Risk Neutral Pricing
Radon-Nikodým Theorem
For equivalent measures P \mathbb{P} P and P ~ \tilde{\mathbb{P}} P~, we have P ~ ( A ) = ∫ A Z ( ω ) d P ( ω ) \tilde{\mathbb{P}}(A)=\int_\mathcal{A} Z(\omega) \, d\mathbb{P}(\omega) P~(A)=∫AZ(ω)dP(ω), where Z ( ω ) ≔ d P ~ d P Z(\omega)\coloneqq \frac{d\tilde{\mathbb{P}}}{d\mathbb{P}} Z(ω):=dPdP~ is the Radon-Nikodým derivative and:
E [ X ] ~ = E [ X Z ] ⟺ Z − 1 defined E [ X ] = E ~ [ X Z − 1 ] \tilde{\mathbb{E}[X]} = \mathbb{E}\left[ XZ \right]\qquad \overset{Z^{-1}\text{ defined}}{\iff} \qquad\mathbb{E}[X] = \tilde{\mathbb{E}}[XZ^{-1}] E[X]~=E[XZ]⟺Z−1 definedE[X]=E~[XZ−1]
Girsanov Theorem
Let W t W_t Wt be a B.M. under P \mathbb{P} P, and Θ ( t ) \Theta(t) Θ(t) an adapted process of filtration F ( t ) \mathcal{F}(t) F(t), and assume E ∫ 0 T Θ 2 ( t ) Z 2 ( u ) d u < ∞ \mathbb{E}\int_0^T \Theta^2(t)Z^2(u) \, du < \infty E∫0TΘ2(t)Z2(u)du<∞, then:
Z ( t ) = exp ( − ∫ 0 t Θ ( u ) d W u − 1 2 ∫ 0 t Θ 2 ( u ) d u ) W ~ t = W t + ∫ 0 t Θ ( u ) d u ⟺ d W ~ t = d W t + Θ ( t ) d t \begin{aligned} Z(t) &= \exp\left( -\int_0^t \Theta(u) \, dW_u - \frac{1}{2}\int_0^t \Theta^2(u) \, du \right) \\ \\ \tilde{W}_t &= W_t + \int_0^t \Theta(u) \, du \qquad\iff\qquad d\tilde{W}_t = dW_t + \Theta(t)dt \end{aligned} Z(t)W~t=exp(−∫0tΘ(u)dWu−21∫0tΘ2(u)du)=Wt+∫0tΘ(u)du⟺dW~t=dWt+Θ(t)dt
Then W ~ t \tilde{W}_t W~t is a B.M. under P ~ \tilde{\mathbb{P}} P~.
(First) Fundamental Theorem of Asset Pricing
A market is arbitrage-free if and only if there exist an equivalent matringale measure Q \mathbb{Q} Q. Specifically, we want the discounted stock process D ( t ) S ( t ) = D t S t D(t)S(t)=D_tS_t D(t)S(t)=DtSt to be a martingale.
Stock Process
d
S
t
=
μ
t
S
t
d
t
+
σ
t
S
t
d
W
t
⟹
S
t
=
S
0
exp
(
∫
0
t
(
μ
s
−
1
2
σ
s
2
)
d
s
+
∫
0
t
σ
s
d
W
s
)
Discount Process
d
D
t
=
−
R
t
D
t
d
t
⟹
D
t
=
e
−
∫
0
t
R
s
d
s
Discounted Stock Process
D
t
S
t
=
S
0
exp
(
∫
0
t
(
μ
s
−
R
s
−
1
2
σ
s
2
)
d
s
+
∫
0
t
σ
s
d
W
s
)
⟹
d
(
D
t
S
t
)
=
σ
t
D
t
S
t
(
μ
t
−
R
t
σ
t
≕
Θ
(
t
)
d
t
+
d
W
t
)
\begin{aligned} &\text{Stock Process}& dS_t &= \mu_tS_tdt + \sigma_tS_tdW_t \\ \\ && \implies S_t &= S_0\exp\left( \int_0^t \left( \mu_s-\frac{1}{2}\sigma^2_s \right) \, ds + \int_0^t \sigma_s \, dW_s \right) \\ \\ \\ &\text{Discount Process}& dD_t &= -R_tD_tdt \\ \\ && \implies D_t &= e^{-\int_0^t R_s \, ds} \\ \\ \\ &\text{Discounted Stock Process}& D_tS_t &= S_0\exp\left( \int_0^t \left( \mu_s- R_s- \frac{1}{2}\sigma^2_s \right) \, ds + \int_0^t \sigma_s \, dW_s \right) \\ \\ && \implies d\left( D_tS_t \right) &= \sigma_tD_tS_t\left( \underset{\eqqcolon \Theta(t)}{\frac{\mu_t-R_t}{\sigma_t}}dt +dW_t\right) \\ \\ \end{aligned}
Stock ProcessDiscount ProcessDiscounted Stock ProcessdSt⟹StdDt⟹DtDtSt⟹d(DtSt)=μtStdt+σtStdWt=S0exp(∫0t(μs−21σs2)ds+∫0tσsdWs)=−RtDtdt=e−∫0tRsds=S0exp(∫0t(μs−Rs−21σs2)ds+∫0tσsdWs)=σtDtSt
=:Θ(t)σtμt−Rtdt+dWt
Where
R
t
R_t
Rt is the adapted I.R. process.
Recall from Fundamental Theorem of Asset Pricing, we want the discounted stock price to be a martingale. This can be achieved by making ( μ t − R t σ t d t + d W t ) \left( {\frac{\mu_t-R_t}{\sigma_t}dt} +dW_t\right) (σtμt−Rtdt+dWt) term a B.M. using Girsanov Theorem, such that:
d W ~ t = ! μ t − R t σ t d t + d W t ⟹ d S t = μ t S t d t + σ t S t d W t = μ t S t d t + σ t S t ( d W ~ t − μ t − R t σ t d t ) = R t S t d t + σ t S t d W ~ t □ \begin{aligned} d\tilde{W}_t \overset{!}= \frac{\mu_t-R_t}{\sigma_t}dt +dW_t\qquad \implies \qquad dS_t &= \mu_tS_tdt + \sigma_t S_t dW_t \\ \\ &= \mu_tS_tdt + \sigma_tS_t\left( d\tilde{W}_t-\frac{\mu_t-R_t}{\sigma_t}dt \right) \\ \\ &= R_tS_tdt + \sigma_tS_td\tilde{W}_t&\square \end{aligned} dW~t=!σtμt−Rtdt+dWt⟹dSt=μtStdt+σtStdWt=μtStdt+σtSt(dW~t−σtμt−Rtdt)=RtStdt+σtStdW~t□
Black-Scholes Formula for Call Option
Stock Price under Q d S t = r S t d t + σ S t d W t Value of Call Option C t = E [ e − r τ ( S T − K ) + ∣ S t ] \begin{aligned} &\text{Stock Price under $\mathbb{Q}$}& dS_t &= rS_tdt + \sigma S_t dW_t \\ \\ &\text{Value of Call Option}& C_t &= \mathbb{E}\left[ e^{-r\tau} \left( S_T-K \right)^+ |S_t \right] \end{aligned} Stock Price under QValue of Call OptiondStCt=rStdt+σStdWt=E[e−rτ(ST−K)+∣St]
By Direct Integration
C t = E [ e − r τ ( S T − K ) + ∣ S t ] = e − r τ E [ ( S t e ( r − 1 2 σ 2 ) τ + σ τ ϵ − K ) + ∣ S t ] = e − r τ [ ∫ R ( S t e ( r − 1 2 σ 2 ) τ + σ τ ϵ − K ) + f ϵ ( ϵ ) d ϵ ] = e − r τ [ ∫ R ( S t e ( r − 1 2 σ 2 ) τ + σ τ ϵ − K ) + 1 2 π e − 1 2 ϵ 2 d ϵ ] = e − r τ [ ∫ − d 2 ∞ S t e ( r − 1 2 σ 2 ) τ + σ τ ϵ 1 2 π e − 1 2 ϵ 2 d ϵ − ∫ − d 2 ∞ K 1 2 π e − 1 2 ϵ 2 d ϵ ] = e − r τ [ ∫ − d 2 ∞ 1 2 π S t e ( r − 1 2 σ 2 ) τ + σ τ ϵ − 1 2 ϵ 2 d ϵ − K ( 1 − N ( − d 2 ) ) ] = e − r τ [ e r τ S t ∫ − d 2 ∞ 1 2 π e 1 2 σ 2 τ + σ τ ϵ − 1 2 ϵ 2 d ϵ − K N ( d 2 ) ] = e − r τ [ e r τ S t ∫ − d 2 ∞ 1 2 π e − 1 2 ( ϵ − σ τ ) 2 d ϵ − K N ( d 2 ) ] = e − r τ [ e r τ S t ∫ − d 2 − σ τ ∞ 1 2 π e − 1 2 u 2 d u − K N ( d 2 ) ] u ≔ ϵ − σ τ = e − r τ [ e r τ S t N ( d 2 + σ τ ) − K N ( d 2 ) ] = S t N ( d 1 ) − K e − r τ N ( d 2 ) □ \begin{aligned} C_t &= \mathbb{E}\left[ e^{-r\tau} \left( S_T-K \right)^+ |S_t \right] \\ \\ &= e^{-r\tau} \mathbb{E}\left[ \left( S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma\sqrt{\tau}\epsilon} - K \right)^+ | S_t \right] \\ \\ &= e^{-r\tau} \left[ \int_\mathbb{R} \left( S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma\sqrt{\tau}\epsilon} - K \right)^+ f_\epsilon(\epsilon) \, d\epsilon \right] \\ \\ &= e^{-r\tau} \left[ \int_\mathbb{R} \left( S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma\sqrt{\tau}\epsilon} - K \right)^+ \frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}\epsilon^2} \, d\epsilon \right] \\ \\ &= e^{-r\tau} \left[ \int_{-d_2}^\infty S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma\sqrt{\tau}\epsilon} \frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}\epsilon^2} \, d\epsilon \;-\; \int_{-d_2}^\infty K \frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}\epsilon^2} \, d\epsilon \right] \\ \\ &= e^{-r\tau} \left[ \int_{-d_2}^\infty \frac{1}{\sqrt{2\pi}}S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma\sqrt{\tau}\epsilon - \frac{1}{2}\epsilon^2} \, d\epsilon \;-\; K\left( 1-N(-d_2) \right) \right] \\ \\ &= e^{-r\tau} \left[ e^{r\tau}S_t \int_{-d_2}^\infty \frac{1}{\sqrt{2\pi}}e^{\frac{1}{2}\sigma^2 \tau + \sigma\sqrt{\tau}\epsilon - \frac{1}{2}\epsilon^2} \, d\epsilon \;-\; KN(d_2) \right] \\ \\ &= e^{-r\tau} \left[ e^{r\tau}S_t \int_{-d_2}^\infty \frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}\left( \epsilon-\sigma\sqrt{\tau} \right)^2} \, d\epsilon \;-\; KN(d_2) \right] \\ \\ &= e^{-r\tau} \left[ e^{r\tau}S_t \int_{-d_2-\sigma\sqrt{\tau}}^\infty \frac{1}{\sqrt{2\pi}}e^{-\frac{1}{2}u^2} \, du \;-\; KN(d_2) \right] &u\coloneqq \epsilon-\sigma\sqrt{\tau} \\ \\ &= e^{-r\tau} \left[ e^{r\tau}S_t N(d_2+\sigma\sqrt{\tau}) - KN(d_2) \right] \\ \\ &= S_tN(d_1) - Ke^{-r\tau}N(d_2) &\square \end{aligned} Ct=E[e−rτ(ST−K)+∣St]=e−rτE[(Ste(r−21σ2)τ+στϵ−K)+∣St]=e−rτ[∫R(Ste(r−21σ2)τ+στϵ−K)+fϵ(ϵ)dϵ]=e−rτ[∫R(Ste(r−21σ2)τ+στϵ−K)+2π1e−21ϵ2dϵ]=e−rτ[∫−d2∞Ste(r−21σ2)τ+στϵ2π1e−21ϵ2dϵ−∫−d2∞K2π1e−21ϵ2dϵ]=e−rτ[∫−d2∞2π1Ste(r−21σ2)τ+στϵ−21ϵ2dϵ−K(1−N(−d2))]=e−rτ[erτSt∫−d2∞2π1e21σ2τ+στϵ−21ϵ2dϵ−KN(d2)]=e−rτ[erτSt∫−d2∞2π1e−21(ϵ−στ)2dϵ−KN(d2)]=e−rτ[erτSt∫−d2−στ∞2π1e−21u2du−KN(d2)]=e−rτ[erτStN(d2+στ)−KN(d2)]=StN(d1)−Ke−rτN(d2)u:=ϵ−στ□
Where
−
d
2
-d_2
−d2 is the lower boundary of the integral to make
S
T
>
K
S_T>K
ST>K:
S
T
>
K
S
t
e
(
r
−
1
2
σ
2
)
τ
+
σ
τ
ϵ
>
K
log
S
t
+
(
r
−
1
2
σ
2
)
τ
+
σ
τ
ϵ
>
log
K
ϵ
>
log
K
S
−
(
r
−
1
2
σ
2
)
τ
σ
τ
≕
−
d
2
⟹
d
2
=
log
S
t
K
+
(
r
−
1
2
σ
2
)
τ
σ
τ
d
1
≔
d
2
+
σ
τ
=
log
S
t
K
+
(
r
+
1
2
σ
2
)
τ
σ
τ
\begin{aligned} S_T &> K \\ \\ S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma\sqrt{\tau}\epsilon} &> K \\ \\ \log S_t + \left( r-\frac{1}{2}\sigma^2 \right)\tau + \sigma\sqrt{\tau}\epsilon &> \log K \\ \\ \epsilon & > \frac{\log \frac{K}{S} - \left( r- \frac{1}{2}\sigma^2 \right)\tau}{\sigma\sqrt{\tau}} \eqqcolon -d_2 \\ \\ \\ \implies\qquad d_2=\frac{\log \frac{S_t}{K} + \left( r- \frac{1}{2}\sigma^2 \right)\tau}{\sigma\sqrt{\tau}}\qquad &d_1\coloneqq d_2+\sigma\sqrt{\tau}=\frac{\log \frac{S_t}{K} + \left( r+ \frac{1}{2}\sigma^2 \right)\tau}{\sigma\sqrt{\tau}} \end{aligned}
STSte(r−21σ2)τ+στϵlogSt+(r−21σ2)τ+στϵϵ⟹d2=στlogKSt+(r−21σ2)τ>K>K>logK>στlogSK−(r−21σ2)τ=:−d2d1:=d2+στ=στlogKSt+(r+21σ2)τ
By Girsanov Theorem
C
t
=
E
[
e
−
r
τ
(
S
T
−
K
)
+
∣
S
t
]
=
e
−
r
τ
E
[
1
S
T
>
K
S
T
]
≕
A
−
e
−
r
τ
E
[
1
S
T
>
K
K
]
≕
B
\begin{aligned} C_t &= \mathbb{E}\left[ e^{-r\tau} \left( S_T-K \right)^+ |S_t \right] \\ \\ &= \underset{\eqqcolon A}{e^{-r\tau}\mathbb{E}\left[ \mathbb{1}_{S_T>K}S_T \right]} - \underset{\eqqcolon B}{e^{-r\tau}\mathbb{E}\left[ \mathbb{1}_{S_T>K}K \right]} \end{aligned}
Ct=E[e−rτ(ST−K)+∣St]==:Ae−rτE[1ST>KST]−=:Be−rτE[1ST>KK]
Notice we remove the conditional expectation notation only for convenience.
For part
A
A
A, we may identify an exponential term as a Ridan-Nikodým derivative and simplify the expectation to a probability.
A
=
e
−
r
τ
E
[
1
S
T
>
K
S
T
]
=
e
−
r
τ
E
[
1
S
T
>
K
S
t
e
(
r
−
1
2
σ
2
)
τ
+
σ
W
τ
]
=
e
−
r
τ
e
r
τ
S
t
E
[
1
S
T
>
K
e
1
2
σ
2
τ
+
σ
W
τ
]
=
S
t
E
[
1
S
T
>
K
e
1
2
σ
2
τ
+
σ
W
τ
]
notice
e
1
2
σ
2
τ
+
σ
W
τ
is a Ridan-Nikodym derivative with
Θ
(
t
)
=
−
σ
=
S
t
E
[
1
S
T
>
K
Z
t
]
=
S
t
E
~
[
1
S
T
>
K
]
=
S
t
P
~
(
S
T
>
K
)
=
S
t
P
~
(
S
t
e
(
r
−
1
2
σ
2
)
τ
+
σ
W
τ
>
K
)
=
S
t
P
~
(
S
t
e
(
r
−
1
2
σ
2
)
τ
+
σ
(
W
~
t
+
σ
τ
)
>
K
)
W
t
=
W
~
t
−
Θ
(
t
)
,
Θ
(
t
)
=
−
σ
=
S
t
P
~
(
S
t
e
(
r
+
1
2
σ
2
)
τ
+
σ
W
~
τ
>
K
)
=
S
t
P
~
(
ϵ
^
>
log
K
S
t
−
(
r
+
1
2
σ
2
)
τ
σ
τ
)
=
S
t
N
(
d
1
)
\begin{aligned} A & = e^{-r\tau}\mathbb{E}\left[ \mathbb{1}_{S_T>K}S_T \right] \\ \\ &= e^{-r\tau}\mathbb{E}\left[ \mathbb{1}_{S_T>K}S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma W_\tau} \right] \\ \\ &= e^{-r\tau}e^{r\tau}S_t\mathbb{E}\left[ \mathbb{1}_{S_T>K}e^{\frac{1}{2}\sigma^2\tau + \sigma W_\tau} \right] \\ \\ &= S_t\mathbb{E}\left[ \mathbb{1}_{S_T>K}e^{\frac{1}{2}\sigma^2\tau + \sigma W_\tau} \right] &\text{notice $e^{\frac{1}{2}\sigma^2\tau + \sigma W_\tau}$ is a Ridan-Nikodym derivative with $\Theta(t)=-\sigma$} \\ \\ &= S_t\mathbb{E}\left[ \mathbb{1}_{S_T>K}Z_t \right] \\ \\ &= S_t\tilde{\mathbb{E}}\left[ \mathbb{1}_{S_T>K} \right] \\ \\ &= S_t\tilde{\mathbb{P}}\left( S_T>K \right) \\ \\ &= S_t\tilde{\mathbb{P}}\left( S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma W_\tau}>K \right) \\ \\ &= S_t\tilde{\mathbb{P}}\left( S_te^{\left( r- \frac{1}{2}\sigma^2 \right)\tau + \sigma \left( \tilde{W}_t+\sigma \tau \right)}>K \right) &W_t=\tilde{W}_t-\Theta(t),\;\Theta(t)=-\sigma \\ \\ &= S_t\tilde{\mathbb{P}}\left( S_te^{\left( r + \frac{1}{2}\sigma^2 \right)\tau + \sigma\tilde{W}_\tau}>K \right) \\ \\ &= S_t\tilde{\mathbb{P}}\left( \hat\epsilon > \frac{\log \frac{K}{S_t}-\left( r+\frac{1}{2}\sigma^2 \right)\tau}{\sigma\sqrt{\tau}} \right) \\ \\ &= S_t N(d_1) \end{aligned}
A=e−rτE[1ST>KST]=e−rτE[1ST>KSte(r−21σ2)τ+σWτ]=e−rτerτStE[1ST>Ke21σ2τ+σWτ]=StE[1ST>Ke21σ2τ+σWτ]=StE[1ST>KZt]=StE~[1ST>K]=StP~(ST>K)=StP~(Ste(r−21σ2)τ+σWτ>K)=StP~(Ste(r−21σ2)τ+σ(W~t+στ)>K)=StP~(Ste(r+21σ2)τ+σW~τ>K)=StP~(ϵ^>στlogStK−(r+21σ2)τ)=StN(d1)notice e21σ2τ+σWτ is a Ridan-Nikodym derivative with Θ(t)=−σWt=W~t−Θ(t),Θ(t)=−σ
The other part is easy:
B = e − r τ E [ 1 S T > K K ] = e − r τ K E [ 1 S T > K ] = e − r τ K P ( S T > K ) = e − r τ K N ( d 2 ) \begin{aligned} B &= e^{-r\tau}\mathbb{E}\left[ \mathbb{1}_{S_T>K}K \right] \\ \\ &= e^{-r\tau}K\mathbb{E}\left[ \mathbb{1}_{S_T>K} \right] \\ \\ &= e^{-r\tau}K\mathbb{P}\left( S_T>K \right) \\ \\ &= e^{-r\tau}KN(d_2) \end{aligned} B=e−rτE[1ST>KK]=e−rτKE[1ST>K]=e−rτKP(ST>K)=e−rτKN(d2)
So we have the result:
C t = e − r τ E [ 1 S T > K S T ] ≕ A − e − r τ E [ 1 S T > K K ] ≕ B = S t N ( d 1 ) − K e − r τ N ( d 2 ) □ \begin{aligned} C_t &= \underset{\eqqcolon A}{e^{-r\tau}\mathbb{E}\left[ \mathbb{1}_{S_T>K}S_T \right]} - \underset{\eqqcolon B}{e^{-r\tau}\mathbb{E}\left[ \mathbb{1}_{S_T>K}K \right]} \\ \\ &= S_tN(d_1) - Ke^{-r\tau} N(d_2)&\square \end{aligned} Ct==:Ae−rτE[1ST>KST]−=:Be−rτE[1ST>KK]=StN(d1)−Ke−rτN(d2)□
自己写的。之后有时间会翻译成中文，如有错误欢迎指出。
扫一扫
2017
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
