Professional Documents
Culture Documents
Research Article
Positive Solutions of European Option Pricing with CGMY
Process Models Using Double Discretization Difference Schemes
Copyright © 2013 R. Company et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
This paper deals with the numerical analysis of PIDE option pricing models with CGMY process using double discretization
schemes. This approach assumes weaker hypotheses of the solution on the numerical boundary domain than other relevant papers.
Positivity, stability, and consistency are studied. An explicit scheme is proposed after a suitable change of variables. Advantages of
the proposed schemes are illustrated with appropriate examples.
jump diffusion models with finite jump intensity. Tavella and dividend paid by the asset, respectively. The payoff function
Randall in [24] use an implicit time discretization and pro- 𝑓(𝑆) for a vanilla call option is given by
pose a stationary rapid convergent iterative method to solve
the full matrix problem quoted above, but with poor numeri- 𝑓 (𝑆) = max (𝑆 − 𝐸, 0) , (4)
cal analysis. A generalization of their iterative method to price
American options is proposed in [25].
One of the most relevant and versatile Lévy models is the where 𝐸 is the strike price.
one proposed by Carr et al., the so-called CGMY model [8], Like [20, 23] for jump diffusion models we transform the
that belongs to the family of KoBoL models [9]. It is con- original PIDE problem in order to remove the convection
sidered a prototype of the general class of models with term to avoid possible numerical oscillations. With respect
jumps and enjoys widespread applicability. The CGMY model to the singularity of the integral kernel quoted above, the
allows diffusions and jumps of both finite and infinite activity. jump component in the neighborhood of log jump size
The CGMY Lévy density is given by zero is approximated by using a Taylor expansion, like
[18, 19].
−𝐺|𝑦|
{ 𝐶𝑒 The selection of the boundary conditions of the numerical
{
{
{ 1+𝑌 , 𝑦 < 0, domain, the discretization of the infinite domain of the
{ 𝑦
{
integration, and matching the discretization of both the
] (𝑦) = { (1)
{
{ 𝐶𝑒−𝑀|𝑦| differential and the integral part are important challenges.
{
{
{ 1+𝑌 , 𝑦 > 0, Some authors, like those of [18], assume a particular behavior
{ 𝑦 of the solution outside of the bounded numerical domain.
where 𝐶 > 0, 𝐺 ≥ 0, 𝑀 ≥ 0, and 𝑌 < 2. The para- In order to weaken these hypotheses we do not truncate the
meter 𝑌 allows controling the fine structure of asset return infinite integral and we use a nonuniform partition of the
distribution. For 𝑌 < 0, the Lévy process is of finite activity; complete unbounded domain, allowing a proper matching
that is, the measure is finite; ∫ ](𝑦)𝑑𝑦 < ∞. For 0 ≤ of the discretizations of the differential and integral parts
𝑌 ≤ 1, it is of infinity activity but finite variance; that is, by assuming asymptotic linear behavior of the solution.
This strategy involves a double discretization with two spa-
∫|𝑦|<1 𝑦](𝑦)𝑑𝑦 < ∞. Finally, for 1 < 𝑌 < 2, both the activity
tial stepsize parameters that will allow a better flexibility
and variation are infinite. Note that, for 𝑌 = 0, one gets the to improve the approximation in different zones of the
well known Variance Gamma process proposed by Madan domain.
and Seneta [29] as a particular case. So CGMY model is an
This paper is organized as follows. In Section 2, the
extension of the Variance Gamma model [7].
integral part of (2) is approximated in a neighborhood of
The authors in [22] use FD methods discretizing the equ-
𝑦 = 0 to obtain a new PIDE integral part extended outside
ation in space by the collocation method and using explicit
a neighborhood of 𝑦 = 0. Then a variable transformation
difference backward schemes focused on the case of infinite
is developed in order to remove both the convection and
activity and finite variation.
reaction terms of the differential part. Following the idea
Very recently [28] proposed an efficient three-time-level
developed in [20] for the finite activity case, a double
finite difference method for the infinite activity Lévy model.
discretization explicit scheme for solving PIDE problem
Second-order convergence rate is shown in numerical expe-
(2) is constructed in Section 3 for the infinite activity case
riments although the numerical analysis of the method is not
affording the new challenges. Positivity and stability of the
developed.
numerical solutions given by the scheme are studied in
The aim of this paper is the construction and numerical
Section 4. Consistency of the scheme is treated in Section 5.
analysis of an explicit finite difference scheme of the PIDE for
In Section 6, some illustrative numerical examples show the
the CGMY model with measure ](𝑦) given by (1):
advantages of the new discretization approach showing how
𝜕𝑉 the double discretization allows flexible improvement of the
𝜕𝜏 accuracy in different zones of the domain.
If V = (V1 , V2 , . . . , V𝑛 )𝑇 is a vector in R𝑛 , we denote its
𝜎2 2 𝜕2 𝑉 𝜕𝑉 infinite norm ‖V‖∞ = max{|V𝑗 |; 𝑖 ≤ 𝑗 ≤ 𝑛}. Vector V is said
= 𝑆 2
+ (𝑟 − 𝑞) 𝑆 − 𝑟𝑉
2 𝜕𝑆 𝜕𝑆 to be nonnegative if V𝑗 ≥ 0 for all 1 ≤ 𝑗 ≤ 𝑛; then we denote
+∞
𝜕𝑉 V ≥ 0. For a matrix 𝐴 = (𝑎𝑖𝑗 )𝑚×𝑛 in R𝑚×𝑛 , we denote ‖𝐴‖∞ =
+∫ ] (𝑦) [𝑉 (𝑆𝑒𝑦 , 𝜏) − 𝑉 (𝑆, 𝜏) − 𝑆 (𝑒𝑦 − 1) ] 𝑑𝑦, max1≤𝑖≤𝑚 {∑𝑛𝑗=1 |𝑎𝑖𝑗 |}. Matrix 𝐴 is said to be nonnegative if
−∞ 𝜕𝑆
𝑎𝑖𝑗 ≥ 0 for all 1 ≤ 𝑖 ≤ 𝑚, 1 ≤ 𝑗 ≤ 𝑛, and we denote 𝐴 ≥ 0.
𝑆 ∈ (0, ∞) , 𝜏 ∈ (0, 𝑇] ,
The exponential integrals have a major role in evaluating
(2) important class of integrals. Let 𝑠 and 𝑧 be continuous (real
𝑉 (𝑆, 0) = 𝑓 (𝑆) , 𝑆 ∈ (0, ∞) . (3) or complex) variables; the exponential integral of order 𝑠
denoted by 𝐸𝑠 (𝑧) is given by [30]
Here 𝑉(𝑆, 𝜏) is the option price depending on the under-
lying asset 𝑆, the time 𝜏 = 𝑇 − 𝑡, 𝜎 is the volatility parameter, ∞
and 𝑟 and 𝑞 are the risk-free interest and the continuous 𝐸𝑠 (𝑧) = ∫ 𝑡−𝑠 exp (−𝑧𝑡) 𝑑𝑡. (5)
1
Abstract and Applied Analysis 3
can be evaluated with high accuracy using the exponential × [𝐸1+𝑌 (𝜀 (𝐺 + 𝑘)) + 𝐸1+𝑌 (𝜀 (𝑀 − 𝑘))] ,
integrals ([30, 31], chapter 7). Let us denote (13)
𝑓̂ (𝛼, 𝑀, 𝜀) = 𝑀𝛼−1 Γ (1 − 𝛼) − 𝜀1−𝛼 𝐸𝛼 (𝜀𝑀) , (9) 𝜆 (𝜀) = 𝐶𝜀−𝑌 (𝐸1+𝑌 (𝐺𝜀) + 𝐸1+𝑌 (𝑀𝜀)) . (14)
where Γ denotes the gamma function and 𝐸𝛼 is the expone- Hence, the problem (2) takes the following form:
ntial integral. Considering the expression (1) and the expo-
nential integral (5) for 𝑀 > 2, one gets 𝜕𝑉 𝜎̂2 2 𝜕2 𝑉 𝜕𝑉
= 𝑆 + (𝑟 − 𝑞 − 𝛾 (𝜀)) 𝑆
𝜕𝜏 2 𝜕𝑆2 𝜕𝑆
2
2 − (𝑟 + 𝜆 (𝜀)) 𝑉
𝜎2 (𝜀) = 𝐶 ∑ (−1)𝑘 ( ) (15)
𝑘
𝑘=0
+ ∫ ] (𝑦) 𝑉 (𝑆𝑒𝑦 , 𝜏) 𝑑𝑦 + O (𝜀3−𝑌 ) ,
× [𝑓̂ (1 + 𝑌, 𝐺 + 𝑘, 𝜀) + 𝑓̂ (1 + 𝑌, 𝑀 − 𝑘, 𝜀)] . 𝑅2
(10) where 𝜎̂ = 𝜎̂ (𝜀) = 𝜎2 + 𝜎2 (𝜀).
2 2
Notice that (10) holds for 𝑌 ∈ (0, 1) ∪ (1, 2). For the particular In order to remove the convection and reaction terms
case where 𝑌 = 0, one gets from (15), let us introduce the following transformation of
variables:
(𝐺 + 1)2 (𝑀 − 1)2 𝑥 = exp [(𝑟 − 𝑞 − 𝛾 (𝜀)) 𝜏] 𝑆, 𝑈 (𝑥, 𝜏)
𝜎2 (𝜀) = 𝐶 [ln
𝐺 (𝐺 + 2) 𝑀 (𝑀 − 2) (16)
= exp [(𝑟 + 𝜆 (𝜀)) 𝜏] 𝑉 (𝑆, 𝜏) .
2
2
− ∑ ( ) (−1)𝑘 Hence the problem (15) is approximated by the following
𝑘 form:
𝑘=0
𝜕𝑈 𝜎̂2 2 𝜕2 𝑈
× (𝐸1 [𝜀 (𝐺 + 𝑘)] + 𝐸1 [𝜀 (𝑀 − 𝑘)]) ] , = 𝑥 + 𝐽, 𝑥 ∈ (0, +∞) , 𝜏 ∈ (0, 𝑇] ,
𝜕𝜏 2 𝜕𝑥2 (17)
(11) 𝑈 (𝑥, 0) = 𝑓 (𝑥) , 𝑥 ∈ (0, +∞) ,
4 Abstract and Applied Analysis
𝑙 𝑙 𝑙
{ 𝑢𝑖−1 − 2𝑢𝑖 + 𝑢𝑖+1
{
{ , 0 ≤ 𝑖 < 𝑁,
{
{ ℎ2
𝑙
Δ𝑖 ≡ { (25)
{
{ 𝑙
𝑢𝑖−1 𝑢𝑖𝑙 𝑙
𝑢𝑖+1
{
{2 [ − + ] , 𝑁 ≤ 𝑖 ≤ 𝑁 + 𝑀 − 2,
{ ℎ𝑖−1 (ℎ𝑖 + ℎ𝑖−1 ) ℎ𝑖 ℎ𝑖−1 ℎ𝑖 (ℎ𝑖−1 + ℎ𝑖 )
value theorem for integrals [34, page 1063], the two remaining Case 3 (𝑥𝑖 ≥ 𝐴𝑒𝜀 ). Here 𝑥𝑖2 > 𝑥𝑖1 ≥ 𝐴 and the approximation
integrals are approximated by of 𝐽𝑖𝑙 is given by
𝑁−1
𝑥𝑖 𝑒−𝜀 𝑔
∫ 𝑔 (𝑥𝑖 , 𝜙) 𝑈 (𝜙, 𝜏 ) 𝑑𝜙 𝑙 𝐽̂𝑖𝑙 = ℎ ∑ 𝑔𝑖𝑗 𝑢𝑗𝑙 + 𝑖𝑁 (ℎ + 𝐴𝛿) 𝑢𝑁
𝑙
𝑥𝑖1 𝑗=1 2
𝑥𝑖 𝑒−𝜀 𝑖 −1
𝐼𝑖𝑙 𝑔 (𝑥𝑖 , 𝜙) 𝑑𝜙) 𝑢𝑖𝑙1 = 𝑔̃𝑖,𝑖1 𝑢𝑖𝑙1 , 𝛿 1
≈ = (∫ + ∑ 𝑔 𝑥2 𝑢 𝑙
𝑥𝑖1 𝐴 𝑗=𝑁+1 𝑖𝑗 𝑗 𝑗
(27)
𝑥𝑖2 (31)
∫ 𝑔 (𝑥𝑖 , 𝜙) 𝑈 (𝜙, 𝜏𝑙 ) 𝑑𝜙 𝑥𝑖21 𝛿 𝑥𝑖22 𝛿
𝑥𝑖 𝑒𝜀 +( 𝑔𝑖,𝑖1 + 𝑔̃𝑖,𝑖2 ) 𝑢𝑖𝑙1 + ( 𝑔𝑖,𝑖2 + 𝑔̃𝑖,𝑖2 ) 𝑢𝑖𝑙2
2𝐴 2𝐴
𝑥𝑖2
≈ 𝐻𝑖𝑙 = (∫ 𝑔 (𝑥𝑖 , 𝜙) 𝑑𝜙) 𝑢𝑖𝑙2 = 𝑔̃𝑖,𝑖2 𝑢𝑖𝑙2 .
𝑥𝑖 𝑒𝜀 𝛿 𝑁+𝑀−1
+ ∑ 𝑔 𝑥2 𝑢 𝑙 .
𝐴 𝑗=𝑖2 +1 𝑖𝑗 𝑗 𝑗
Let us denote
Assuming that 𝑈(𝜙, 𝜏) tends to zero at least linearly as
] (ln 𝑥𝑗 /𝑥𝑖 ) 𝜙 tends to zero one has 𝑔(𝑥, 𝜙)𝑢(𝜙, 𝜏) → 0 by (1)
𝑔𝑖,𝑗 = . (28) and (28). On the other hand, assuming linear behavior of
𝑥𝑗
the solution for large values of 𝜙, the integrand of (20)
𝑔(𝑥, 𝐴/𝑧)𝑈(𝐴/𝑧, 𝜏)(1/𝑧2 ) → 0, as 𝑧 → 0. Thus, both the
Depending on the location of 𝑥𝑖 for each 𝑖 with 1 < 𝑖 ≤ 𝑁 +
terms involving 𝑢0𝑙 and 𝑢𝑁+𝑀
𝑙
do not appear in the expressions
𝑀 − 2, we approximate 𝐽𝑖𝑙 given by (26) in the following form.
of (29)–(31). Taking into account (25)–(31) the resulting
Case 1 (𝑥𝑖 < 𝐴𝑒−𝜀 ). Note that in this case 𝑥𝑖1 < 𝐴, and thus difference scheme for the PIDE problem (21) takes the form
𝑙
𝐽𝑖,1
𝑖1 −1
is approximated by ℎ ∑𝑗=1 𝑔𝑖𝑗 𝑢𝑗𝑙 + ((ℎ/2)𝑔𝑖,𝑖1 + 𝑔̃𝑖,𝑖1 )𝑢𝑖𝑙1 . 𝜎2 2 𝑙
𝑘̂
𝑢𝑖𝑙+1 = 𝑢𝑖𝑙 + 𝑥 Δ + 𝑘𝐽̂𝑖𝑙 , 1 ≤ 𝑖 ≤ 𝑁 + 𝑀 − 2. (32)
Also one has 𝑥𝑖2 < 𝐴 in the domain of the integral 𝐽𝑖,2 𝑙
and 2 𝑖 𝑖
𝑁−1
is approximated by ((ℎ/2)𝑔𝑖,𝑖2 + 𝑔̃𝑖,𝑖2 )𝑢𝑖2 + ℎ ∑𝑗=𝑖2 +1 𝑔𝑖𝑗 𝑢𝑗𝑛 +
𝑙 In order to obtain a complete difference scheme, we include
the initial and boundary conditions. From (22), we have
(𝑔𝑖𝑁/2)(ℎ + 𝐴𝛿)𝑢𝑁 𝑙
+ (𝛿/𝐴) ∑𝑁+𝑀−1 2 𝑙
𝑗=𝑁+1 𝑔𝑖𝑗 𝑥𝑗 𝑢𝑗 , taking into
account (20) for 𝑥𝑗 > 𝐴. Thus 𝑢𝑖0 = max (𝑥𝑖 − 𝐸, 0) , 1 ≤ 𝑖 ≤ 𝑁 + 𝑀 − 1. (33)
On the other hand, for a vanilla call option the boundary
𝑖1 −1
ℎ condition for 𝑖 = 0 is
𝐽̂𝑖𝑙 = ℎ ∑ 𝑔𝑖𝑗 𝑢𝑗𝑙 + ( 𝑔𝑖,𝑖1 + 𝑔̃𝑖,𝑖1 ) 𝑢𝑖𝑙1
𝑗=1 2 𝑢0𝑙 = 0, 0 ≤ 𝑙 ≤ 𝐿, (34)
ℎ and by assuming the linear behavior of the solution for large
+ ( 𝑔𝑖,𝑖2 + 𝑔̃𝑖,𝑖2 ) 𝑢𝑖𝑙2 values of the spatial variable, we have 𝜕2 𝑈/𝜕𝑥2 → 0 and
2
(29) thus Δ𝑙𝑁+𝑀−1 = 0 and the null integral term approximation
𝑁−1
𝑔𝑖𝑁 𝐽𝑁+𝑀−1 = 0, for all time level 𝑙. Thus from (32) for 𝑖 =
+ ℎ ∑ 𝑔𝑖𝑗 𝑢𝑗𝑙 + 𝑙
(ℎ + 𝐴𝛿) 𝑢𝑁
2 𝑁 + 𝑀 − 1, one gets
𝑗=𝑖2 +1
𝑙+1 𝑙 0
𝑁+𝑀−1
𝑢𝑁+𝑀−1 = 𝑢𝑁+𝑀−1 = 𝑢𝑁+𝑀−1 , 0 ≤ 𝑙 ≤ 𝐿 − 1. (35)
𝛿
+ ∑ 𝑔 𝑥2 𝑢 𝑙 . For the sake of convenience to study the stability, we now
𝐴 𝑗=𝑁+1 𝑖𝑗 𝑗 𝑗
introduce the vector formulation of scheme (32)–(35). Let us
denote the vector in R𝑁+𝑀−1 as
Case 2 (𝐴𝑒−𝜀 ≤ 𝑥𝑖 < 𝐴𝑒𝜀 ). As 𝑥𝑖1 < 𝐴 and 𝑥𝑖2 ≥ 𝐴, the appro- 𝑡
𝑈𝑙 = [𝑢1𝑙 𝑢2𝑙 ⋅ ⋅ ⋅ 𝑢𝑁+𝑀−1
𝑙
], (36)
ximation of 𝐽𝑖𝑙 becomes
and let 𝑃 = (𝑝𝑖𝑗 ) be a tridiagonal matrix in R(𝑁+𝑀−1)×(𝑁+𝑀−1)
𝑖1 −1
ℎ related to the differential part, defined by
𝐽̂𝑖𝑙 = ℎ ∑ 𝑔𝑖𝑗 𝑢𝑗𝑙 + ( 𝑔𝑖,𝑖1 + 𝑔̃𝑖,𝑖1 ) 𝑢𝑖𝑙1
𝑗=1 2 𝛼𝑖 , 2 ≤ 𝑖 ≤ 𝑁 + 𝑀 − 1, 𝑗 = 𝑖 − 1,
{
{
{
{
𝑥𝑖22 𝛿 {
{
{
+( 𝑔𝑖,𝑖2 + 𝑔̃𝑖,𝑖2 ) 𝑢𝑖𝑙2 (30) {𝛽𝑖 ,
{ 1 ≤ 𝑖 ≤ 𝑁 + 𝑀 − 1, 𝑗 = 𝑖,
2𝐴 𝑝𝑖𝑗 = { (37)
{
{
{
{ 𝛾𝑖 , 1 ≤ 𝑖 ≤ 𝑁 + 𝑀 − 2, 𝑗 = 𝑖 + 1,
{
{
𝛿 𝑁+𝑀−1 {
{
+ ∑ 𝑔 𝑥2 𝑢 𝑙 .
𝐴 𝑗=𝑖2 +1 𝑖𝑗 𝑗 𝑗 {0, otherwise,
6 Abstract and Applied Analysis
where
𝑘 2 { ℎ𝑔𝑖𝑗 , 1 ≤ 𝑗 ≤ 𝑖1 − 1,
𝛾1 = 𝜎̂ , {
{
2 {
{
{
{ ℎ
{
{
𝛼𝑁+𝑀−1 = 0, { 2 𝑔𝑖,𝑗 + 𝑔̃𝑖,𝑗 ,
{
{
𝑗 = 𝑖1 ,
{
{
{
{
𝛽𝑁+𝑀−1 = 1, {0, 𝑖1 + 1 ≤ 𝑗 ≤ 𝑖2 − 1,
𝑏𝑖𝑗(2) ={ (41)
{
{
𝑘 {
{
𝛾𝑖 = 𝛼𝑖 = 2 𝜎̂2 𝑥𝑖2 , 2 ≤ 𝑖 ≤ 𝑁 − 1, { 𝑥𝑖22 𝛿
{
2ℎ {
{ 𝑔 + 𝑔̃𝑖,𝑗 , 𝑗 = 𝑖2 ,
{
{ 2𝐴 𝑖,𝑗
{
{
𝑘 2 2 {
{
𝛽𝑖 = 1 − 𝜎̂ 𝑥𝑖 , 1 ≤ 𝑖 ≤ 𝑁 − 1, {
{ 𝛿 𝑥2 𝑔 ,
ℎ2 𝑖2 + 1 ≤ 𝑗 ≤ 𝑁 + 𝑀 − 1,
{ 𝐴 𝑗 𝑖𝑗
2 2
𝜎 𝐴
𝑘̂
𝛼𝑁 = , { ℎ𝑔𝑖𝑗 , 1 ≤ 𝑗 ≤ 𝑁 − 1,
ℎ (ℎ + 𝐴𝛿/ (1 − 𝛿)) {
{
{
{
{1
{
𝜎2 𝐴 (1 − 𝛿)
𝑘̂ {
{ (ℎ + 𝐴𝛿) 𝑔𝑖𝑁, 𝑗 = 𝑁,
𝛽𝑁 = 1 − , (38) {
{ 2
{
{
𝛿ℎ {
{
{
{ 𝛿 2
{
{ 𝑥𝑗 𝑔𝑖𝑗 , 𝑁 + 1 ≤ 𝑗 ≤ 𝑖1 − 1,
𝜎2 𝐴 (1 − 𝛿)
𝑘̂ {
{ 𝐴
𝛾𝑁 = , 𝑏𝑖𝑗(3) = { 2 (42)
𝛿 (ℎ + 𝐴𝛿/ (1 − 𝛿)) {
{ 𝛿𝑥𝑗
{
{
𝑘̂𝜎2 𝑥𝑖2 { 2𝐴 𝑔𝑖,𝑗 + 𝑔̃𝑖,𝑗 ,
{
{
𝑗 = 𝑖1 , 𝑖2 ,
𝛼𝑖 = , {
{
ℎ𝑖−1 (ℎ𝑖 + ℎ𝑖−1 ) {
{
{
{ 0, 𝑖1 + 1 ≤ 𝑗 ≤ 𝑖2 − 1,
{
{
{
{
𝜎2 𝑥𝑖2
𝑘̂ {
{
𝛽𝑖 = 1 − , {𝛿 2
ℎ𝑖 ℎ𝑖−1 { 𝐴 𝑥𝑗 𝑔𝑖𝑗 , 𝑖2 + 1 ≤ 𝑗 ≤ 𝑁 + 𝑀 − 1.
𝜎2 𝑥𝑖2
𝑘̂ Hence scheme (32)–(35) can be written in the form
𝛾𝑖 = ,
ℎ𝑖 (ℎ𝑖 + ℎ𝑖−1 ) 𝑈𝑙+1 = (𝑃 + 𝐵) 𝑈𝑙 = (𝑃 + 𝐵)𝑙 𝑈0 ,
𝑁 + 1 ≤ 𝑖 ≤ 𝑁 + 𝑀 − 2. 0 ≤ 𝑙 ≤ 𝐿 − 1, (43)
𝑡
Let 𝐵 = (𝑏𝑖𝑗 ) be the matrix in ∈ R(𝑁+𝑀−1)×(𝑁+𝑀−1) related to 𝑈0 = [𝑓 (𝑥1 ) 𝑓 (𝑥2 ) ⋅ ⋅ ⋅ 𝑓 (𝑥𝑁+𝑀−1 )] .
the integral part whose entries 𝑏𝑖𝑗 for each fixed 𝑖 in 1 ≤ 𝑖 ≤
𝑁 + 𝑀 − 2 are defined by 4. Positivity and Stability of
(1)
𝑘𝑏𝑖𝑗 , 1 ≤ 𝑖 ≤ 𝑖1 (𝑁) − 1,
the Numerical Solution
{
{
{
{
{
{ The price of contracts modelled by PIDE must be a non-
{ (2)
{𝑘𝑏𝑖𝑗 ,
{ 𝑖1 (𝑁) ≤ 𝑖 ≤ 𝑖2 (𝑁) − 1, negative value. Our objective here is to demonstrate that the
𝑏𝑖𝑗 = { (39) solution of scheme (32)–(35) is conditionally nonnegative
{
{
{
{ 𝑘𝑏𝑖𝑗(3) , 𝑖2 (𝑁) ≤ 𝑖 ≤ 𝑁 + 𝑀 − 2, and stable.
{
{
{
{ First we study the positivity of the matrix 𝑃. The following
{0, 𝑖 = 𝑀 + 𝑁 − 1, lemma has been proved in [20].
The next result will be used below to guarantee stability. If the property (49) is satisfied for appropriate relation-
ships between the stepsizes ℎ, 𝛿 and 𝑘, then one says that the
Lemma 3. Let matrices 𝑃 and 𝐵 be defined by (37)–(39), and strong uniform stability is conditional.
let 𝜀 > 0; then the following results hold.
Theorem 5. With the previous notation, the numerical solu-
(1) Under conditions (C1) and (C2) of Lemma 1, ‖𝑃‖∞ = 1. tion {𝑢𝑖𝑙 } of scheme (32)–(35) is strongly uniformly ‖ ⋅ ‖∞ stable
(2) ‖𝐵‖∞ ≤ 𝑘(𝜆(𝜀) + 1), where 𝜆(𝜀) is defined by (14). if one satisfies the condition 0 < 𝛿 ≤ 1/3 together with
In order to prove the consistency, we must show that From the first mean value theorem for integrals [34, page
1063], one gets
𝑇𝑖𝑙 (𝑈) → 0, as ℎ → 0, 𝛿 → 0, 𝑘 → 0. (57)
𝑥𝑖 𝑒−𝜀
Assuming that 𝑈 is twice continuously partially differentiable 𝐼 (𝑥𝑖 , 𝜀) = ∫ 𝑔 (𝑥𝑖 , 𝜙) 𝑈 (𝜙, 𝜏𝑙 ) 𝑑𝜙
𝑥𝑖1
with respect to 𝜏 and four times partially differentiable with
respect to 𝑥, and using Taylor’s expansion about (𝑥𝑖 , 𝜏𝑙 ), it 𝑥𝑖 𝑒−𝜀
(62)
follows that = (∫ 𝑔 (𝑥𝑖 , 𝜙) 𝑑𝜙) 𝑈 (𝑐, 𝜏𝑙 )
𝑥𝑖1
𝑈𝑖𝑙+1 − 𝑈𝑖𝑙 𝜕𝑈
= (𝑥 , 𝜏𝑙 ) + 𝑘𝐸𝑖𝑙 (1) , = 𝑔̃𝑖,𝑖1 𝑈 (𝑐, 𝜏𝑙 ) , 𝑥𝑖1 < 𝑐 < 𝑥𝑖 𝑒−𝜀 ,
𝑘 𝜕𝜏 𝑖
1 𝜕2 𝑈 and since
𝐸𝑖𝑙 (1) = (𝑥 , 𝜁) , 𝜏𝑙 < 𝜁 < 𝜏𝑙+1 ,
2 𝜕𝜏2 𝑖 𝜕𝑈
𝑈 (𝑐, 𝜏𝑙 ) = 𝑈 (𝑥𝑖1 , 𝜏𝑙 ) + (𝑐 − 𝑥𝑖1 ) (𝜉, 𝜏𝑙 ) ,
𝑙 1 𝑙 𝜕𝑥 (63)
𝐸𝑖 (1) ≤ 𝑊𝑖 (1)
2 𝑥𝑖1 < 𝜉 < 𝑐,
2
1 𝜕 𝑈
= max { 2 (𝑥𝑖 , 𝜁) ; 𝜏𝑙 ≤ 𝜁 ≤ 𝜏𝑙+1 } , it follows that
2 𝜕𝜏
𝐼 (𝑥𝑖 , 𝜀) − 𝑔̃𝑖,𝑖1 𝑈𝑖𝑙 ≤ 𝑔̃𝑖,𝑖1 ℎΛ𝑙𝑖 (1) ≤ ℎ2 𝑊𝑖𝑙 (3) , (64)
1
𝜕2 𝑈
Δ𝑙𝑖 = (𝑥 , 𝜏𝑙 ) + ℎ2 𝐸𝑖𝑙 (2) , where
𝜕𝑥2 𝑖 𝜕𝑈
1 𝜕4 𝑈 ̃ 𝑙 Λ𝑙𝑖 (1) = max { (𝑥, 𝜏𝑙 ) ; 𝑥𝑖1 ≤ 𝑥 ≤ 𝑥𝑖 𝑒−𝜀 } ,
𝐸𝑖𝑙 (2) = (𝜁, 𝜏 ) , 𝑥𝑖 − ℎ < 𝜁̃ < 𝑥𝑖 + ℎ, 𝜕𝑥
(65)
12 𝜕𝑥4
𝑊𝑖𝑙 (3) = Λ𝑙𝑖 (1) max {𝑔 (𝑥𝑖 , 𝑥) ; 𝑥𝑖1 ≤ 𝑥 ≤ 𝑥𝑖 𝑒 } . −𝜀
𝑙 1
𝐸𝑖 (2) ≤ 𝑊𝑖𝑙 (2)
12 Analogously,
4
1 𝜕 𝑈 ̃ 𝑙 𝐻 (𝑥𝑖 , 𝜀) − 𝑔̃𝑖,𝑖2 𝑈𝑖𝑙 ≤ 𝑔̃𝑖,𝑖1 ℎΛ𝑙𝑖 (1) ≤ ℎ2 𝑊𝑖𝑙 (4) ,
= max { 4 (𝜁, 𝜏 ) ; 𝑥𝑖 − ℎ ≤ 𝜁̃ ≤ 𝑥𝑖 + ℎ} . 2
12 𝜕𝑥
𝑥𝑖2
(58)
𝐻 (𝑥𝑖 , 𝜀) = ∫ 𝑔 (𝑥𝑖 , 𝜙) 𝑈 (𝜙, 𝜏𝑙 ) 𝑑𝜙,
𝑥𝑖 𝑒𝜀
In accordance with [36, page 101] let us denote the local (66)
𝑙 𝜕𝑈
consistency error of 𝐽𝑖,1 (see (29)) by
𝑊𝑖𝑙 (4) = (max { (𝑥, 𝜏𝑙 ) ; 𝑥𝑖 𝑒−𝜀 ≤ 𝑥 ≤ 𝑥𝑖2 })
𝜕𝑥
𝑙 𝑙 𝑙
𝐶𝑖,1 = 𝐽𝑖,1 − 𝑇𝑖,1 ([0, 𝑥𝑖 𝑒−𝜀 ]) ,
× (max {𝑔 (𝑥𝑖 , 𝑥) ; 𝑥𝑖 𝑒−𝜀 ≤ 𝑥 ≤ 𝑥𝑖2 }) .
−𝜀
𝑇𝑖,1 ([0, 𝑥𝑖 𝑒 ])
(59) Let 𝑊𝑖𝑙 (5), 𝑊𝑖𝑙 (6), and 𝑊𝑖𝑙 (7) be defined as
𝑖1 −1
ℎ (2)
= ℎ ∑ 𝑔𝑖𝑗 𝑈𝑗𝑙 + ( 𝑔𝑖,𝑖1 + 𝑔̃𝑖,𝑖1 ) 𝑈𝑖𝑙1 . 𝑊𝑖𝑙 (5) = sup {(𝑔 (𝑥𝑖 , 𝑥) 𝑈 (𝑥, 𝜏𝑙 )) ; 0 < 𝑥 ≤ 𝑥𝑖 𝑒−𝜀 } ,
𝑗=1 2
(2)
𝑙
By (26) and (29), the local consistency error for 𝐽𝑖,2 is given 𝑊𝑖𝑙 (6) = sup {(𝑔 (𝑥𝑖 , 𝑥) 𝑈 (𝑥, 𝜏𝑙 )) ; 𝑥𝑖 𝑒𝜀 ≤ 𝑥 ≤ 𝐴} ,
by
𝐴 𝐴 1 (2)
𝑙
𝐶𝑖,2 𝑙
= 𝐽𝑖,2 𝑙
− (𝑇𝑖,2 ([𝑥𝑖 𝑒𝜀 , 𝐴]) + 𝑇𝑖,3
𝑙
([0, 1])) , (60) 𝑊𝑖𝑙 (7) = sup {(𝑔 (𝑥𝑖 , ) 𝑈 ( , 𝜏𝑙 ) 2 ) ; 0 < 𝑥 ≤ 1} ,
𝑧 𝑧 𝑧
where (67)
𝑙
𝑇𝑖,2 ([𝑥𝑖 𝑒𝜀 , 𝐴]) where the second derivatives appearing in (67) are taken with
respect to the variable 𝑥 for 𝑊𝑖𝑙 (5) and 𝑊𝑖𝑙 (6), and with respect
𝑁−1
ℎ to the variable 𝑧 for 𝑊𝑖𝑙 (7). From the expression of the error
= ( 𝑔𝑖,𝑖2 + 𝑔̃𝑖,𝑖2 ) 𝑈𝑖𝑙2 + ℎ ∑ 𝑔𝑖𝑗 𝑈𝑗𝑙
2 𝑗=𝑖 +1
of the trapezoidal rule, [35, page 54], (59)–(67), one gets
2
(61) 𝑙 𝑥 𝑒−𝜀
ℎ 𝑙 𝐶𝑖,1 ≤ ℎ2 (𝑊𝑖𝑙 (3) + 𝑖 𝑊𝑖𝑙 (5)) ,
+ 𝑔𝑖𝑁𝑈𝑁 , 12
2
𝑁+𝑀−1 𝑙 1 𝐴𝛿2 𝑙
𝑙 𝛿 1 2 𝑙 𝐶𝑖,2 ≤ ℎ2 (𝑊𝑖𝑙 (4) + (𝐴 − 𝑥 𝑒−𝜀
) 𝑊 𝑙
(6)) + 𝑊 (7) .
𝑇𝑖,3 ([0, 1]) = ( 𝑔𝑖𝑁𝑥𝑁 𝑈𝑁 + ∑ 𝑔𝑖𝑗 𝑥𝑗2 𝑈𝑗𝑙 ) . 12 𝑖 𝑖
12 𝑖
𝐴 2 𝑗=𝑁+1 (68)
Abstract and Applied Analysis 9
100 80
80 60
40
Option price V
60
Option price V
20
40
0
20
−20
0
−40
−20
0 20 40 60 80 100 0 20 40 60 80 100 120
S S
Figure 1: The variation of 𝑉 with 𝑆 for several values of 𝑌. Figure 2: The effect of positivity conditions on 𝑉.
ℎ2 2 2 𝑙
𝑇𝑖𝑙 (𝑈) = 𝑘𝐸𝑖𝑙 (1) − 𝑙
𝜎̂ 𝑥𝑖 𝐸𝑖 (2) + 𝐶𝑖,1 𝑙
+ 𝐶𝑖,2 , Example 7. Here in this example the parameters have been
2 selected as follows: 𝑇 = 1, 𝐸 = 80, 𝐴 = 3𝐸, 𝜎 = 0.2, 𝑟 = 0.01,
𝑞 = 0, 𝐶 = 1, 𝐺 = 20, 𝑀 = 30, 𝑌 = 1.5, 𝜀 = 0.1, 𝑁 = 100, and
𝑙 𝑊𝑙 (2) 𝑥 𝑒−𝜀
𝑇𝑖 (𝑈) ≤ ℎ2 ( 𝑖 + 𝑊𝑖𝑙 (3) + 𝑊𝑖𝑙 (4) + 𝑖 𝑊𝑖𝑙 (5) 𝛿 = 0.15. Positivity conditions hold for 𝑘 = 0.002, while for
24 12 𝑘 = 0.01, the positivity conditions are broken and the values
of the option price become unreliable as shown in Figure 2.
1
+ (𝐴 − 𝑥𝑖 𝑒−𝜀 ) 𝑊𝑖𝑙 (6) )
12 Next example shows that the variation of the absolute
and relative error of the solution in light of the stability and
𝐴𝛿2 𝑙 positivity conditions hold at the strike for two cases: first, for
+ 𝑊 (7) + 𝑘𝑊𝑖𝑙 (1) .
12 𝑖 several values of the stepsize discretization ℎ and second, for
(69) different values of the parameter 𝜀.
𝑌 = 0.5 𝑌 = 1.5 6
ℎ Absolute Relative Absolute Relative
𝛼 𝛼 4
error error error error
𝑌 = 1.98
ℎ −6
Absolute error Relative error 𝛼 0 10 20 30 40 50 60
0.8 3.87 × 10−5 3.87 × 10−7 — S
0.4 9.76 × 10−6 9.76 × 10−8 1.9873 h = 0.25
0.2 2.46 × 10−6 2.46 × 10−8 1.9882 h = 0.5
h = 0.75
𝑌 = 1.98
𝜀
Absolute error Relative error
0.8 4.19 × 10−4 4.19 × 10−6
0.4 5.76 × 10−5 5.76 × 10−7
0.2 5.92 × 10−6 5.92 × 10−8
−1
0 10 20 30 40 50 60
S
[3] R. C. Merton, “Option pricing when underlying stock returns SIAM Journal on Scientific Computing, vol. 29, no. 1, pp. 93–117,
are discontinuous,” Journal of Financial Economics, vol. 3, no. 2007.
1-2, pp. 125–144, 1976. [23] E. W. Sachs and A. K. Strauss, “Efficient solution of a partial
[4] O. Barndor-Nielsen, “Processes of normal inverse Gaussian integro-differential equation in finance,” Applied Numerical
type,” Finance and Stochastics, vol. 2, no. 1, pp. 41–68, 1997. Mathematics, vol. 58, no. 11, pp. 1687–1703, 2008.
[5] E. Eberlein, “Applications of generalized hyperbolic Lévy [24] D. Tavella and C. Randall, Pricing Financial Instruments, John
motion to Fi-nance,” in Lévy Processes—Theory and Applica- Wiley & Sons, New York, NY, USA, 2000.
tions, O. Barndorff-Nielsen, T. Mikosch, and S. Resnick, Eds., [25] S. Salmi and J. Toivanen, “An iterative method for pricing Amer-
pp. 319–336, Birkhauser, Boston, Mass, USA, 2001. ican options under jump-diffusion models,” Applied Numerical
[6] I. Koponen, “Analytic approach to the problem of convergence Mathematics, vol. 61, no. 7, pp. 821–831, 2011.
of truncated Lévy flights towards the Gaussian stochastic [26] J. Toivanen, “Numerical valuation of European and American
process,” Physical Review E, vol. 52, no. 1, pp. 1197–1199, 1995. options under Kou’s jump-diffusion model,” SIAM Journal on
[7] D. Madan and F. Milne, “Option pricing with variance gamma Scientific Computing, vol. 30, no. 4, pp. 1949–1970, 2008.
martingale components,” Mathematical Finance, vol. 1, no. 4, pp. [27] A. Almendral and C. W. Oosterlee, “Numerical valuation of
39–55, 1991. options with jumps in the underlying,” Applied Numerical
[8] P. Carr, H. Geman, D. B. Madan, and M. Vor, “The fine structure Mathematics, vol. 53, no. 1, pp. 1–18, 2005.
of asset returns: an empirical investigation,” Journal of Business, [28] J. Lee and Y. Lee, “Tridiagonal implicit method to evaluate
vol. 75, no. 2, pp. 305–332, 2002. European and American options under infinite activity Lévy
[9] S. I. Boyarchenko and S. Z. Levendorskii, “Option pricing for models,” Journal of Computational and Applied Mathematics,
truncated Lévy processes,” International Journal of Theoretical vol. 237, pp. 234–243, 2013.
and Applied Finance, vol. 3, no. 3, pp. 549–552, 2000. [29] D. B. Madan and E. Seneta, “The variance gamma (V.G.) model
[10] R. Cont and P. Tankov, Financial Modelling With Jump Processes, for share market returns,” The Journal of Business , vol. 63, pp.
Chapman & Hall/CRC Financial Mathematics, Chapman & 511–524, 1990.
Hall/CRC, Boca Raton, Fla, USA, 2004. [30] M. S. Milgram, “The generalized integro-exponential function,”
[11] A. Pascucci, PDE and Martingale Methods in Option Pricing, vol. Mathematics of Computation, vol. 44, no. 170, pp. 443–458, 1985.
2 of Bocconi & Springer, Springer, Milan, Italy, 2011.
[31] N. M. Temme, Special Functions an Introduction to the Classical
[12] A.-M. Matache, P.-A. Nitsche, and C. Schwab, “Wavelet Functions of Mathematical Physics, John Wiley & Sons, New
Galerkin pricing of American options on Lévy driven assets,” York, NY, USA, 1996.
Quantitative Finance, vol. 5, no. 4, pp. 403–424, 2005.
[32] A.-M. Matache, T. Von Petersdorff, and C. Schwab, “Fast deter-
[13] D. B. Madan and M. Yor, “CGMY and Meixner subordinators ministic pricing of options on Lévy driven assets,” Mathematical
are absolutely continuous with respect to one sided stable Modelling and Numerical Analysis, vol. 38, no. 1, pp. 37–71, 2004.
subordinators,” 2006, http://arxiv.org/pdf/math/0601173.pdf.
[33] R. Kangro and R. Nicolaides, “Far field boundary conditions for
[14] J. Poirot and P. Tankov, “Monte Carlo option pricing for tem- Black-Scholes equations,” SIAM Journal on Numerical Analysis,
pered stable (CGMY) processes,” Asia-Pacific Financial Mar- vol. 38, no. 4, pp. 1357–1368, 2001.
kets, vol. 13, no. 4, pp. 327–344, 2006.
[34] I. S. Grandshteyn, I. M. Ryzhik, and A. Je, Eds.I. S. Grandshteyn,
[15] F. Fang and C. W. Oosterlee, “A novel pricing method for I. M. Ryzhik, and A. Jefferey, Eds., Table of Integrals, Series and
european options based on fourier-cosine series expansions,” Product, Academic Press, New York, NY, USA, 7th edition, 2007.
SIAM Journal on Scientific Computing, vol. 31, no. 2, pp. 826–
[35] P. J. Davis and P. Rabinowitz, Methods of Numerical Integration,
848, 2008.
Academic Press, New York, NY, USA, 2nd, edition, 1984.
[16] S. Pagliarani, A. Pascucci, and C. Riga, Adjoint Expansions in
Local Lévy Models, SSRN eLibrary, 2011. [36] P. Linz, Analytic and Numerical Methods for Volterra Equations,
SIAM, Philadelphia, Pa, USA, 1985.
[17] E. Benhamou, E. Gobet, and M. Miri, “Smart expansion and fast
calibration for jump diffusions,” Finance and Stochastics, vol. 13, [37] G. D. Smith, Numerical Solution of Partial Differential Equa-
no. 4, pp. 563–589, 2009. tions: Finite Difference Methods, Clarendon Press, Oxford, UK,
3rd edition, 1985.
[18] R. Cont and E. Voltchkova, “A finite difference scheme for
option pricing in jump diffusion and exponential Lévy models,” [38] D. B. Madan, P. Carr, and E. C. Chang, “The variance gamma
SIAM Journal on Numerical Analysis, vol. 43, no. 4, pp. 1596– process and option pricing,” European Finance Review, vol. 2,
1626, 2005. pp. 79–105, 1998.
[19] I. R. Wang, J. W. Wan, and P. A. Forsyth, “Robust numerical
valuation of European and American options under the CGMY
process,” Journal of Computational Finance, vol. 10, no. 4, pp.
31–69, 2007.
[20] M. C. Casaban, R. Company, L. Jodar, and J. V. Romero, “Double
discretization difference schemes for partial integro-differential
option pricing jump diffusion models,” Abstract and Applied
Analysis, vol. 2012, Article ID 120358, 20 pages, 2012.
[21] L. Andersen and J. Andreasen, “Jump-diffusion processes: vola-
tility smile fitting and numerical methods for option pricing,”
Review of Derivatives Research, vol. 4, no. 3, pp. 231–262, 2000.
[22] A. Almendral and C. W. Oosterlee, “Accurate evaluation of
european and American options under the CGMY process,”
Advances in Advances in Journal of Journal of
Operations Research
Hindawi Publishing Corporation
Decision Sciences
Hindawi Publishing Corporation
Applied Mathematics
Hindawi Publishing Corporation
Algebra
Hindawi Publishing Corporation
Probability and Statistics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014
International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences