Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal...

40
1 / 24 Dynamic Protection for Bayesian Optimal Portfolio Hideaki Miyata Department of Mathematics, Kyoto University Jun Sekine Institute of Economic Research, Kyoto University Jan. 6, 2009, Kunitachi, Tokyo

Transcript of Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal...

Page 1: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

1 / 24

Dynamic Protection for Bayesian Optimal Portfolio

Hideaki MiyataDepartment of Mathematics, Kyoto University

Jun SekineInstitute of Economic Research, Kyoto University

Jan. 6, 2009, Kunitachi, Tokyo

Page 2: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Problem

⊲ Problem

⊲ Outline

Applications

Results

2 / 24

Optimal stopping problem,

sup0≤τ≤T

ENτXτ ,

with the running maximum,

Nt := 1 ∨ maxs∈[0,t]

Ks

Xs

, Xt := f(t, wt)e−δt,

and 1-dim BM w, where f ∈ C1,2([0, T ] × R) that satisfies

(

∂t +1

2∂xx

)

f = 0, f, ∂xf > 0,

δ ∈ R≥0, and K ∈ C1([0, T ], R>0) are given.

Page 3: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Problem

⊲ Problem

⊲ Outline

Applications

Results

2 / 24

Optimal stopping problem,

sup0≤τ≤T

ENτXτ ,

with the running maximum,

Nt := 1 ∨ maxs∈[0,t]

Ks

Xs

, Xt := f(t, wt)e−δt,

and 1-dim BM w, where f ∈ C1,2([0, T ] × R) that satisfies

(

∂t +1

2∂xx

)

f = 0, f, ∂xf > 0,

δ ∈ R≥0, and K ∈ C1([0, T ], R>0) are given.(f(t, x) := x0 exp ax − (a2t)/2: BS-case.)

Page 4: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Outline

⊲ Problem

⊲ Outline

Applications

Results

3 / 24

(Financial) Applications

Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model

(Mathematical) Results/Contributions

Extension of Peskir (2005)’s analysis for BS-model.

Page 5: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Outline

⊲ Problem

⊲ Outline

Applications

Results

3 / 24

(Financial) Applications

Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model

(Mathematical) Results/Contributions

Extension of Peskir (2005)’s analysis for BS-model.

Characterize with a free-boundary problem for 3-dim.Markov (t, X, N).

Page 6: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Outline

⊲ Problem

⊲ Outline

Applications

Results

3 / 24

(Financial) Applications

Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model

(Mathematical) Results/Contributions

Extension of Peskir (2005)’s analysis for BS-model.

Characterize with a free-boundary problem for 3-dim.Markov (t, X, N).(Reduction to 2-dim. Markov(t, NX) is known in BS-case.)

Page 7: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Outline

⊲ Problem

⊲ Outline

Applications

Results

3 / 24

(Financial) Applications

Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model

(Mathematical) Results/Contributions

Extension of Peskir (2005)’s analysis for BS-model.

Characterize with a free-boundary problem for 3-dim.Markov (t, X, N).(Reduction to 2-dim. Markov(t, NX) is known in BS-case.)

Free-boundary is a unique sol. of a nonlinearintegral equation.

Page 8: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Outline

⊲ Problem

⊲ Outline

Applications

Results

3 / 24

(Financial) Applications

Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model

(Mathematical) Results/Contributions

Extension of Peskir (2005)’s analysis for BS-model.

Characterize with a free-boundary problem for 3-dim.Markov (t, X, N).(Reduction to 2-dim. Markov(t, NX) is known in BS-case.)

Free-boundary is a unique sol. of a nonlinearintegral equation.

A numerical computation scheme.

Page 9: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Financial Applications

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

4 / 24

Page 10: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Application (1): Dynamic Fund Protection

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

5 / 24

In a complete market with the risk-neutral P,

Page 11: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Application (1): Dynamic Fund Protection

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

5 / 24

In a complete market with the risk-neutral P,

X: one unit of (the discounted) investment fund,δ: dividend rate, paid to customers (not re-invested in thefund).

Page 12: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Application (1): Dynamic Fund Protection

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

5 / 24

In a complete market with the risk-neutral P,

X: one unit of (the discounted) investment fund,δ: dividend rate, paid to customers (not re-invested in thefund).

Nt: minimal aggregate number nt of the fund units in acustomer’s account s.t.

(i) n0 = 1,

(ii) nt ≥ ns for t ≥ s ≥ 0, and

(iii) ntXt ≥ Kt (: floor) for ∀t ≥ 0.

Page 13: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Application (1): Dynamic Fund Protection

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

5 / 24

In a complete market with the risk-neutral P,

X: one unit of (the discounted) investment fund,δ: dividend rate, paid to customers (not re-invested in thefund).

Nt: minimal aggregate number nt of the fund units in acustomer’s account s.t.

(i) n0 = 1,

(ii) nt ≥ ns for t ≥ s ≥ 0, and

(iii) ntXt ≥ Kt (: floor) for ∀t ≥ 0.

The protection scheme is called the dynamic fund protection

(Gerber-Pafumi, 2000 and Gerber-Shiu, 2003).

Page 14: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Dynamic Protetion vs. European-OBPI

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

6 / 24

A simple European-OBPI (i.e., buy European put written onX with strike k) controls a down-side-risk of X at theterminal T :

XT + (k − XT )+ = XT ∨ k ≥ k.

DFP controls a down-side-risk of the process X (or NX)

NtXt ≥ Kt for ∀t ∈ [0, T ].

Page 15: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

“Attractive” Feature of Dynamic Fund Protetion

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

7 / 24

DFP may be more “attractive” than European-OBPI forcustomers...

Suppose XT < KT (= k). Then,

E-OBPI: XT + (KT − XT )+ = KT , : floor value,

DFP: NT XT =

(

1 + max0≤t≤T

Kt

Xt

)

XT

≥(

1 +KT

XT

)

XT = XT + KT .

Page 16: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

“Attractive” Feature of Dynamic Fund Protetion

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

7 / 24

DFP may be more “attractive” than European-OBPI forcustomers...

Suppose XT < KT (= k). Then,

E-OBPI: XT + (KT − XT )+ = KT , : floor value,

DFP: NT XT =

(

1 + max0≤t≤T

Kt

Xt

)

XT

≥(

1 +KT

XT

)

XT = XT + KT .

This comparison is not “fair”, of course..

Page 17: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Pricing Dynamic Fund Protection

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

8 / 24

DFP-scheme is not self-financing. To compute the fair value of “DFP-option”, compute the

Snell envelope,

Vt := ess supt≤τ≤T

E [NτXτ |Ft] , (t ∈ [0, T ]),

i.e., the minimal superreplicating self-financing portfolio ofNX.

Apply a standard no-arbitrage pricing argument in completemarket.

Page 18: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Extension: Beyond BS

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

9 / 24

BS: Xt := x0 exp awt − (a2/2 + δ)t (δ = 0 byGerber-Pafumi; δ > 0 and T = ∞ by Gerber–Shiu)

Page 19: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Extension: Beyond BS

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

9 / 24

BS: Xt := x0 exp awt − (a2/2 + δ)t (δ = 0 byGerber-Pafumi; δ > 0 and T = ∞ by Gerber–Shiu)

CEV (with δ = 0) by Imai-Boyle.

Page 20: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Extension: Beyond BS

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

9 / 24

BS: Xt := x0 exp awt − (a2/2 + δ)t (δ = 0 byGerber-Pafumi; δ > 0 and T = ∞ by Gerber–Shiu)

CEV (with δ = 0) by Imai-Boyle. Dynamically invested fund,

X :=X(x0,π,δ), where

dXt

Xt

=πt

dSt

St

+ (1 − πt)rdt − δdt, X0 = x0.

Page 21: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Extension: Beyond BS

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

9 / 24

BS: Xt := x0 exp awt − (a2/2 + δ)t (δ = 0 byGerber-Pafumi; δ > 0 and T = ∞ by Gerber–Shiu)

CEV (with δ = 0) by Imai-Boyle. Dynamically invested fund,

X :=X(x0,π,δ), where

dXt

Xt

=πt

dSt

St

+ (1 − πt)rdt − δdt, X0 = x0.

πt := σ−2(µ − r): growth-optimal fund,

Page 22: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Extension: Beyond BS

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

9 / 24

BS: Xt := x0 exp awt − (a2/2 + δ)t (δ = 0 byGerber-Pafumi; δ > 0 and T = ∞ by Gerber–Shiu)

CEV (with δ = 0) by Imai-Boyle. Dynamically invested fund,

X :=X(x0,π,δ), where

dXt

Xt

=πt

dSt

St

+ (1 − πt)rdt − δdt, X0 = x0.

πt := σ−2(µ − r): growth-optimal fund, πt := σ−2 (E[µ|Ft] − r): Bayesian growth optimal fund.

Page 23: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Application (2): DP for Bayesian Optimal Portfolio

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

10 / 24

Financial market with a bank-account B ≡ 1 and astock(-index)

dSt = Stσ(t, S)(dzt + λdt), S0 > 0

on (Ω,F , P0, (Gt)t∈[0,T ]), where

Gt := σ(zu; u ∈ [0, t]) ∨ σ(λ),

1-dim. BM z and r.v. λ ∼ ν, independent of z. Another filtration,

St := σ(Su; u ∈ [0, t]),

is regarded as the available information for investors.

Page 24: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

DP for Dynamically Invested Fund

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

11 / 24

Treat DP ofX := X(x0,π,δ),

wheredXt

Xt

= πt

dSt

St

− δdt, X0 = x0,

where π := (πt)t∈[0,T ] belongs to

AT :=

(ft)t∈[0,T ] : St-prog. m’ble,

∫ T

0

|ftσt|2dt < ∞ a.s.

.

In particular, we are interested in the optimally invested fund,X, so that

supπ∈AT

EU(

Xx0,π,δT

)

= EU(

XT

)

.

Page 25: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Reference Measure (EMM under Partial Information)

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

12 / 24

IntroducedP

dP0

Gt

= exp

(

−λzt −1

2λ2t

)

to seewt := zt + λt

is a (P,Gt)-BM and

St := σ(Su; u ≤ t) = σ(wu; u ≤ t) =: Ft for ∀t ∈ [0, T ]

sincedSt = Stσ(t, S)dwt : unique strong sol.

and

dwt =dSt

Stσ(t, S).

Page 26: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Bayesian Optimal Portfolio (Karatzas-Zhao, 2000)

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

13 / 24

Xt = X (t, wt) (t ∈ [0, T ]).

Here, defineX (t, y) := X (t,Y(x0), y),

where

F (t, y) :=

R≥0

exp

(

xy − 1

2|x|2t

)

ν(dx),

X (t, x, y) :=eδ(T−t)

R

I(

eδT x

F (T, y +√

T − tz)

)

1√2π

e−|z|2

2 dz,

I := (U ′)−1, and Y(·) := X (0, ·, 0)−1.

Page 27: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Russian Option for Local-volatility Model

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

14 / 24

Optimal stopping problem is rewritten as

f(0, 0) × sup0≤τ≤T

E

[

e−δτ

1 ∨ maxs∈[0,τ ]

Ys

]

.

Here, P on (Ω,FT ) is defined by

dP

dP

Ft

:=f(t, wt)

f(0, 0),

Yt := Kt/Xt satisfies the Markovian SDE,

dYt = Yt [a(t, Yt)dwt + (κt + δ) dt]

with a(t, y) := ∂x log f (t, f−1 (t, Kt/y)), κt := ∂t log Kt, and

the P-BM, wt := −

wt −∫ t

0∂x log f(u, wu)du

.

Page 28: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Related Works on Russian Options

⊲ Problem

⊲ Outline

Applications

⊲ DFP

⊲ vs. E-OBPI

⊲ Feature

⊲ Pricing

⊲ Extension

⊲ DP-BOP(1)

⊲ DP-BOP(2)

⊲ EMM

⊲ BOP

⊲ Russian Option

⊲ Related Woks

Results

15 / 24

BS-case, i.e., Xt := x0 exp awt − (δ + a2/2)t, Kt := K0eαt,

anddYt = Yt adwt + (α + δ)dt .

T = ∞: explicit results by Shepp-Shiryaev, Duffie-Harrison,Salminen, etc.

T < ∞: studied by Duistermaat-Kyprianou-van Schaik,Ekstrom, Peskir, etc.

Some extenstions: Guo-Shepp, Pedersen, Gapeev, etc.

General treatment with viscocity of variational inequality:Barles-Daher-Romano.

Page 29: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Mathematical Results

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

16 / 24

Page 30: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Dynamic Version of Optimal Stopping

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

17 / 24

By dynamic-programming and the Markov property of (t, X, N),or (t, Y, N) := (t, K/X, N), deduce

ess supt≤τ≤T

E [NτXτ |Ft] = V (t, Xt, Nt) = e−δtXtV (t, Kt/Xt, Nt) ,

where

V (t, y, z) := sup0≤τ≤T−t

Ee−δτN (t,y,z)τ ,

N (t,y,z)s :=z ∨ max

u∈[0,s]Y (t,y)

u ,

and Y (t,y) solves

dYs = Ys a(t + s, Ys)dws + (κt+s + δ)ds , Y0 = y.

Page 31: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Free Boundary Problem

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

18 / 24

(∂t + L − δ)V (t, y, z) =0 for (t, y, z) ∈ C,

V (t, y, z)|y=b(t,z)+ =z, (instantaneous stopping),

∂yV (t, y, z)|y=b(t,z)+ =0, (smooth-pasting),

∂zV (t, y, z)|z=y+ =0, (normal-reflection),

V (t, y, z) >z for (t, y, z) ∈ C,

V (t, y, z) =z for (t, y, z) ∈ D,

where L := 12y2a(t, y)2∂yy + y(κt + δ)∂y,

C := (t, y, z) ∈ [0, T ) × E; b(t, z) < y ≤ z ,

D := (t, y, z) ∈ [0, T ) × E; 0 < y ≤ b(t, z) ,

E :=

(y, z) ∈ R2; 0 < y ≤ z, z ≥ 1

.

Page 32: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Main Theorem (1)

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

19 / 24

(1) For V , ∃b : [0, T ) × [1,∞) → R>0, s.t. (V, b) solves theFree-boundary Prob.

(2) b is the unique solution to

z = e−δ(T−t)E

[

N(t,b(t,z),z)T−t

]

+ δ

∫ T−t

0

dse−δs

× E[

N (t,b(t,z),z)s I

(

Y (t,b(t,z))s < b

(

t + s, N (t,b(t,z),z)s

))]

so that

(i) b: continuous, nondecreasing w.r.t. t,(ii) b(t, z) < z and b(T−, z) = z.

Page 33: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Main Theorem (2)

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

20 / 24

(3) Optimal stopping time:

τt := inf 0 ≤ t ≤ T ; Yt ≤ b(t, Nt) .

(4) “Closed-form” expression:

V (t, y, z) = e−δ(T−t)E

[

N(t,y,z)T−t

]

+ δ

∫ T−t

0

dse−δs

× E[

N (t,y,z)s I

(

Y (t,y)s < b

(

t + s, N (t,y,z)s

))]

.

Page 34: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Remark

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

21 / 24

The result may not be surprising to some extent...

Page 35: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Remark

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

21 / 24

The result may not be surprising to some extent...

Point: the integral equation for b is the “characteristic” ofthe problem in the sense that

free boundary b is determined as a unique sol. of theequation, and,

value function V is represented “in closed-form” with b.

Page 36: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Remark

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

21 / 24

The result may not be surprising to some extent...

Point: the integral equation for b is the “characteristic” ofthe problem in the sense that

free boundary b is determined as a unique sol. of theequation, and,

value function V is represented “in closed-form” with b.

Peskir’s change of variable formula for continuoussemimartingale with local-time on curve, (an extension ofIto-Meyer-Tanaka formula), plays an important role.

Page 37: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

How to Numerically Compute b and V ?

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

22 / 24

Using

q(s, dy′, dz′; t, y, z) := P(

Y (t,y)s ∈ dy′, N (t,y,z) ∈ dz′

)

,

we see

E

[

N(t,y,z)T−t

]

=KT y

Kt

E

(

z′

y′

)

q(T − t, dy′, dz′; t, y, z),

E[

N (t,y,z)s I

(

Y (t,y)s < b

(

t + s, N (t,y,z)s

))]

=Kt+sy

Kt

E

(

z′

y′

)

I(y′ < b(t + s, z′))q(s, dy′, dz′; t, y, z).

So, the integral equation for b is represented with q.

Page 38: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Boundary Crossing Probability

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

23 / 24

We see

N (t,y,z)s ≤ d

⇔ z ≤ d and wu ≥ ℓ(u) for all u ∈ [0, s].

where

ℓ(u) = ℓ(u; d, t, y, k) := f−1

(

t + u,Kt+u

d

)

− Kt+u

y.

Approximation schema to compute the joint probability

P (ws ≤ c1, wu ≥ ℓ(u) for all u ∈ [0, s])

have been studied by several works.

Page 39: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Related Topics for Future Works

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

24 / 24

More general (multi-dimensional, possibly) model withnumerics.

Large-time asymptotics (as T → ∞).

Page 40: Dynamic Protection for Bayesian Optimal Portfolio · Dynamic protection for Bayesian optimal portfolio Russian option pricing for local volatility model (Mathematical) Results/Contributions

Related Topics for Future Works

⊲ Problem

⊲ Outline

Applications

Results

⊲ Dynamic Prob.

⊲ Free-bdry Prob.

⊲ Main Theorem (1)

⊲ Main Theorem (2)

⊲ Remark

⊲ How to compute ?

⊲ Bdry Cross Prob.

⊲ Related Topics

24 / 24

More general (multi-dimensional, possibly) model withnumerics.

Large-time asymptotics (as T → ∞).

Comparison with American-OBPI strategy (by ElKaroui-Jeanblanc-Lacoste, El Karoui-Meziou, etc.)