View
1
Download
0
Category
Preview:
Citation preview
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Malliavin Calculus and its RecentApplications in Financial Mathematics
Yeliz Yolcu Okur
Institute of Applied Mathematics, Financial MathematicsMiddle East Technical University, Ankara
Istanbul Analysis Seminars11 April 2014, Karaköy, Istanbul
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Agenda
1 Introduction2 Preliminaries
Brownian Motion and its Properties3 Stochastic Differential Equations
Itô FormulaApplication
Black Scholes Model
4 Evolution of Malliavin Calculus TheoryStochastic Calculus of Variations: Approach I
The Derivative and Divergence OperatorsStochastic Calculus of Variations: Approach II
Wiener-Itô Chaos Expansion
5 Recent Studies on Malliavin CalculusComputation of Greeks under General StochasticVolatility Models
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Observations:Stock prices move randomly because of the efficientmarket hypothesis.Absolute change in the price of a stock is not a usefulquantity. Increase of one unit in a stock worth 10 units ismuch more significant than the one with a stock worth 100units.Hence, relative change must be modeled: dS/SdS/S = (µ+ σnoise)dt (more realistic, consider someenvironmental random effects)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Random variable
(Ω,F , Ftt≥0,P)
DefinitionA real valued random variable, X , is a real valuedF-measurable function defined on Ω. i.e.
X : Ω→ R
(Ω,F ,P)→ (R,B,PX )
B = (−∞, x ], x ∈ R,
PX ((−∞, x ]) = P(ω : X (ω) ∈ (−∞, x ])= P(ω : X (ω) ≤ x) = FX (x)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Process
DefinitionA real valued stochastic process is just a sequence of realrandom variables, i.e.
(t , ω)→ Xt (ω) : T × Ω→ R
is a stochastic process which is a parameterized collection ofreal-valued rvs Xtt∈T .
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Brownian Motion and its Properties
Brownian Motion
DefinitionA Brownian is a real valued, continuous stochastic processB(t , ω) , t ≥ 0, ω ∈ Ω, with independent and stationaryincrements, i.e.,
continuity: The map t → B(t , ω) is continous.independent increments: ∀s ≤ t , B(t)− B(s) is indepedentof Fs = σ(B(u),u ≤ s).
stationary increments: ∀s ≤ t , h > 0, B(t + h)− B(s + h)have the same probability law with B(t)− B(s).
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Brownian Motion and its Properties
Standard Brownian Motion
A standard Brownian motion, B(t , ω), t ∈ [0,T ], ω ∈ Ω, is aBrownian motion with the following properties:
B(0) = 0 P− a.s.E[B(t)] = 0E[B(t)2] = t
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Brownian Motion and its Properties
Simulation of Brownian Motion
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Brownian Motion and its Properties
Properties of Brownian Motion
cont. everywhere but nowhere differentiable,hits any and every real value no matter how large or smallit is,It has 1/2-self similarity, i.e. B(λt) = λ1/2B(t) for λ ≥ 0.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Brownian Motion and its Properties
Variation of Brownian Motion
Taylor and Duke (1972) showed that
Vp(B; a,b) <∞⇐⇒ p > 2
For p = 2, instead of considering the supremum over allpartitions, if we restrict ourselves to the sequences of partitions4n for which ||4n|| → 0, then we obtain the following result:
Theorem
Q2(B; a,b,4n)→ b − a in L2
as ||4n|| → 0. Moreover, the quadratic variation of a Brownianmotion B on the interval [a,b] is the limit of Q2(B; a,b,4n) anddenoted by < B,B >[a,b].
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Brownian Motion and its Properties
Stochastic Integration for Simple Processes
H(t , ω) =∑p
i=1 φi(ω)I[ti−1,ti )(t)where φ is Fti−1-measurable and bounded,0 = t0 < t1 < . . . < tp = T .
I(H)(t) :=∑
1≤i≤k
φi(B(ti)− B(ti−1)) + φk+1(B(t)− B(tk ))
for any t ∈ (tk , tk+1].
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Brownian Motion and its Properties
Stochastic Integration
PropositionLet H(t , ω), t ∈ [0,T ], ω ∈ Ω be a simple process. Then,
(∫ T
0 H(s)dB(s))0≤t≤T is a continuous Ft -martingale.
E
[(∫ T0 H(s)dB(s)
)2]
= E(∫ T
0 |H(s)|2ds)
E(
supt≤T |∫ t
0 H(s)dB(s)|2)≤ 4E
(∫ T0 |H(s)|2ds
)H =
(H(t , ω))0≤t≤T , (Ft )t≥0 − adapted process, E(∫ T
0 H(s)2ds)<∞
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Itô Process
X (t , ω), t ∈ [0,T ] is called Itô process if it can be written as
X (t) = X (0) +
∫ t
0K (s)ds +
∫ t
0H(s)dB(s),
whereX (0) is F0-measurable.∫ T
0 |K (s)|ds <∞ P− a.s.∫ T0 |H(s)|2ds <∞ P− a.s.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Differential Equations
More generally, consider the following equation:
dX (t) = b(t ,X (t))dt + σ(t ,X (t))dB(t) (1)
where X (0) = x ∈ R,b : [0,T ]× R→ R,σ : [0,T ]× R→ R
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Questions Come to Our Minds at First Glance
1 Can we obtain existence and uniqueness theorem for suchequations?
2 How can we solve?3 Answers:
1. b and σ satisfy Lipschitz and polynomial growthcondition and E[|x |2] <∞2. Itô formula
4 E = (Xs)0≤s≤T ,Ft −measurable continuous processess.t. E ( sups≤T |Xs|2) <∞
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Questions Come to Our Minds at First Glance
1 Can we obtain existence and uniqueness theorem for suchequations?
2 How can we solve?
3 Answers:1. b and σ satisfy Lipschitz and polynomial growthcondition and E[|x |2] <∞2. Itô formula
4 E = (Xs)0≤s≤T ,Ft −measurable continuous processess.t. E ( sups≤T |Xs|2) <∞
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Questions Come to Our Minds at First Glance
1 Can we obtain existence and uniqueness theorem for suchequations?
2 How can we solve?3 Answers:
1. b and σ satisfy Lipschitz and polynomial growthcondition and E[|x |2] <∞2. Itô formula
4 E = (Xs)0≤s≤T ,Ft −measurable continuous processess.t. E ( sups≤T |Xs|2) <∞
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Questions Come to Our Minds at First Glance
1 Can we obtain existence and uniqueness theorem for suchequations?
2 How can we solve?3 Answers:
1. b and σ satisfy Lipschitz and polynomial growthcondition and E[|x |2] <∞2. Itô formula
4 E = (Xs)0≤s≤T ,Ft −measurable continuous processess.t. E ( sups≤T |Xs|2) <∞
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Itô Formula
Itô Formula
Let X (t , ω) , t ∈ [0,T ] be an Itô process
X (t) = X (0) +
∫ t
0K (s)ds +
∫ t
0H(s)dB(s),
and f ∈ C1,2, then
f (t ,X (t)) = f (0,X (0)) +
∫ t
0
∂f∂s
(s,X (s))ds +
∫ t
0
∂f∂x
(s,X (s))dX (s)
+12
∫ t
0
∂f∂x2 (s,X (s))d < X ,X >s
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Itô Formula
Itô Formula in Practice
f (t , x) = x2, X (t) = B(t);
B(t)2 = t + 2∫ t
0B(s)dB(s)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Application
Financial Market
Market is ideal (no constraints on consumption, notransaction costs or taxes).
Bank account process (riskless asset) S0(t),
dS0(t) = S0(t)r(t)dt ,S0(0) = 1.
1 risky asset and price process is given by S1(t),
dS1(t) = S1(t)[µdt + σdB(t)],
S1(0) > 0.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Application
Financial Market
Market is ideal (no constraints on consumption, notransaction costs or taxes).Bank account process (riskless asset) S0(t),
dS0(t) = S0(t)r(t)dt ,S0(0) = 1.
1 risky asset and price process is given by S1(t),
dS1(t) = S1(t)[µdt + σdB(t)],
S1(0) > 0.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Application
Financial Market
Market is ideal (no constraints on consumption, notransaction costs or taxes).Bank account process (riskless asset) S0(t),
dS0(t) = S0(t)r(t)dt ,S0(0) = 1.
1 risky asset and price process is given by S1(t),
dS1(t) = S1(t)[µdt + σdB(t)],
S1(0) > 0.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Application
Simulation of Stock Price
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Malliavin calculus
Paul Malliavin1925-2010
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Books on Calculus of Variations
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Books on Calculus of Variations
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Evolution of Malliavin Calculus Theory
DefinitionMalliavin calculus is an infinite dimensional differential calculusfor functionals of an arbitrary Gaussian process.
1 originally introduced to study the regularity of the law offunctionals of Brownian Motion:work on Wiener space (Ω = C0[0,T ] of continuousfunctions ω : [0,T ]→ R, ||ω||∞ := supt∈[0,T ] |ω(t)|)
2 Applications to finance:Representation of functionals of stochastic processesStochastic control theoryStochastic differential equationsComputation of GreeksGeneral stochastic differential games
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Evolution of Malliavin Calculus Theory
DefinitionMalliavin calculus is an infinite dimensional differential calculusfor functionals of an arbitrary Gaussian process.
1 originally introduced to study the regularity of the law offunctionals of Brownian Motion:work on Wiener space (Ω = C0[0,T ] of continuousfunctions ω : [0,T ]→ R, ||ω||∞ := supt∈[0,T ] |ω(t)|)
2 Applications to finance:Representation of functionals of stochastic processesStochastic control theoryStochastic differential equationsComputation of GreeksGeneral stochastic differential games
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
Settings in Infinite Dimensional Framework
Let H be a separable Hilbert space (|| · ||H , 〈·, ·〉H )There exists probability space (Ω,G, µ)
M = W (h); h ∈ H := Family of centered rvs defined onthis space s.t.
h→W (h) is linear.E[W (h1)W (h2)] = 〈h1,h2〉HEach W (h) is Gaussian.
M is a closed subspace of L2(Ω) that is isometric to H.W (h1) and W (h2) are independent if and only if h1 and h2are orthogonal.The rvs eW (h),h ∈ H form a total subset of L2(Ω).
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
Theorem(Revuz & Yor Thm 1.2) Given a probability measure P on R,there exists a probability space (Ω,F , µ) and a sequence ofindependent rvs Xndefined on Ω s.t. Xn(P) = µ for every n.
Take an orthonormal basis en,n ≥ 1 in H. By the thm, thereis a probability space (Ω,G, µ) on which one can define asequence of independent reduced real Gaussian variables gnfor n ≥ 1. That is Ω = R⊗N, G = B⊗N and µ = µ⊗N1 , where µ1 is
the Gaussian measure on R (µ1(dx) = (2π)−1/2 exp(− |x |2
2 )dx).For each h ∈ H, the series
∑n≥1〈h,en〉Hgn converges in
L2(Ω,G, µ) to a random variable and
W (h) = limn→∞
∑n≥1
〈h,en〉Hgn
Indeed, Hn = Hn(W (h)),h ∈ H, ||h||H = 1
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
The Derivative Operator
There are several ways to define the Malliavin derivative forrandom vectors F : Ω→ Rn. Note that, the values of thederivative of a random variable will be H-valued.
DefinitionThe random variables of the type
F = f (W (h1), . . . , f (W (hn)),
with h1,h2, . . . ,hn ∈ H is called smooth functionals.If f ∈ C∞p (Rn), then the corresponding classes of smoothfunctionals is denoted by SIf f ∈ C∞b (Rn), then the family is denoted by SbIf f is polynomial, then the family is denoted by P
P ⊂ S, Sb ⊂ S.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
Derivative Operator on Smooth Functionals
DF =n∑
i=1
∂i f (W (h1), . . . ,W (hn))hi
For a fixed h ∈ H; set
F εh = f (W (h1) + ε〈h,h1〉H , . . .W (hn) + ε〈h,hn〉H)
Then,
DhF := 〈DF ,h〉H =ddε
F εh|ε=0.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
Properties
For any F ∈ SD(FG) = FDG + GDFD is closable operator; D : Lp(Ω)→ Lp(Ω; H), for anyp ≥ 1.E[〈DF ,h〉H ] = E[FW (h)]
E[G〈DF ,h〉H ] = −E[F 〈DG,h〉H ] + E[FGW (h)]
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
Let D1,p be the closure of the set S with respect to theseminorm for p ∈ [1,∞):
||F ||1,p :=(E[||F ||p] + E[||DF ||pH ]
)1/p
Observations:D1,p is dense in Lp(Ω),
||F ||k ,p :=(E[||F ||p] +
∑kj=1E[||DjF ||pH⊗j ]
)1/p
Dj,q ⊂ Dk ,p, for k ≤ j and p ≤ qD0,p = Lp(Ω)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
The Divergence Operator
DefinitionThe adjoint of the derivative operator D is denoted with δ. It isan unbounded operator on L2 (Ω; H) with values in L2 (Ω) and ithas the following properties:
1 The domain of δ is the set of H valued square integrablerandom variables u ∈ L2 (Ω; H) such that,
|E [〈DF ,u〉H ]| ≤ c ‖F‖2 , (2)
for all F ∈ D1,2 and some constant c, which depends on u.2 If u ∈ Dom (δ) then δ (u) ∈ L2 (Ω) and characterized by
E [Fδ (u)] = E [〈DF ,u〉H ] , (3)
for any F ∈ D1,2.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach I
Skorohod Integral
H = L2(T ,B, λ), where λ is the σ-finite atomless measuredefined on the set T , then, δ(u) is the Skorohod stochasticintegral of the square integrable process u ∈ L2(Ω× T );
δ(u) =
∫T
u(t)δB(t)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach II
Wiener-Itô Chaos Expansion
For the sake of simplicity assume T = [0,T ].
Definition
Let F be an FT measurable random variable in L2(Ω). Thenthere exists a unique sequence of functions, fn∞n=0, wherefn ∈ L2(T n) such that
F =∞∑
n=0
In(fn).
Here the convergence is in L2(Ω). Moreover, we have thefollowing isometry: ‖F‖2L2(Ω)
=∑∞
n=0 n!‖fn‖2L2(T n)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach II
Malliavin Derivative
Definition
Let F ∈ L2(Ω) be FT -measurable random variable with thefollowing chaos expansion:
F =∞∑
n=0
In(fn)
Then the Malliavin derivative DtF of F at time t can be definedas the following expansion:
DtF =∞∑
n=1
nIn−1(fn(·, t)), t ∈ T
where In−1(fn(·, t)) is the (n − 1)-fold iterated Itô integral offn(t1, t2, . . . , tn−1, t).
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach II
Skorohod Integral
Definition
Let u ∈ L2(Ω× T ) and for all t ∈ T , the rv u(t) is FTmeasurable. Let it has Wiener-Itô chaos expansion be
u(t) =∞∑
n=0
In(fn(·, t)),
where fn ∈ L2(T n+1) symmetric function of first n variable.Then;
δ(u) :=
∫T
u(t)δB(t) :=∞∑
n=0
In+1(fn)
when convergent in L2(Ω).
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach II
Here, fn, n = 1,2, . . . are the symmetric functions obtained by
fn(t1, . . . , tn, t) =1
n + 1
[fn(t1, . . . , tn, t) +
n∑i=1
fn(t1, · · · , ti−1, t , ti , . . . , tn, ti )
]
Remark
Note that
E
[∫T
u(t)2λ(dt)]
=∞∑
n=0
n!||fn||2L2(T n+1).
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Stochastic Calculus of Variations: Approach II
Example1 F = B(T ), DtF = 12 F = B2(T ), DtF = 2B(T )
3 F =∫ T
0 g(s)dB(s), DtF = g(t) (g is a deterministicfunction)
4 F =∫ T
0 g(s)dB(s), DtF =∫ T
t Dt (g(s))dB(s) + g(t) (g is astochastic process)
5 δ(1) = B(T )
6∫ T
0 B(T ) δB(t) = B(T )2 − T
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Research Areas
1 White noise generalization of the Clark-Ocone formulaunder change of measure.
2 A Malliavin calculus approach to general stochasticdifferential games with partial information.
3 SDE solutions in the space of smooth random variables.4 Analysis of volatility feedback and leverage effects using
high frequency data.5 Computation of Greeks in stochastic volatility models.6 A general formula for the Laplace transform of the hitting
times of non-Gaussian processes.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
General Stochastic Volatility Model
(joint work with B. Yılmaz, A. Inkaya and T. Sayer)
dSt
St= rtdt + σ (t ,Vt ) dWt , (4)
dVt = u (t ,Vt ) dt + v (t ,Vt ) dZt , (5)
where
dZt =[ρdWt +
√1− ρ2dW 1
t
],
d〈W ,Z 〉t = ρdt ,
for t ∈ [0,T ]. Here, r represents the risk free interest rate,u (t ,Yt ) and v (t ,Yt ) ∈ C2 ([0,T ]× R).
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
Computation of Greeks under General Stochastic Volatility Models
For the general stochastic volatility model, define
G (t ,T ) = σ (t ,Vt ) +
∫ T
t
∂σ
∂y(s,Vs) Dt Vs dWs −
∫ T
t
∂σ
∂y(s,Vs) Dt Vsσ (s,Vs) ds (6)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
Proposition
If S ∈ D2,2 and V ∈ D2,2 the following equalities hold fors ≤ t ≤ v ≤ T
DtST = ST G (t ,T )
DuST := 〈DST ,u〉L2([0,T ]) = ST∫ T
0 ut G (t ,T ) dt
DuDuST = ST
(∫ T0 ut G (t ,T ) dt
)2
+∫ T
0
∫ Ts usut DsG (t ,T ) dtds
DsG (t ,T ) = ∂σ
∂y (t ,Vt )DsVt +∫ T
t DsVv Dt Vv
[∂2σ∂y2 (v ,Vv ) (dWv − σ (v ,Vv ) dv)
]+∫ T
t∂σ∂y (v ,Vv ) DsDt Vv (dWv − σ (v ,Vv ) dv)
+∫ T
t
((∂σ∂y (v ,Vv )
)Dt Vv
(∂σ∂y (v ,Vv )
)DsVv
)dv ,
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
PropositionSuppose I is an open interval of R. Consider the families ofrandom variables
(F ζ)ζ∈I and
(Hζ)ζ∈I . These families are
continuously differentiable in Dom (δ) with respect to theparameter ζ ∈ I. Assume that (ut )t∈[0,T ] ∈ D1,2 satisfying
DuF ζ 6= 0 a.s. on∂ζF ζ 6= 0
, ζ ∈ I.
Furthermore assume that uHζ∂ζFζ
DuFζ is continuous with respect toζ in Dom (δ). Then we have
∂
∂ζEQ
[Hζ f
(Fζ)]
= EQ
[f(
Fζ)(Hζ∂ζFζ
DuFζδ (u)− Du
(Hζ∂ζFζ
DuFζ
)+ ∂ζHζ
)]
for any function f such that f(Fζ)∈ L2 (Ω), ζ ∈ I.
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
Proposition
Assume that F S0 := ST ∈ Dom (δ), HS0 := e−∫ T
0 rt dt ∈ Dom (δ),(ut )t∈[0,T ] ∈ D1,2 and satisfies
DuST 6= 0 a.s. on ∂S0ST 6= 0.
Furthermore assume thatuHS0∂S0
ST
DuSTis continuous with respect
to S0 in Dom (δ). Then, the delta under the general SV modeldefined in equations (4) and (5) is given as
∆ =e−
∫ T0 rt dt
S0E
f (ST )
1∫ T0 G (t ,T ) dt
δ (u.)−∫ T
0
∫ Ts usut DsG (t ,T ) dtds(∫ T
0 ut G (t ,T ) dt)2
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
First Variation and Malliavin Derivative
The first variation Yt := ∂∂V0
Vt of Vt is defined by:
Yt = exp(∫ t
0(u
′(s,Vs)−
12
(ρv′(s,Vs))2)ds +
∫ t
0v′(s,Vs)[ρdWs +
√1− ρ2dW 1
s ]
),
for all t ∈ [0,T ]. The Malliavin derivative of the stochastic volatility process Vt at time sis given by:
DsVt = v(s,Vs)ρYt
Ysexp
(∫ t
s(v
′(r ,Vr ))2 1− ρ2
2dr −
∫ t
sv′(r ,Vr )
√1− ρ2dW 1
r
)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
Stein & Stein Model
dSt
St= rdt + VtdWt , (7)
dVt = γ(Θ− Vt )dt + κdZt , (8)
where
dZt =[ρdWt +
√1− ρ2dW 1
t
],
〈dWt ,dZt〉 = ρdt ,
for t ∈ [0,T ].
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
DsVt = κρexp(−γ(t − s)),
for s ≤ t ≤ T .
DsG(t ,T ) = DsVt −∫ T
t
(∂σ
∂y(u,Vu)
)(∂σ
∂y(u,Vu)
)DtVuDsVudu
= κρe−γ(t−s)
(1 +
κρ
2γe−2Tγ +
κρ
2γ
). (9)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
Delta of Stein & Stein Model via Malliavin Calculus
∆ =e−rT
S0E
[f (ST )
(WT∫ T
0
(Vt + κρeγt
(∫ Tt e−γu (dWu − Vudu)
))dt
−∫ T
0
∫ Ts DsG(t ,T )dtds(∫ T
0
(Vt + κρeγt
(∫ Tt e−γu (dWu − Vudu)
))dt)2
)]
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
Stein & Stein Model for ρ = 0
CorollaryUnder the assumption that W and Z are two independentBrownian motion in equation (7), the delta of a European optionunder the Stein & Stein model is given by
∆ =e−rT
S0E
[f (ST )
(WT∫ T
0 Vtdt
)]. (10)
Introduction Preliminaries Stochastic Differential Equations Evolution of Malliavin Calculus Theory Recent Studies on Malliavin Calculus
Computation of Greeks under General Stochastic Volatility Models
THANK YOU FOR YOUR ATTENTION !
E-mail: yyolcu@metu.edu.tr
Recommended