getdocdd94. 162KB Jun 04 2011 12:05:04 AM

(1)

in PROBABILITY

A NOTE ON EXISTENCE AND UNIQUENESS FOR SOLUTIONS OF

MULTIDIMENSIONAL REFLECTED BSDES

JEAN-FRANÇOIS CHASSAGNEUX

Dpt. de Math., Université d’Evry Val d’Essonne&CREST

email: [email protected]

ROMUALD ELIE

CEREMADE, Université Paris-Dauphine&CREST

IDRIS KHARROUBI1

CEREMADE, Université Paris-Dauphine&CREST

SubmittedApril 1, 2010, accepted in final formDecember 5, 2010 AMS 2000 Subject classification: 93E20, 65C99, 60H30.

Keywords: BSDE with oblique reflections, Switching problems.

Abstract

In this note, we provide an innovative and simple approach for proving the existence of a unique solution for multidimensional reflected BSDEs associated to switching problems. Getting rid of a monotonicity assumption on the driver function, this approach simplifies and extends the recent results of Hu & Tang[4]or Hamadene & Zhang[3].

1 Introduction

The theory of Backward Stochastic Differential Equations (BSDEs for short) offers a large number of applications in the field of stochastic control or mathematical finance. Recently, Hu and Tang

[4]introduced and studied a new type of BSDE, constrained by oblique reflections, and associated

to optimal switching problems. Via heavy arguments, Hamadène and Zhang[3]generalized the

form of these multidimensional reflected BSDEs. They allow for the consideration of more general oblique constraints, as well as drivers depending on the global solution of the BSDE. Unfortunately, their framework requires the driver function to be increasing with respect to the components of the global solution of the BSDE. In the context of linear reflections of switching type, we are able to get rid of this limiting monotonicity assumption.

We provide in this note a new method to prove existence and uniqueness for such type of BSDEs.

We follow the classical scheme introduced for e.g. in[2], which consists in proving that a well

chosen operator is a contraction for a given norm. This is done via the introduction of a convenient

1_{RESEARCH PARTLY SUPPORTED BY THE EUROPEAN RESEARCH COUNCIL UNDER THE GRANT 228053-FIRM.}

(2)

one dimensional dominating BSDE and the use of a standard comparison theorem.

The rest of this note is organized as follows. In Section 2, we present our main result, namely Theorem 2.1. In Section 3, we give its proof which requires several intermediary results.

2 Framework and main result

LetT >0 be a given time horizon and(Ω,F,P₎_{be a stochastic basis supporting a}_q_-dimensional

Brownian motionW,q≥1. F_{= (}_F

t)t≤T is the completed filtration generated by the Brownian

motionW, andPdenotes theσ−algebra on[0,T]×Ωgenerated byF₋_{progressively measurable}

processes. In the following, we shall omit the dependence onω∈Ω when it is clearly given by

the context.

We introduce the following spaces of processes:

• S2_(resp. _S2

c ) is the set of R−valued, adapted and càdlàg

2_{(resp. continuous) processes}

(Yt)0≤t≤Tsuch thatkYkS2 := E

h

sup_t∈[0,T]|Yt|2

i1₂

< ∞,

• H2is the set ofRq₋_{valued, progressively measurable process}₍_Z_t₎₀_≤_t_≤_T_{such that}_k_Z_k_H2 := Eh RT

0 |Zt| 2_d_ti

1 2

< ∞,

• K2_(resp._K2

c ) is the subset of nondecreasing processes(Kt)0≤t≤T∈ S2(resp.(Kt)0≤t≤T∈

S2

c ), starting fromK0=0.

We are given a matrix valued continuous processC= (Ci j₎

1≤i,j≤d such thatCi j∈ Sc2, fori,j∈ I,

whereI :={1, . . . ,d},d≥2, and satisfying the following structure condition







(i)Cii

. =0 , fori∈ I ;

(ii)inf0≤t≤TC i j

t >0 , fori,j∈ I2, withi6=j;

(iii)inf₀_≤_t_≤_TC_ti j+C_tjl−Cil

t >0, fori,j,l∈ I3, withi6= j, j6=l;

(2.1)

In “switching problems”, the quantity C interprets as a cost of switching. As in [4] or[3] in a

more general framework, this assumption makes instantaneous switching irrelevant. We introduce the family of random closed convex sets(C_t)₀_≤_t_≤_T associated toC:

Ct:=

y∈Rd_| _yi_≥_max j∈I(y

j₋_Ci j

t ), i∈ I

, 0≤t≤T ,

and the oblique projection operatorP ontoC defined by

P_t:y∈Rd _7→

max

j∈I

¦

yj−C_ti j©

i∈I

, 0≤t≤T.

We are also given a terminal random variableξ∈[L2(F_T)]d _{valued in}_C

T, where L2(FT)is the

set ofF_T-measurable random variableX satisfyingE|X|2

<∞.

(3)

We then consider the following system of reflected BSDEs: find(_Y˙_{, ˙}_Z_{, ˙}_K)∈[S2

c × H

2_{× K}2

c ]

d _{such that}



 

 

˙

Y_ti=ξi₊RT

t f i₍_s_{, ˙}_Y

s, ˙Zs)ds−

RT

t Z˙ i

s.dWs+K˙Ti −K˙ i

t, 0≤t≤T,

˙

Yi

t ≥maxj∈I{Y˙

j t −C

i j

t }, 0≤t≤T,

RT

0[Y˙

i

t −maxj∈I{Y˙tj−C i j

t }]d ˙Kti=0 , 0≤i≤d,

(2.2)

where f :Ω×[0,T]×Rd×Rq×d →Rm _isP⊗ B(Rd)⊗ B(Rq×d)−measurable and satisfies the

following Lipschitz property: there exists a constant L>0 such that

|f(t,y,z)−f(t,y′,z′)| ≤ L(|y−y′|+|z−z′|), for all(t,y,y′,z,z′)∈[0,T]×[Rd_]2_×_[Rq×d_]2_,P₋_{a.s. We also assume that}

E





ZT

0

|f(t, 0, 0)|2_{d t}



 < ∞.

The existence and uniqueness of a solution to the system (2.2) has already been derived with the addition of one of the following assumptions:

• (H1)The cost processCis constant and, fori∈ I, thei−th component of the driver

func-tionf depends only on yi_{, the}_i₋_{th component of} _y_{, and}_zi_{, the}_i₋_{th column of}_z_{. (see Hu}

& Tang[4])

• (H2)For i ∈ I, the i−th component of the random driver f depends on y and thei−th

column ofz, and is nondecreasing in yjfor j6=i. (see Hamadène & Zhang[3])

We now introduce the following weaker assumption:

• (Hf)Fori∈ I, thei−th component of the random function f depends on y and thei−th

column of the variablez:

fi₍_t_,_y_,_z_{) =}_fi₍_t_,_y_,_zi₎_, ₍_t_,_y_,_z₎_∈_[0,_T_]_×_Rd_×_Rq×d_.

The main contribution of this note is the following result.

Theorem 2.1. Under(Hf), the BSDE(2.2)admits a unique solution.

3 Proof of Theorem 2.1

The proof divides in three steps. First, we slightly generalize the switching representation result

presented in Theorem 3.1 of[4], allowing for the consideration of random driver and costs.

(4)

3.1 The optimal switching representation property

We first give a key representation property for the solution of (2.2) under the following assump-tion:

• (H3)Fori ∈ I, the i−th component of the random driver f depends only on yi and the

i−th column of the variablez:

fi(t,y,z) =fi(t,yi,zi), (i,t,y,z)∈ I ×[0,T]×Rd×Rq×d_.

Observe that(H1) =⇒ (H3) =⇒ (H2). We first provide the existence of a solution to (2.2)

under Assumption(H3).

Proposition 3.1. Under(H3), the BSDE(2.2)has at least one solution.

Proof. Since(H3)is stronger than(H2), the existence of a solution is provided by Theorem 3.2 in [3]. Due to the the particular form of switching oblique reflections and for sake of completeness, we decide to report here a simplified sketch of proof.

We use Picard iteration. Let (Y.,0_,_Z.,0₎ _∈ _[_S2_{× H}2_]d _{be the solution to the following BSDE}

For i∈ I andn≥1, define recursively Yi,n_{as the first component of the unique solution (see}

Theorem 5.2 in[1]) of the reflected BSDE



upper bounded by ¯Y the first component of the unique solution to

¯

for anyi∈ I, the smallest fi_{-supermartingale ˜}_Yi _{with lower barrier max}

(5)

Fori∈ I, we directly deduce from (3.2) that ˙Yi≥_Y˜i_{, and, since}₍_Y.,n₎

n∈Nis increasing, a direct

comparison argument leads toYi,n≤_Y˜i_{, for}_n_∈_N_{. Therefore, we get ˙}_Y ₌_Y_˜ _{so that ˙}_Y _satisfies

(2.2).

It only remains to prove that ˙Y is continuous. In the sequel, we consider ˜Ωa measurable subset of

Ω, such thatP₍_{Ω) =}˜ _{1 and (3.3)-(2.1) hold}_everywhere_{on ˜}_{Ω. We now look towards a contradiction.}

We suppose that ˙Y is not continuous, i.e. there exists a measurable subset B of ˜Ω satisfying

P₍_B₎_>_{0 and such that}

∀ω∈B, ∃{t(ω),i₀(ω)} ∈[0,T]× I s.t. ∆Y˙_ti0₍_ω(ω₎):=Y˙_ti0₍_ω(ω₎)−Y˙_ti0₍_ω(ω₎−)6=0 .

We fix ω ∈ B and deduce from (3.3) that ∆Y˙i0(ω)

t(ω) = −∆K

i0(ω)

t(ω) < 0 so that ˙Y

i0(ω) t(ω)− = Y˙

i1(ω) t(ω)−−

C_ti0₍_ω(ω₎)i1(ω)for somei1(ω)6=i0(ω). This leads directly to ˙

Yi1(ω) t(ω)−=Y˙

i0(ω) t(ω)−+C

i0(ω)i1(ω) t(ω) > Y˙

i0(ω) t(ω) +C

i0(ω)i1(ω) t(ω) ≥ Y˙

i1(ω) t(ω) ,

which implies∆Y˙_ti1₍_ω(ω₎)<0. The exact same argument leads to the existence of an integeri₂(ω)∈ I such thati₂(ω)6=i₁(ω)and

˙

Y_ti2₍_ω(ω₎−)=Y˙

i1(ω)

t(ω)−+C

i1(ω)i2(ω)

t(ω) > Y˙

i1(ω)

t(ω) +C

i1(ω)i2(ω)

t(ω) ≥ Y˙

i2(ω)

t(ω) ,

so that∆Y˙i2(ω)

t(ω) <0. SinceI is finite, we get, after repeating this argument at mostdtimes, the

existence of a finite sequence(ik(ω))0≤k≤n(ω) such thatin(ω)(ω) =i0(ω)and ˙Y

ik(ω)

t(ω)−= Y˙

ik+1(ω) t(ω)− −

Cik(ω)ik+1(ω)

t(ω) for anyk<n(ω). Therefore, we have

n(ω) X

k=1

Cik−1(ω)ik(ω)

t(ω) =0 . (3.4)

Furthermore, using recursively the structural condition (2.1) (iii), we easily compute

n(ω) X

k=1

Cik−1(ω)ik(ω)

t(ω) ≥ C

i0(ω)i1(ω)

t(ω) +C

i1(ω)in(ω)(ω)

t(ω) .

Sincei_n(ω)(ω) =i0(ω), we deduce from (2.1) (iii) and (i) that

n(ω) X

k=1

Cik−1(ω)ik(ω)

t(ω) > C

i0(ω)i0(ω)

t(ω) = 0 . (3.5)

Since equations (3.4) and (3.5) are satisfied for any ω ∈ B, this leads to a contradiction and

concludes the proof.

Under(H3), a slight generalization of Theorem 3.1 in[4]allows to represent the process(Y˙i₎ i∈I as the value process of an optimal switching problem on a family of one-dimensional BSDEs.

More precisely, we consider the set of admissible3 _strategies _A _{consisting in the sequence} _a₌

(θ_k,α_k)_k≥0with

3_{Our definition of admissible strategies slightly differs from the one given in}_[4]_{Definition 3.1. Indeed, the authors}

assume thatNis inL2(FT)rather thanA∈ S2. In fact, it is clear that in the context of constant cost processCsatisfying

(6)

• (θk)k≥0a nondecreasing sequence of stopping times valued in[0,T], and such that there

exists an integer valued random variableNa,F_Tmeasurable andθNa=T,P−a.s.,

• (α_k)_k_≥₀a sequence of random variables valued inI such thatα_k isF_θ_k-measurable for all

k≥0,

• the processAadefined byAa_t:=P_k_≥₁Cαk−1αk

θk 1θk≤tbelonging toS 2_.

The state process associated to a strategya∈ A is denoted(a_t)₀≤t≤T and defined by at =

X

k≥1

αk−11(θk−1,θk](t), 0≤t≤T.

For(t,i)∈[0,T]× I, we also defineA_t_,_ithe subset of admissible strategies restricted to start in stateiat timet. For(t,i)∈[0,T]× I anda∈ A_t_,_i, we consider(Ua_,_Va₎_{∈ S}2_{× H}2_{the unique} solution of the switched BSDE:

U_ra = ξaT₊

Z T

r

fas₍_s_,_Ua

s,V a s)ds−

Z T

r

V_sa.dWs−AaT+A a

r, t≤r≤T.

As in Theorem 3.1 in[4], the correspondence betweenY and the switched BSDEs is given by the

following proposition which provides, as a by-product, the uniqueness of solution to (2.2) under (H3).

Proposition 3.2. Under(H3), there exists a∗_{∈ A}

t,i such that

˙

Yi

t = U a∗

t = ess sup

a∈At,i

Ua

t , (t,i)∈[0,T]× I.

Proof. The proof follows from the exact same reasoning as in the proof of Theorem 3.1 in[4]

having in mind that in our framework, the driver f and the costsCare random. This implies that

Na∗, the number of switch of the optimal strategy, is only almost surely finite. This is the only

difference with[4]whereNa∗ belongs toL2(F_T).

3.2 The contraction operator

We suppose that(Hf)is in force and introduce the operatorΦ:[H2_]d_→_[_H2_]d_,_Γ_7→_Y _{:= Φ(Γ),}

where(Y,Z,K)∈(S2_{× H}2_{× K}2₎d _{is the unique solution of the BSDE}



 

 

Yi t =ξi+

RT

t f i₍_s_,_Γ

s,Zsi)ds−

RT

t Z i

s.dWs+KTi −Kti, 0≤t≤T, Yi

t ≥maxj∈I{Ytj−C i j

t }, 0≤t≤T,

RT

0[Y

i

t −maxj∈I{Y

j t −C

i j

t }]dKti=0 , i∈ I .

(3.6)

Since(Hf)holds, the random driver(ω,t,z)7→[fi(t,Γ_t(ω),zi)]_i_∈I satisfies(H3), for anyΓ∈

[H2_]d_{. Therefore, the existence of a unique solution to (3.6) is given by Proposition 3.1 and}

Proposition 3.2 andΦis well defined. Furthermore, observe thatΦis valued in[S2_]d_.

In order to prove thatΦis a contraction on[H2_]d_{, we introduce, as in e.g.} _[₂_{], the norm}_k_._k

2,β defined on[H2_]d _by

kYk2,β :=

E

hZ T

0

eβt|Y_t|2d ti 1 2

(7)

We now state the contraction property forΦ, whose proof is postponed to the last section of this note.

Proposition 3.3. Forβlarge enough, the operatorΦis a contraction on the Banach space([H2_]d_,_k_._k

2,β),

i.e. Φis k−Lipschitz continuous with k<1for this norm.

Proof of Theorem 2.1.As a consequence, there exists a unique fixed point in[H2]d forΦ. Since Φis valued in[S2]d, there is a continuous version (still denotedY) of this fixed point belonging

to[S2_]d_{, see footnote 5 p. 21 in}_[₂_{]. Hence, this version} _Y _{and the corresponding processes}

(Z,K)∈[H2_{× K}2_]d _{are the unique solution of (2.2).}

3.3 Contraction via domination

This section is dedicated to the proof of Proposition 3.3.

Proof of Proposition 3.3. We consider two processes1Γand2Γbelonging to[H2_]d _{and denote}

1_Y _{:= Φ(}1_Γ)_and2_Y _{:= Φ(}2_Γ).

Step 1:Auxiliary dominating BSDE.

Let us introduce the following BSDE:



Step 2:Switching representation.

We fix(t,i)∈[0,T]× I and, for anya∈ A_t_,_i, denote by(Uˇa_{, ˇ}_Va₎_and₍j_Ua_,j_Va_{), for} _j₌_{1 and}

2, the respective solutions of the following one-dimensional BSDEs:

ˇ

where ˇa∈ At,iis the associated optimal strategy given in Proposition 3.2.

Using a comparison argument , we easily check that ˇUa _≥1_Ua_∨2_Ua_{, for any strategy}_a_{∈ A}

t,i.

This estimate combined with (3.8) leads to ˇYi

t ≥1Y i t∨2Y

i

t . Since ˇais an admissible strategy for

(8)

This directly leads to

Step 3:Contraction property.

We first control the first term on the right hand side of (3.9). Denoting δUaˇ _:= _U_ˇaˇ₋1_Uˇa _and

Combining these two estimates with the inequality 2x y≤ 1

βx

forβ≥L. Taking the expectation in the previous expression and observing that the second term

in (3.9) is treated similarly, we deduce

E

2-Lipschitz continuous for thek · k2,β norm, which concludes the

proof of the proposition.

References

[1] El Karoui N., C. Kapoudjan, E. Pardoux, S. Peng, M.C. Quenez,Reflected solutions of

Back-ward SDE’s and related obstacle problems for PDE’s, The Annals of Probability,25(2) (1997),

702-737.MR1434123

[2] El Karoui N., S. Peng, M.C. Quenez,Backward Stochastic Differential Equation in finance,

Mathematical finance,7(1) (1997), 1-71.MR1434407

[3] Hamadène S. and J. Zhang,Switching problem and related system of reflected backward

(9)

[4] Hu Y. and S. Tang,Multi-dimensional BSDE with oblique Reflection and optimal switching,

Prob. Theory and Related Fields,147(1-2) (2008), 89-121.MR2594348

[5] Peng S., Monotonic limit theory of BSDE and nonlinear decomposition theorem of

Doob-Meyer’s type, Prob. Theory and Related Fields,113(1999), 473-499.MR1717527

[6] Peng S. and M. Xu,The smallest g-supermartingale and reflected BSDE with single and double