = 1 ∞ , 有 x n ≿ y n ( ∀ n ) , 那么对 x = l i m n → ∞ x n 与 y = l i m n → ∞ y n , 必然有 x ≿ y \{(x^n,y^n\}_{n=1}^{\infty},有x^n\succsim y^n(\forall n),那么对x = lim_{n\to \infty} x^n与y = lim_{n\to \infty} y^n,必然有x\succsim y {(xn,yn}n=1∞,有xn≿yn(∀n),那么对x=limn→∞xn与y=limn→∞yn,必然有x≿y
如果一个偏好是理性且连续的,那么就可以用一个连续的函数 u ( x ) u(x) u(x)来表示。
将效用拓展到不确定情况下
将前面的确定性情况下的偏好和效用理论拓展到不确定状况,有关键的三步
- Modelling choiceset under uncertainty
- Modelling Perference under uncertainty
- Finding Expected utility function
Modelling choiceset under uncertainty
用lottery(彩票)的形式来刻画,未来的可能结果
- Simple lottery(简单彩票): L = ( p 1 , … , p N ) , p i ≥ 0 , ∑ n = 1 N p n = 1 L = (p_1,\ldots,p_N), p_i \geq 0, \sum_{n=1}^N p_n = 1 L=(p1,…,pN),pi≥0,∑n=1Npn=1
- Compound lotter(复合彩票): 有K张Simple lottery L k = ( p 1 k , … , p N k ) , k = 1 , … , K , 概率 α k ≥ 0 L_k = (p_1^k,\ldots,p_N^k),k=1,\ldots,K,概率\alpha_k\geq 0 Lk=(p1k,…,pNk),k=1,…,K,概率αk≥0,复合彩票 ( L 1 , … , L K ; α 1 , … , α K ) (L_1,\ldots,L_K;\alpha_1,\ldots,\alpha_K) (L1,…,LK;α1,…,αK)表示为以 α k \alpha_k αk的概率产生结果 L k L_k Lk; (很显然复合彩票可以化为简单彩票)
所以我们只研究Simple lottery,并把所有简单彩票所组成的集合叫做彩票空间 L \mathcal{L} L;
Modelling Perference under uncertainty
Independence axiom(独立性公理):
A
,
B
,
C
∈
L
,
∀
α
∈
(
0
,
1
)
A
≿
B
⇔
α
A
+
(
1
−
α
)
C
≿
α
B
+
(
1
−
α
)
C
A,B,C\in \mathcal{L}, \quad \forall \alpha \in (0, 1) \\ A \succsim B \Leftrightarrow \alpha A + (1-\alpha)C \succsim \alpha B + (1-\alpha) C
A,B,C∈L,∀α∈(0,1)A≿B⇔αA+(1−α)C≿αB+(1−α)C
实际上,这是一个比较强的假设,即假设A、C之间,B、C之间没有“化学反应”
期望效用定理: 在
L
\mathcal{L}
L上, Perference满足(compete,transitivity,continuous,Independence axion),那么这种偏好就能表示成期望效用的形式
U
(
L
)
=
∑
n
=
1
N
p
n
u
(
x
n
)
U(L) = \sum_{n=1}^Np_nu(x_n)
U(L)=n=1∑Npnu(xn)
风险厌恶的度量
在图中,不确定性表现为未来消费的波动
从图中可以得到两点结论
- 引入不确定性后,效用下降了 E u ( c ± △ ) < E u ( c ) Eu(c\pm\triangle)<Eu(c) Eu(c±△)<Eu(c)
- 边际效用递减地越快越是风险厌恶者,风险溢价就要更多(为了消除风险而支付的)
Coefficient of Absolute Risk Aversion(ARA)
a gamble(固定大小) : 以 π \pi π的概率赢得数额为h的钱,反之则输掉h的钱; 对于不同风险厌恶程度的人来说, π \pi π的数额至关重要,所以可以用 π \pi π对风险厌恶程度进行度量
u ( y ) = π ∗ u ( y + h ) + ( 1 − π ∗ ) π ( y − h ) = T a y l o r ( 在 y 处展开 ) π ∗ [ u ( y ) + h u ′ ( y ) + h 2 2 u ′ ′ ( y ) ] + ( 1 − π ∗ ) [ u ( y ) − h u ′ ( y ) + h 2 2 u ′ ′ ( y ) ] ⇛ 0 = ( 2 π ∗ − 1 ) h u ′ ( y ) + h 2 2 u ′ ′ ( y ) ⇛ π ∗ = 1 2 + h 4 [ − u ′ ′ ( y ) u ′ ( y ) ] \begin{aligned} u(y) &= \pi^*u(y+h) + (1-\pi^*)\pi(y-h)\\ &\overset{Taylor(在y处展开)}{=} \pi^*[u(y)+hu'(y)+\frac{h^2}{2}u''(y)] + (1-\pi^*)[u(y)-hu'(y)+\frac{h^2}{2}u''(y)] \\ \Rrightarrow 0&=(2\pi^*-1)hu'(y) + \frac{h^2}{2}u''(y) \\ \Rrightarrow \pi^*&=\frac{1}{2} + \frac{h}{4}[-\frac{u''(y)}{u'(y)}] \end{aligned} u(y)⇛0⇛π∗=π∗u(y+h)+(1−π∗)π(y−h)=Taylor(在y处展开)π∗[u(y)+hu′(y)+2h2u′′(y)]+(1−π∗)[u(y)−hu′(y)+2h2u′′(y)]=(2π∗−1)hu′(y)+2h2u′′(y)=21+4h[−u′(y)u′′(y)]
定义绝对风险厌恶系数ARA
R
A
(
y
)
≡
−
u
′
′
(
y
)
u
′
(
y
)
R_A(y) \equiv -\frac{u''(y)}{u'(y)}
RA(y)≡−u′(y)u′′(y)
- 对 u ′ ′ ( y ) u''(y) u′′(y)表示效用的二阶导,二阶导越大,边际效用的下降的速度就越快,风险厌恶程度越大
- 对 u ′ ( y ) u'(y) u′(y)取决于收入水平,收入水平越低值越大,风险厌恶程度越小
Coefficient of Relative Risk Aversion(RRA)
a gamble(固定比例) : 以 π \pi π的概率赢得数额为 θ y \theta y θy的钱,反之则输掉 θ y \theta y θy的钱;
u ( y ) = π ∗ u ( y + θ y ) + ( 1 − π ∗ ) π ( y − θ y ) = T a y l o r ( 在 y 处展开 ) π ∗ [ u ( y ) + θ y u ′ ( y ) + θ 2 2 y 2 u ′ ′ ( y ) ] + ( 1 − π ∗ ) [ u ( y ) − θ y u ′ ( y ) + θ 2 2 y 2 u ′ ′ ( y ) ] ⇛ 0 = ( 2 π ∗ − 1 ) θ y u ′ ( y ) + θ 2 2 y 2 u ′ ′ ( y ) ⇛ π ∗ = 1 2 + θ 4 [ − y u ′ ′ ( y ) u ′ ( y ) ] \begin{aligned} u(y) &= \pi^*u(y+\theta y) + (1-\pi^*)\pi(y-\theta y)\\ &\overset{Taylor(在y处展开)}{=} \pi^*[u(y)+\theta yu'(y)+\frac{\theta^2}{2}y^2u''(y)] + (1-\pi^*)[u(y)-\theta yu'(y)+\frac{\theta^2}{2}y^2u''(y)] \\ \Rrightarrow 0&=(2\pi^*-1)\theta yu'(y) + \frac{\theta^2}{2}y^2u''(y) \\ \Rrightarrow \pi^*&=\frac{1}{2} + \frac{\theta}{4}[-\frac{yu''(y)}{u'(y)}] \end{aligned} u(y)⇛0⇛π∗=π∗u(y+θy)+(1−π∗)π(y−θy)=Taylor(在y处展开)π∗[u(y)+θyu′(y)+2θ2y2u′′(y)]+(1−π∗)[u(y)−θyu′(y)+2θ2y2u′′(y)]=(2π∗−1)θyu′(y)+2θ2y2u′′(y)=21+4θ[−u′(y)yu′′(y)]
定义相对风险厌恶系数RRA
R
A
(
y
)
≡
−
y
u
′
′
(
y
)
u
′
(
y
)
R_A(y) \equiv -\frac{yu''(y)}{u'(y)}
RA(y)≡−u′(y)yu′′(y)
常见的效用函数
- CARA: ARA为常数
U ( c ) = − e − α c R A ( c ) = α U(c) = -e^{-\alpha c} \\ R_A(c) = \alpha U(c)=−e−αcRA(c)=α - CRRA: RRA为常数
U ( c ) = c 1 − γ − 1 1 − γ R A ( c ) = γ U(c) = \frac{c^{1-\gamma}-1}{1-\gamma} \\ R_A(c) = \gamma U(c)=1−γc1−γ−1RA(c)=γ
当 γ = 1 \gamma=1 γ=1时,CRRA函数退化为对数效用函数 u ( c ) = l n c u(c)=lnc u(c)=lnc - linear:
U ( c ) = α c U(c) = \alpha c U(c)=αc
一旦是liner就看不出风险厌恶了,至少展成二阶才能解释风险厌恶 - 二次型效用或CARA+lognormal return
U ( r ) = E ( r ) − A σ 2 ( r ) U(r) = E(r) - A\sigma^2(r) U(r)=E(r)−Aσ2(r)
用期望效用研究人面对风险的行为
投资组合优化问题(I)
结论: 无论一个人的风险厌恶程度多高,他都会投资一部分在风险资产上
无风险资产 | 风险资产 | |
---|---|---|
收益率 | r f r_f rf | r ~ \tilde{r} r~ |
投资数额 | w 0 − a w_0-a w0−a | a a a |
对于投资者来说
max
a
E
[
u
(
w
~
)
]
=
max
a
E
{
u
[
w
0
(
1
+
r
f
)
+
a
(
r
~
−
r
f
)
]
}
F
O
C
:
E
{
u
′
[
w
0
(
1
+
r
f
)
+
a
(
r
~
−
r
f
)
]
(
r
~
−
r
f
)
}
=
0
\max_a E[u(\tilde{w})] = \max_a E\{u[w_0(1+r_f)+a(\tilde{r}-r_f)]\}\\ FOC: E\{u'[w_0(1+r_f)+a(\tilde{r}-r_f)](\tilde{r}-r_f)\} = 0
amaxE[u(w~)]=amaxE{u[w0(1+rf)+a(r~−rf)]}FOC:E{u′[w0(1+rf)+a(r~−rf)](r~−rf)}=0
下证:
a
∗
>
0
⇔
E
(
r
~
)
>
r
f
a^*>0 \Leftrightarrow E(\tilde{r}) > r_f
a∗>0⇔E(r~)>rf
(
充分性
)
令
V
(
a
)
=
E
{
u
[
w
0
(
1
+
r
f
)
+
a
(
r
~
−
r
f
)
]
}
F
O
C
:
V
′
(
a
∗
)
=
0
(
a
∗
的定义
)
V
′
′
(
a
∗
)
=
E
{
u
′
′
[
w
0
(
1
+
r
f
)
+
a
∗
(
r
~
−
r
f
)
]
(
r
~
−
r
f
)
2
}
<
0
(
u
′
′
(
⋅
)
<
0
)
∴
V
′
(
a
∗
)
是减函数
又
∵
a
∗
>
0
∴
V
′
(
0
)
>
V
′
(
a
∗
)
=
0
V
′
(
0
)
=
E
{
u
′
[
w
0
(
1
+
r
f
)
]
(
r
~
−
r
f
)
}
>
0
=
u
′
[
w
0
(
1
+
r
f
)
]
E
(
r
~
−
r
f
)
>
0
又
∵
u
′
(
⋅
)
>
0
,
∴
E
(
r
~
−
r
f
)
>
0
⇛
E
(
r
~
)
>
r
f
(
必要性
)
简单的将上面的过程从后往前即可
(
V
′
(
a
∗
)
是减函数的部分先说明,其余从后往前
)
\begin{aligned} (充分性) &\qquad 令V(a) = E\{u[w_0(1+r_f)+a(\tilde{r}-r_f)]\}\\ FOC&: V'(a^*) = 0 \quad(a^*的定义) \\ V''(a^*) &= E\{u''[w_0(1+r_f)+a^*(\tilde{r}-r_f)](\tilde{r}-r_f)^2\} <0 \quad (u''(\cdot)<0)\\ \therefore &V'(a^*)是减函数\\ 又\because& a^* > 0 \quad \therefore V'(0) > V'(a^*) = 0 \\ V'(0) &= E\{u'[w_0(1+r_f)](\tilde{r}-r_f)\} > 0 \\ &=u'[w_0(1+r_f)]E(\tilde{r}-r_f) > 0 \\ 又\because &u'(\cdot) > 0, \quad \therefore E(\tilde{r}-r_f) > 0 \Rrightarrow E(\tilde{r}) > r_f \\ (必要性)& 简单的将上面的过程从后往前即可 \\ &(V'(a^*)是减函数的部分先说明,其余从后往前) \end{aligned}
(充分性)FOCV′′(a∗)∴又∵V′(0)又∵(必要性)令V(a)=E{u[w0(1+rf)+a(r~−rf)]}:V′(a∗)=0(a∗的定义)=E{u′′[w0(1+rf)+a∗(r~−rf)](r~−rf)2}<0(u′′(⋅)<0)V′(a∗)是减函数a∗>0∴V′(0)>V′(a∗)=0=E{u′[w0(1+rf)](r~−rf)}>0=u′[w0(1+rf)]E(r~−rf)>0u′(⋅)>0,∴E(r~−rf)>0⇛E(r~)>rf简单的将上面的过程从后往前即可(V′(a∗)是减函数的部分先说明,其余从后往前)
这个命题的结论比它看起来更震撼。它说的是,只要风险资产的期望收益率高于无风险利率(哪怕幅度非常微小),风险厌恶的投资者就会愿意买入风险资产(而不管投资者的风险厌恶程度是多么的高);
Arrow-Pratt 近似
接着上面的结论,Arrow-Pratt 近似给出了风险资产对一个完全持有无风险资产的人来说的两方面的影响
- E U ∝ a EU \propto a EU∝a(因为 E ( r ~ ) > r f E(\tilde{r}) > r_f E(r~)>rf)
- E U ∝ 1 a 2 EU \propto \frac{1}{a^2} EU∝a21
当 a 2 a^2 a2很小的时候, a 2 < a a^2 < a a2<a,所以只要 E ( r ~ ) > r f E(\tilde{r}) > r_f E(r~)>rf,所有人都会持有一些风险资产;
下面给出相关证明:
- 假设 x ~ \tilde{x} x~一个均值为零的随机变量, E x ~ = 0 E\tilde{x}=0 Ex~=0,表示风险
- k k k是一个正的常数,用于调节风险 k x ~ k\tilde{x} kx~的大小
- 假设人的初始财富为 w 0 w_0 w0, g ( k ) g(k) g(k)为在这个初始财富上,对应 k x ~ k\tilde{x} kx~的风险溢价
根据上面的不确定条件上的期望效用的图,很容易写出如下等式(风险溢价的定义)
E
u
(
w
0
+
k
x
~
)
=
u
(
w
0
−
g
(
k
)
)
Eu(w_0+k\tilde{x}) = u(w_0-g(k))
Eu(w0+kx~)=u(w0−g(k))
对上式左右两边对k求二阶导
E
[
x
~
u
′
(
w
0
+
k
x
~
)
]
=
−
g
′
(
k
)
u
′
(
w
0
−
g
(
k
)
)
(
1
)
E
[
x
~
2
u
′
′
(
w
0
+
k
x
~
)
]
=
−
g
′
′
(
k
)
u
′
(
w
0
−
g
(
k
)
)
+
[
g
′
(
k
)
]
2
u
′
′
(
w
0
−
g
(
k
)
)
(
2
)
\begin{aligned} E[\tilde{x}u'(w_0+k\tilde{x})] &= -g'(k)u'(w_0-g(k)) \qquad (1) \\ E[\tilde{x}^2u''(w_0+k\tilde{x})] &= -g''(k)u'(w_0-g(k)) + [g'(k)]^2u''(w_0-g(k)) \qquad (2)\\ \end{aligned}
E[x~u′(w0+kx~)]E[x~2u′′(w0+kx~)]=−g′(k)u′(w0−g(k))(1)=−g′′(k)u′(w0−g(k))+[g′(k)]2u′′(w0−g(k))(2)
当
k
=
0
,
g
(
0
)
=
0
,
g
′
(
0
)
=
0
(
由
(
1
)
式
)
k=0,g(0)=0,g'(0)=0(由(1)式)
k=0,g(0)=0,g′(0)=0(由(1)式),将0代入(2)式,得
g
′
′
(
0
)
=
−
u
′
′
(
w
0
)
u
′
(
w
0
)
E
x
~
2
g''(0) = -\frac{u''(w_0)}{u'(w_0)}E\tilde{x}^2
g′′(0)=−u′(w0)u′′(w0)Ex~2
接着对
g
(
k
)
在
k
=
0
g(k)在k=0
g(k)在k=0处做泰勒展开,展到二次项
g
(
k
)
≈
g
(
0
)
+
k
g
′
(
0
)
+
1
2
k
2
g
′
′
(
0
)
=
1
2
k
2
R
A
(
w
0
)
E
x
~
2
\begin{aligned} g(k) &\approx g(0) + kg'(0) + \frac{1}{2}k^2g''(0) \\ &= \frac{1}{2}k^2R_A(w_0)E\tilde{x}^2 \end{aligned}
g(k)≈g(0)+kg′(0)+21k2g′′(0)=21k2RA(w0)Ex~2
回到投资组合优化问题(I)
无风险资产 | 风险资产 | |
---|---|---|
收益率 | r f r_f rf | r ~ \tilde{r} r~ |
投资数额 | w 0 − a w_0-a w0−a | a a a |
- 投资者期末的财富为:
w ~ = w 0 ( 1 + r f ) + a ( r ~ − r f ) = w 0 ( 1 + r f ) + a ( μ − r f ) + a ( r ~ − μ ) = w ˉ + a ϵ ~ \begin{aligned} \tilde{w} &= w_0(1+r_f) + a(\tilde{r} - r_f) \\ &= w_0(1+r_f) + a(\mu-r_f) + a(\tilde{r} - \mu) \\ &= \bar{w} + a\tilde{\epsilon} \end{aligned} w~=w0(1+rf)+a(r~−rf)=w0(1+rf)+a(μ−rf)+a(r~−μ)=wˉ+aϵ~
其中 μ = E [ r ~ ] , w ˉ = w 0 ( 1 + r f ) + a ( μ − r f ) , ϵ ~ = r ~ − μ , E [ ϵ ~ ] = 0 \mu=E[\tilde{r}] ,\quad \bar{w} = w_0(1+r_f) + a(\mu-r_f) ,\quad \tilde{\epsilon}=\tilde{r} - \mu,\quad E[\tilde{\epsilon}] = 0 μ=E[r~],wˉ=w0(1+rf)+a(μ−rf),ϵ~=r~−μ,E[ϵ~]=0
E u ( w ˉ + a ϵ ~ ) = u ( w ˉ − g ( a ) ) ( ∗ ) ⇒ g ( a ) ≈ 1 2 a 2 R A ( w ˉ ) E ϵ ~ 2 ⇒ g ( a ) a 2 ≈ 1 2 R A ( w ˉ ) E ϵ ~ 2 ( 常数 ) Eu(\bar{w} + a\tilde{\epsilon}) = u(\bar{w} - g(a)) \quad (*)\\ \Rightarrow g(a) \approx \frac{1}{2}a^2R_A(\bar{w})E\tilde{\epsilon}^2 \\ \Rightarrow \frac{g(a)}{a^2} \approx \frac{1}{2}R_A(\bar{w})E\tilde{\epsilon}^2 \quad(常数) \\ Eu(wˉ+aϵ~)=u(wˉ−g(a))(∗)⇒g(a)≈21a2RA(wˉ)Eϵ~2⇒a2g(a)≈21RA(wˉ)Eϵ~2(常数)
当 a → 0 a\to 0 a→0时,有以下极限关系式成立
a ( μ − r f ) − g ( a ) a 2 = μ − r f a − g ( a ) a − → + ∞ \frac{a(\mu-r_f)-g(a)}{a^2} = \frac{\mu-r_f}{a} - \frac{g(a)}{a} -\to +\infty a2a(μ−rf)−g(a)=aμ−rf−ag(a)−→+∞
因此当 a a a是足够小的正数时候,必然有 a ( μ − r f ) − g ( a ) > 0 a(\mu-r_f)-g(a)>0 a(μ−rf)−g(a)>0,有以下不等式成立
w ˉ − g ( a ) = w 0 ( 1 + r f ) + a ( μ − r f ) − g ( a ) > w 0 ( 1 + r f ) \bar{w} - g(a) = w_0(1+r_f) + a(\mu-r_f) - g(a) > w_0(1+r_f) wˉ−g(a)=w0(1+rf)+a(μ−rf)−g(a)>w0(1+rf)
因此有
u ( w 0 ( 1 + r f ) ) < u ( w ˉ − g ( a ) ) = E u ( w ˉ + a ϵ ~ ) = E u ( w ~ ) u(w_0(1+r_f)) < u(\bar{w} - g(a)) = Eu(\bar{w} + a\tilde{\epsilon}) = Eu(\tilde{w}) u(w0(1+rf))<u(wˉ−g(a))=Eu(wˉ+aϵ~)=Eu(w~)
所以只要 E ( r ~ ) > r f E(\tilde{r}) > r_f E(r~)>rf,将一部分( a a a)资产投在无风险资产上是有利可图的; 接着稍微改写一下期末资产的确定性等值, ( ∗ ) (*) (∗)左边是不确定值,右边是确定值
w ˉ − g ( a ) = w 0 ( 1 + r f ) + a ( μ − r f ) − g ( a ) ≈ w 0 ( 1 + r f ) + a ( μ − r f ) − 1 2 a 2 R A ( w ˉ ) E ϵ ~ 2 \begin{aligned} \bar{w} - g(a) & = w_0(1+r_f) + a(\mu-r_f) - g(a) \\ &\approx w_0(1+r_f) + a(\mu-r_f) - \frac{1}{2}a^2R_A(\bar{w})E\tilde{\epsilon}^2 \end{aligned} wˉ−g(a)=w0(1+rf)+a(μ−rf)−g(a)≈w0(1+rf)+a(μ−rf)−21a2RA(wˉ)Eϵ~2
这个式子自然验证了 - E U ∝ a EU \propto a EU∝a
- E U ∝ 1 a 2 EU \propto \frac{1}{a^2} EU∝a21
投资组合优化问题(II) – 定性分析
由**投资组合优化问题(I)**的一阶条件得知,
a
∗
a^*
a∗是其初始函数
w
0
w_0
w0的函数
F
O
C
:
E
{
u
′
[
w
0
(
1
+
r
f
)
+
a
(
r
~
−
r
f
)
]
(
r
~
−
r
f
)
}
=
0
FOC: E\{u'[w_0(1+r_f)+a(\tilde{r}-r_f)](\tilde{r}-r_f)\} = 0
FOC:E{u′[w0(1+rf)+a(r~−rf)](r~−rf)}=0
接下来研究,不同初始财富对风险投资量
a
∗
a^*
a∗的影响;
结论:随着财富增加,ARA(绝对风险厌恶系数)递减 R A ′ ( w 0 ) < 0 ⇔ a ∗ ′ ( w 0 ) > 0 R_A'(w_0)<0 \Leftrightarrow {a^{*}}'(w_0)>0 RA′(w0)<0⇔a∗′(w0)>0 投到风险资产上的资产增加
下面给出相关证明:
( 充分性 ) R A ′ ( w 0 ) < 0 ⇒ a ∗ ′ ( w 0 ) > 0 F O C : E { u ′ [ w 0 ( 1 + r f ) + a ∗ ( r ~ − r f ) ] ( r ~ − r f ) } = 0 对 w 0 求导 : E { u ′ ′ [ w 0 ( 1 + r f ) + a ∗ ( r ~ − r f ) ] ( r ~ − r f ) [ ( 1 + r f ) + ( r ~ − r f ) d a ∗ d w 0 ] } = 0 将 : w 0 ( 1 + r f ) + a ∗ ( r ~ − r f ) 记为 w ~ 上式为 : ( 1 + r f ) E [ u ′ ′ ( w ~ ) ( r ~ − r f ) ] + E [ u ′ ′ ( w ~ ) ( r ~ − r f ) 2 d a ∗ d w 0 ] = 0 (充分性) \quad R_A'(w_0)<0 \Rightarrow {a^{*}}'(w_0)>0 \\ \begin{aligned} FOC:& E\{u'[w_0(1+r_f)+a^*(\tilde{r}-r_f)](\tilde{r}-r_f)\} = 0 \\ 对w_0求导:& E\{u''[w_0(1+r_f)+a^*(\tilde{r}-r_f)](\tilde{r}-r_f)[(1+r_f)+(\tilde{r}-r_f)\frac{da^*}{dw_0}]\} = 0 \\ 将:& w_0(1+r_f)+a^*(\tilde{r}-r_f) 记为 \tilde{w} \\ 上式为:& (1+r_f) E[u''(\tilde{w})(\tilde{r}-r_f)]+E[u''(\tilde{w})(\tilde{r}-r_f)^2 \frac{da^*}{dw_0}] = 0 \\ \end{aligned} (充分性)RA′(w0)<0⇒a∗′(w0)>0FOC:对w0求导:将:上式为:E{u′[w0(1+rf)+a∗(r~−rf)](r~−rf)}=0E{u′′[w0(1+rf)+a∗(r~−rf)](r~−rf)[(1+rf)+(r~−rf)dw0da∗]}=0w0(1+rf)+a∗(r~−rf)记为w~(1+rf)E[u′′(w~)(r~−rf)]+E[u′′(w~)(r~−rf)2dw0da∗]=0
由于
d
a
∗
d
w
0
\frac{da^*}{dw_0}
dw0da∗不是随机变量(因为
a
∗
a^*
a∗是事前决定的)
d
a
∗
d
w
0
=
−
(
1
+
r
f
)
E
[
u
′
′
(
w
~
)
(
r
~
−
r
f
)
]
E
[
u
′
′
(
w
~
)
(
r
~
−
r
f
)
2
]
\frac{da^*}{dw_0} = -\frac{(1+r_f) E[u''(\tilde{w})(\tilde{r}-r_f)]}{E[u''(\tilde{w})(\tilde{r}-r_f)^2]}
dw0da∗=−E[u′′(w~)(r~−rf)2](1+rf)E[u′′(w~)(r~−rf)]
- ( 1 + r f ) > 0 (1+r_f)>0 (1+rf)>0
- u ′ ′ ( ⋅ ) < 0 u''(\cdot)<0 u′′(⋅)<0
- ( r ~ − r f ) 2 > 0 (\tilde{r}-r_f)^2>0 (r~−rf)2>0
d
a
∗
d
w
0
的符号
=
E
[
u
′
′
(
w
~
)
(
r
~
−
r
f
)
]
的符号
\frac{da^*}{dw_0} 的符号 = E[u''(\tilde{w})(\tilde{r}-r_f)] 的符号
dw0da∗的符号=E[u′′(w~)(r~−rf)]的符号
用ARA替换
u
′
′
(
w
~
)
u''(\tilde{w})
u′′(w~),
R
A
(
y
)
≡
−
u
′
′
(
y
)
u
′
(
y
)
R_A(y) \equiv -\frac{u''(y)}{u'(y)}
RA(y)≡−u′(y)u′′(y)
E
[
u
′
′
(
w
~
)
(
r
~
−
r
f
)
]
=
E
[
−
u
′
(
w
~
)
R
A
(
w
~
)
(
r
~
−
r
f
)
]
=
∑
n
=
1
N
P
n
(
−
u
′
(
w
n
)
R
A
(
w
n
)
(
r
n
−
r
f
)
)
\begin{aligned} E[u''(\tilde{w})(\tilde{r}-r_f)] &= E[-u'(\tilde{w})R_A(\tilde{w})(\tilde{r}-r_f)] \\ &= \sum_{n=1}^N P_n (-u'(w_n)R_A(w_n)(r_n-r_f)) \end{aligned}
E[u′′(w~)(r~−rf)]=E[−u′(w~)RA(w~)(r~−rf)]=n=1∑NPn(−u′(wn)RA(wn)(rn−rf))
其中,
w
n
w_n
wn为第n种情况下,投资者期末的财富值,
r
n
r_n
rn为第n种情况下的风险资产回报率; 接下来对
r
n
r_n
rn分类讨论
-
当 r n ≥ r f r_n \geq r_f rn≥rf时,
∵ a ∗ > 0 , 则必有 w n ≥ w 0 ( 1 + r f ) 又 ∵ R A ′ ( w 0 ) < 0 , R A ( w n ) ( r n − r f ) ≤ R A ( w 0 ( 1 + r f ) ) ( r n − r f ) ( ∗ 1 ) \because a^* > 0, 则必有 w_n \geq w_0(1+r_f) \\ 又\because R_A'(w_0)<0, \\ R_A(w_n)(r_n-r_f) \leq R_A(w_0(1+r_f))(r_n-r_f) \qquad (*1) ∵a∗>0,则必有wn≥w0(1+rf)又∵RA′(w0)<0,RA(wn)(rn−rf)≤RA(w0(1+rf))(rn−rf)(∗1) -
当 r n < r f r_n < r_f rn<rf时,
∵ a ∗ > 0 , 则必有 w n < w 0 ( 1 + r f ) 又 ∵ R A ′ ( w 0 ) < 0 , R A ( w n ) ( r n − r f ) ≤ R A ( w 0 ( 1 + r f ) ) ( r n − r f ) ( ∗ 2 ) \because a^* > 0, 则必有 w_n < w_0(1+r_f) \\ 又\because R_A'(w_0)<0, \\ R_A(w_n)(r_n-r_f) \leq R_A(w_0(1+r_f))(r_n-r_f) \qquad (*2) ∵a∗>0,则必有wn<w0(1+rf)又∵RA′(w0)<0,RA(wn)(rn−rf)≤RA(w0(1+rf))(rn−rf)(∗2)
对
(
∗
1
)
与
(
∗
2
)
(*1)与(*2)
(∗1)与(∗2) 加上边际效用大于零
u
′
(
w
)
>
0
u'(w)>0
u′(w)>0
(
−
u
′
(
w
n
)
)
R
A
(
w
n
)
(
r
n
−
r
f
)
≥
(
−
u
′
(
w
n
)
)
R
A
(
w
0
(
1
+
r
f
)
)
(
r
n
−
r
f
)
又
∵
r
n
=
r
f
不恒成立,上面不等号取严格不等号
⇒
∑
n
=
1
N
P
n
(
−
u
′
(
w
n
)
R
A
(
w
n
)
(
r
n
−
r
f
)
)
>
∑
n
=
1
N
P
n
(
−
u
′
(
w
n
)
)
R
A
(
w
0
(
1
+
r
f
)
)
(
r
n
−
r
f
)
=
E
[
(
−
u
′
(
w
~
)
)
R
A
(
w
0
(
1
+
r
f
)
)
(
r
~
−
r
f
)
]
=
−
R
A
(
w
0
(
1
+
r
f
)
)
E
[
u
′
(
w
~
)
(
r
~
−
r
f
)
]
=
0
(
E
[
u
′
(
w
~
)
(
r
~
−
r
f
)
]
=
0
是一阶条件
)
(-u'(w_n))R_A(w_n)(r_n-r_f) \geq (-u'(w_n))R_A(w_0(1+r_f))(r_n-r_f) \\ 又\because r_n = r_f不恒成立,上面不等号取严格不等号 \\ \begin{aligned} \Rightarrow \sum_{n=1}^N P_n (-u'(w_n)R_A(w_n)(r_n-r_f)) &> \sum_{n=1}^N P_n(-u'(w_n))R_A(w_0(1+r_f))(r_n-r_f) \\ &= E[(-u'(\tilde{w}))R_A(w_0(1+r_f))(\tilde{r}-r_f)] \\ &= -R_A(w_0(1+r_f))E[u'(\tilde{w})(\tilde{r}-r_f)] \\ &= 0 \qquad (E[u'(\tilde{w})(\tilde{r}-r_f)]=0是一阶条件) \end{aligned}
(−u′(wn))RA(wn)(rn−rf)≥(−u′(wn))RA(w0(1+rf))(rn−rf)又∵rn=rf不恒成立,上面不等号取严格不等号⇒n=1∑NPn(−u′(wn)RA(wn)(rn−rf))>n=1∑NPn(−u′(wn))RA(w0(1+rf))(rn−rf)=E[(−u′(w~))RA(w0(1+rf))(r~−rf)]=−RA(w0(1+rf))E[u′(w~)(r~−rf)]=0(E[u′(w~)(r~−rf)]=0是一阶条件)
所以, d a ∗ d w 0 的符号 = E [ u ′ ′ ( w ~ ) ( r ~ − r f ) ] 的符号 = + \frac{da^*}{dw_0} 的符号 = E[u''(\tilde{w})(\tilde{r}-r_f)] 的符号 = + dw0da∗的符号=E[u′′(w~)(r~−rf)]的符号=+,所以 a ∗ ′ ( w 0 ) > 0 {a^{*}}'(w_0)>0 a∗′(w0)>0;
必要性的证明也用上面的框架,稍微改一下就能得证…
除了DARA的结论: R A ′ ( w 0 ) < 0 ⇔ a ∗ ′ ( w 0 ) > 0 R_A'(w_0)<0 \Leftrightarrow {a^{*}}'(w_0)>0 RA′(w0)<0⇔a∗′(w0)>0,还有类似的两个(CARA,IARA)
- R A ′ ( w 0 ) < 0 ⇔ a ∗ ′ ( w 0 ) > 0 R_A'(w_0)<0 \Leftrightarrow {a^{*}}'(w_0)>0 RA′(w0)<0⇔a∗′(w0)>0 (DARA)
- R A ′ ( w 0 ) = 0 ⇔ a ∗ ′ ( w 0 ) = 0 R_A'(w_0)=0 \Leftrightarrow {a^{*}}'(w_0)=0 RA′(w0)=0⇔a∗′(w0)=0 (CARA)
- R A ′ ( w 0 ) > 0 ⇔ a ∗ ′ ( w 0 ) < 0 R_A'(w_0)>0 \Leftrightarrow {a^{*}}'(w_0)<0 RA′(w0)>0⇔a∗′(w0)<0 (IARA)
在现实世界中,财富越多的人通常会在风险资产上投资更多,因此,从现实世界的观察来推断,人们大概应该都是绝对风险厌恶程度下降(DARA)的。这样的人,财富越多,投资在风险资产上的财富量就越大。
投资组合优化问题(II) – 定量分析
上面只是解决了不同风险厌恶水平下,财富与风险资产财富分配的定性分析,而我们想知道当初始财富 w 0 w_0 w0增加1%的时候, a ∗ a^* a∗会增加百分之多少,即定量问题
- 定义
a
∗
对
w
0
a^*对w_0
a∗对w0的弹性
e
(
w
0
)
e(w_0)
e(w0),(初始财富
w
0
w_0
w0增加1%的时候,
a
∗
a^*
a∗会增加百分之多少)
e ( w 0 ) ≜ d a ∗ a ∗ / d w 0 w 0 e(w_0) \triangleq \frac{da^*}{a^*} / \frac{dw_0}{w_0} e(w0)≜a∗da∗/w0dw0
先给出结论
-
R
R
′
(
w
0
)
<
0
⇔
e
(
w
0
)
>
1
R_R'(w_0)<0 \Leftrightarrow e(w_0)>1
RR′(w0)<0⇔e(w0)>1 (DRRA)
- 如果是DRRA,随着经济增长,财富增长,持有的风险资产比例随之增加
-
R
R
′
(
w
0
)
=
0
⇔
e
(
w
0
)
=
1
R_R'(w_0)=0 \Leftrightarrow e(w_0)=1
RR′(w0)=0⇔e(w0)=1 (CRRA)
- 如果是CRRA,随着经济增长,财富增长,持有的风险资产比例不变
-
R
R
′
(
w
0
)
>
0
⇔
e
(
w
0
)
<
1
R_R'(w_0)>0 \Leftrightarrow e(w_0)<1
RR′(w0)>0⇔e(w0)<1 (IARA)
- 如果是IRRA,随着经济增长,财富增长,持有的风险资产比例随之减少
e ( w 0 ) = w 0 a ∗ d a ∗ d w 0 = − w 0 ( 1 + r f ) E [ u ′ ′ ( w ~ ) ( r ~ − r f ) ] a ∗ E [ u ′ ′ ( w ~ ) ( r ~ − r f ) 2 ] e ( w 0 ) − 1 = − w 0 ( 1 + r f ) E [ u ′ ′ ( w ~ ) ( r ~ − r f ) ] + a ∗ E [ u ′ ′ ( w ~ ) ( r ~ − r f ) 2 ] a ∗ E [ u ′ ′ ( w ~ ) ( r ~ − r f ) 2 ] = − E [ u ′ ′ ( w ~ ) ( r ~ − r f ) ( w 0 ( 1 + r f ) + a ∗ ( r ~ − r f ) ) ] a ∗ E [ u ′ ′ ( w ~ ) ( r ~ − r f ) 2 ] = − 1 a ∗ E [ u ′ ′ ( w ~ ) ( r ~ − r f ) w ~ ] E [ u ′ ′ ( w ~ ) ( r ~ − r f ) 2 ] = − 1 a ∗ E [ − u ′ ( w ~ ) R R ( w ~ ) ( r ~ − r f ) ] E [ u ′ ′ ( w ~ ) ( r ~ − r f ) 2 ] ( R A ( y ) ≡ − u ′ ′ ( y ) u ′ ( y ) ) \begin{aligned} e(w_0) &= \frac{w_0}{a^*} \frac{da^*}{dw_0} = -\frac{w_0(1+r_f) E[u''(\tilde{w})(\tilde{r}-r_f)]}{a^*E[u''(\tilde{w})(\tilde{r}-r_f)^2]} \\ e(w_0) -1 &=-\frac{w_0(1+r_f) E[u''(\tilde{w})(\tilde{r}-r_f)]+a^*E[u''(\tilde{w})(\tilde{r}-r_f)^2]}{a^*E[u''(\tilde{w})(\tilde{r}-r_f)^2]} \\ &=-\frac{ E[u''(\tilde{w})(\tilde{r}-r_f)(w_0(1+r_f)+a^*(\tilde{r}-r_f))]}{a^*E[u''(\tilde{w})(\tilde{r}-r_f)^2]} \\ &=-\frac{1}{a^*}\frac{ E[u''(\tilde{w})(\tilde{r}-r_f)\tilde{w}]}{E[u''(\tilde{w})(\tilde{r}-r_f)^2]} \\ &=-\frac{1}{a^*}\frac{ E[-u'(\tilde{w})R_R(\tilde{w})(\tilde{r}-r_f)]}{E[u''(\tilde{w})(\tilde{r}-r_f)^2]} \qquad (R_A(y) \equiv -\frac{u''(y)}{u'(y)})\\ \end{aligned} e(w0)e(w0)−1=a∗w0dw0da∗=−a∗E[u′′(w~)(r~−rf)2]w0(1+rf)E[u′′(w~)(r~−rf)]=−a∗E[u′′(w~)(r~−rf)2]w0(1+rf)E[u′′(w~)(r~−rf)]+a∗E[u′′(w~)(r~−rf)2]=−a∗E[u′′(w~)(r~−rf)2]E[u′′(w~)(r~−rf)(w0(1+rf)+a∗(r~−rf))]=−a∗1E[u′′(w~)(r~−rf)2]E[u′′(w~)(r~−rf)w~]=−a∗1E[u′′(w~)(r~−rf)2]E[−u′(w~)RR(w~)(r~−rf)](RA(y)≡−u′(y)u′′(y))
接着 e ( w 0 ) − 1 e(w_0) -1 e(w0)−1的符号 = E [ − u ′ ( w ~ ) R R ( w ~ ) ( r ~ − r f ) ] E[-u'(\tilde{w})R_R(\tilde{w})(\tilde{r}-r_f)] E[−u′(w~)RR(w~)(r~−rf)]的符号,对 r n r_n rn分类讨论,构造一个不等式,分别对 R R ′ ( w 0 ) R_R'(w_0) RR′(w0)分类讨论,得到 e ( w 0 ) − 1 e(w_0) -1 e(w0)−1的符号,基本套路与上面的证明一致,这里省略…
在现实世界中,居民持有的财富在几百年来持续上升。如果人是 IRRA 或 DRRA 的偏好,那么我们应该观察到财富逐步被完全投资在风险资产上,或是完全投资到无风险资产上。但在现实中,投资在风险资产上的财富比例大致保持不变。因此,CRRA 是贴近现实偏好的效用函数形式。所以,在经济与金融分析中,CRRA 型效用函数应用得最为广泛。
风险与储蓄
确定性条件下的储蓄
符号 | 说明 | |
---|---|---|
初始财富 | w w w | |
储蓄 | s s s | |
时间价值 | δ \delta δ | 人心不耐的折扣 |
回报率 | R R R | 总回报率 R = 1 + r R=1+r R=1+r |
max
s
u
(
w
−
s
)
+
δ
u
(
s
R
)
F
O
C
:
u
′
(
w
−
s
)
=
δ
u
′
(
s
R
)
R
两边对
R
求导,
(
s
是
R
的函数,
s
(
R
)
)
−
d
s
d
R
u
′
′
(
w
−
s
)
=
δ
u
′
(
s
R
)
+
δ
R
[
u
′
′
(
s
R
)
(
s
+
R
d
s
d
R
)
]
d
s
d
R
=
δ
u
′
(
s
R
)
+
δ
s
R
u
′
′
(
s
R
)
−
u
′
′
(
w
−
s
)
−
δ
R
u
′
′
(
s
R
)
\begin{aligned} \max_s& u(w-s) + \delta u(sR) \\ FOC:& u'(w-s) = \delta u'(sR) R \\ 两边对R&求导,(s是R的函数,s(R)) \\ -\frac{ds}{dR}u''(w-s) &= \delta u'(sR) + \delta R[u''(sR)(s+R\frac{ds}{dR})] \\ \frac{ds}{dR} &=\frac{\delta u'(sR)+\delta sRu''(sR)}{-u''(w-s)-\delta Ru''(sR)} \end{aligned}
smaxFOC:两边对R−dRdsu′′(w−s)dRdsu(w−s)+δu(sR)u′(w−s)=δu′(sR)R求导,(s是R的函数,s(R))=δu′(sR)+δR[u′′(sR)(s+RdRds)]=−u′′(w−s)−δRu′′(sR)δu′(sR)+δsRu′′(sR)
对于分母,
u
′
′
(
⋅
)
<
0
u''(\cdot)<0
u′′(⋅)<0,所以分母大于0
d
s
d
R
的符号
=
δ
u
′
(
s
R
)
+
δ
s
R
u
′
′
(
s
R
)
的符号
\frac{ds}{dR}的符号 = \delta u'(sR)+\delta sRu''(sR)的符号
dRds的符号=δu′(sR)+δsRu′′(sR)的符号
δ
u
′
(
s
R
)
+
δ
s
R
u
′
′
(
s
R
)
=
δ
u
′
(
s
R
)
[
1
+
s
R
u
′
′
(
s
R
)
u
′
(
s
R
)
]
=
δ
u
′
(
s
R
)
[
1
−
R
R
(
s
R
)
]
\begin{aligned} \delta u'(sR)+\delta sRu''(sR) &= \delta u'(sR)[1+\frac{sRu''(sR)}{u'(sR)}] \\ &= \delta u'(sR)[1-R_R(sR)] \end{aligned}
δu′(sR)+δsRu′′(sR)=δu′(sR)[1+u′(sR)sRu′′(sR)]=δu′(sR)[1−RR(sR)]
- 如果 R R ( s R ) < 1 , d s d R > 0 R_R(sR) < 1, \frac{ds}{dR} > 0 RR(sR)<1,dRds>0,当R上升的时候,saving上升
- 如果 R R ( s R ) > 1 , d s d R < 0 R_R(sR) > 1, \frac{ds}{dR} < 0 RR(sR)>1,dRds<0,当R上升的时候,saving下降
回顾储蓄的两个效应
- substitution effect(替代效应): 今天储蓄在明天产生的财富更多,促使消费者多储蓄、少消费
- Income effect(收入效应): R越高,今明两天的总财富就越多,今天就应该多消费
而上面的结论则说明了这两个效应孰强孰弱是由RRA决定!
资源跨时间、跨状态的平滑配置
-
Intertemporary 跨期调配
max w 1 , w 2 = u ( w 1 ) + δ u ( w 2 ) s . t . w 1 + w 2 = w F O C : u ′ ( w 1 ) = δ u ′ ( w 2 ) \max_{w_1,w_2} = u(w_1) + \delta u(w_2) \\ s.t. \quad w_1 + w_2 = w \\ FOC: u'(w_1) = \delta u'(w_2) w1,w2max=u(w1)+δu(w2)s.t.w1+w2=wFOC:u′(w1)=δu′(w2)
一阶条件表示当期边际效用等于末期贴现的边际效用 -
跨状态调配
max w 1 , w 2 = P 1 u ( w 1 ) + P 2 u ( w 2 ) s . t . w 1 + w 2 = w F O C : P 1 u ′ ( w 1 ) = P 2 u ′ ( w 2 ) \max_{w_1,w_2} = P_1 u(w_1) + P_2 u(w_2) \\ s.t. \quad w_1 + w_2 = w \\ FOC: P_1u'(w_1) = P_2 u'(w_2) w1,w2max=P1u(w1)+P2u(w2)s.t.w1+w2=wFOC:P1u′(w1)=P2u′(w2)
一阶条件表示某情形下边际效用等于另一种情形下的边际效用
上面两个一阶条件表明
- 在考虑贴现后,消费者会在两个时点之间平滑财富配置
- 在考虑概率后,消费者会在两个状态之间平滑财富配置
最终平滑配置的结论:
- 风险厌恶度越大的人,跨期平滑配置的意愿更强
- 替代效应会促使人们做出非平滑配置(如上图所示,会使一个增加一个减少)
- 收入效应会促使人们做出平滑配置(如上图所示,两个都增加)
不确定情况下的储蓄
前面研究了当 E ( R ~ ) E(\tilde{R}) E(R~)变化的时候,saving的变化; 按照金融的逻辑,肯定还要研究当 R ~ \tilde{R} R~的风险变大的时候,saving的变化;
- 对
R
~
\tilde{R}
R~做一个mean-preserving spread(保均展形)
对于这个问题有两种效应
- 替代效应: 如果回报率的风险度上升,那么意味着储蓄的价值下降,这时还不如减少储蓄,增加当前确定的消费。
- 预防性储蓄:还有人可能会认为正因为未来不确定性上升,所以更应该多储蓄来为未来可能出现的不利局面做好准备。
下面给出证明:
max
s
u
(
w
−
s
)
+
δ
E
[
u
(
s
R
~
)
]
F
O
C
:
u
′
(
w
−
s
)
=
δ
E
[
R
~
u
′
(
s
R
~
)
]
将等式
左边记为
L
H
S
,右边记为
R
H
S
d
L
H
S
d
s
>
0
⇒
d
R
H
S
d
s
>
0
\begin{aligned} \max_s& u(w-s) + \delta E[u(s\tilde{R})] \\ FOC:& u'(w-s) = \delta E[\tilde{R}u'(s\tilde{R})] \\ 将等式&左边记为LHS ,右边记为RHS \\ \frac{dLHS}{ds} &> 0 \Rightarrow \frac{dRHS}{ds} > 0 \end{aligned}
smaxFOC:将等式dsdLHSu(w−s)+δE[u(sR~)]u′(w−s)=δE[R~u′(sR~)]左边记为LHS,右边记为RHS>0⇒dsdRHS>0
令
g
(
R
)
≜
R
u
′
(
s
R
)
g(R)\triangleq Ru'(sR)
g(R)≜Ru′(sR),当
g
(
R
)
g(R)
g(R)为凸函数时,
σ
2
(
R
)
↑
⇒
R
H
S
↑
⇒
d
R
H
S
d
s
>
0
\sigma^2(R) \uparrow \Rightarrow RHS \uparrow \Rightarrow \frac{dRHS}{ds} > 0
σ2(R)↑⇒RHS↑⇒dsdRHS>0
g
′
(
R
)
=
u
′
(
s
R
)
+
s
R
u
′
′
(
s
R
)
⇒
g
′
′
(
R
)
=
2
s
u
′
′
(
s
R
)
+
s
2
R
u
′
′
′
(
s
R
)
=
s
u
′
′
(
s
R
)
[
2
−
P
R
(
s
R
)
]
\begin{aligned} g'(R) &= u'(sR) + sRu''(sR) \\ \Rightarrow g''(R) &= 2su''(sR) + s^2R u'''(sR) \\ &=su''(sR)[2-P_R(sR)] \end{aligned}
g′(R)⇒g′′(R)=u′(sR)+sRu′′(sR)=2su′′(sR)+s2Ru′′′(sR)=su′′(sR)[2−PR(sR)]
P
R
(
y
)
≜
−
y
u
′
′
′
(
y
)
u
′
′
(
y
)
P_R(y) \triangleq -\frac{yu'''(y)}{u''(y)}
PR(y)≜−u′′(y)yu′′′(y),称为相对审慎系数; 类似地,还有绝对审慎系数,
P
A
(
y
)
≜
−
u
′
′
′
(
y
)
u
′
′
(
y
)
P_A(y) \triangleq -\frac{u'''(y)}{u''(y)}
PA(y)≜−u′′(y)u′′′(y)(根据之前讨论组合优化的问题一样,还可以考虑绝对审慎系数与风险扩大对saving量的变化,这里就不再做讨论了); 下面给出不确定下储蓄与风险变化的定性结论:
- 当 P R ( s R ) < 2 , g ′ ′ ( R ) < 0 P_R(sR) <2, g''(R)<0 PR(sR)<2,g′′(R)<0, g ( R ) g(R) g(R)为凹函数, σ 2 ( R ) ↑ , s a v i n g ↓ \sigma^2(R) \uparrow ,saving \downarrow σ2(R)↑,saving↓
- 当 P R ( s R ) > 2 , g ′ ′ ( R ) > 0 P_R(sR) >2, g''(R)>0 PR(sR)>2,g′′(R)>0, g ( R ) g(R) g(R)为凸函数, σ 2 ( R ) ↑ , s a v i n g ↑ \sigma^2(R) \uparrow ,saving \uparrow σ2(R)↑,saving↑(越是审慎的人,预防性动机超过了替代效应)