Good Notes On FDM For Option Pricing PDF

Chapter 5 Finite Difference
Methods
Math6911 S08, HM Zhu

References
1. Chapters 5 and 9, Brandimarte
2. Section 17.8, Hull
3. Chapter 7, “Numerical analysis”, Burden and Faires
2
Math6911, S08, HM ZHU
Outline
• Finite difference (FD) approximation to the derivatives

• Explicit FD method
• Numerical issues
• Implicit FD method
• Crank-Nicolson method
• Dealing with American options
• Further comments
3
Methods
5.1 Finite difference approximations

Finite-difference mesh
• Aim to approximate the values of the continuous function

f(t, S) on a set of discrete points in (t, S) plane
• Divide the S-axis into equally spaced nodes at distance
∆S apart, and, the t-axis into equally spaced nodes a
distance ∆t apart
• (t, S) plane becomes a mesh with mesh points on
(i ∆t, j∆S)
• We are interested in the values of f(t, S) at mesh points
(i ∆t, j∆S), denoted as
fi , j = f ( i∆t, j ∆S ) 5
The mesh for finite-difference
approximation
fi , j = f ( i∆t, j ∆S )
S
Smax=M∆S
?
j ∆S
i ∆t T=N ∆t t
6
Black-Scholes Equation for a
European option with value V(S,t)
∂V 1 2 2 ∂ 2V ∂V
+ σ S + rS − rV = 0 (5.1)
∂t 2 ∂ S
2
∂S
where 0 < S < +∞ and 0 ≤ t < T
with proper final and boundary conditions
Notes:
This is a second-order hyperbolic, elliptic, or parabolic,
forward or backward partial differential equation
Its solution is sufficiently well behaved ,i.e. well-posed

Finite difference approximations
The basic idea of FDM is to replace the partial derivatives

by approximations obtained by Taylor expansions near the
point of interests
For example,
∂f ( S,t ) f ( S ,t + ∆t ) − f ( S ,t ) f ( S ,t + ∆t ) − f ( S ,t )
= lim ≈
∂t ∆t → 0 ∆t ∆t
for small ∆t, using Taylor expansion at point ( S ,t )
∂f ( S ,t )
f ( S ,t + ∆t ) = f ( S ,t ) +
∂t
(
∆t + O ( ∆ t )
2
)
8
Forward-, Backward-, and Central-
difference approximation to 1st order
derivatives
central
backward forward
t − ∆t t t + ∆t
∂f ( t,S ) f ( t + ∆t,S ) − f ( t,S )
Forward: ≈ + O ( ∆t )
∂t ∆t
∂f ( t,S ) f ( t,S ) − f ( t − ∆t,S )
Backward: ≈ + O ( ∆t )
∂t ∆t
∂f ( t,S ) f ( t + ∆t,S ) − f ( t − ∆t,S )
Central:
∂t
≈
2∆t
(
+ O ( ∆t )
2
)
Symmetric Central-difference
approximations to 2nd order derivatives
∂ 2 f ( t,S ) f ( t,S + ∆S ) − 2 f ( t,S ) + f ( t,S − ∆S )

∂S 2
≈
( ∆S )
2 (
+ O ( ∆S )
2
)
Use Taylor's expansions for f ( t,S + ∆S ) and f ( t,S − ∆S )
around point ( t,S ) :
f ( t,S + ∆S ) = ?
+
f ( t,S − ∆S ) = ?
10
∂f fi +1, j − fi , j ∂f fi , j +1 − fi , j
Forward Difference: ≈ , ≈
∂t ∆t ∂S ∆S
∂f fi , j − fi −1, j ∂f fi , j − fi , j −1
Backward Difference: ≈ , ≈
∂t ∆t ∂S ∆S
∂f fi +1, j − fi −1, j ∂f fi , j +1 − fi , j −1
Central Difference: ≈ , ≈
∂t 2∆t ∂S 2∆S
As to the second derivative, we have:
∂ 2 f ⎛ fi , j +1 − fi , j fi , j − fi , j −1 ⎞
≈⎜ − ⎟ ∆S
∂S 2
⎝ ∆S ∆S ⎠
fi , j +1 − 2 fi , j + fi , j −1
=
( )
2
∆ S 11
• Depending on which combination of schemes we use in

discretizing the equation, we will have explicit, implicit, or
Crank-Nicolson methods
• We also need to discretize the boundary and final
conditions accordingly. For example, for European Call,
Final Condition:
f N , j = max ( j ∆S − K , 0 ) , for j = 0 ,1,...,M
Boundary Conditions:
⎧⎪ fi ,0 = 0
⎨ − r ( N − i ) ∆t
, for i = 0 ,1,...,N
⎪⎩ fi ,M = S max − Ke
12
Math6911, S08, HM ZHU where Smax = M ∆S.
Methods
5.2.1 Explicit Finite-Difference Method

Explicit Finite Difference Methods
∂ f ∂f 1 2 2∂2f
In + rS + σ S = rf , at point (i ∆ t , j ∆ S ), set
∂t ∂S 2 ∂S 2
∂ f f i , j − f i −1 , j
backward difference: ≈
∂t ∆t
∂f f i , j +1 − f i , j −1
central difference: ≈ ,
∂S 2∆S
and
∂2f f i , j + 1 + f i , j −1 − 2 f i , j
≈ , r f = rf i , j , S = j ∆ S
∂S 2
∆S 2

Rewriting the equation, we get an explicit schem e:

f i −1 , j = a *j f i , j −1 + b*j f i , j + c*j f i , j +1 (5.2)
where
1
(
a = ∆ t σ 2 j 2 − rj
*
j
2
)
(
b j = 1 − ∆t σ j + r
* 2 2
)
∆ t (σ + rj )
1
c = *
j
2
j 2
2
for i = N - 1, N - 2 , ..., 1, 0 and j = 1, 2 , ..., M - 1 .
Numerical Computation Dependency
S
Smax=M∆S
(j+1)∆S x
j∆S x x
(j-1)∆S x
0
0 (i-1)∆t i∆t T=N ∆t t
Implementation
1. Starting with the final values f N , j , we apply (5.2) to solve

f N −1, j for 1 ≤ j ≤ M − 1. We use the boundary condition
to determine f N −1,0 and f N -1,M .
2. Repeat the process to determine f N − 2 , j and so on
17
Example
We compare explicit finite difference solution for a

European put with the exact Black-Scholes formula, where
T = 5/12 yr, S0=$50, K = $50, σ=30%, r = 10%.
Black-Scholes Price: $2.8446

EFD Method with Smax=$100, ∆S=2, ∆t=5/1200: $2.8288
18
Example (Stability)
We compare explicit finite difference solution for a European

put with the exact Black-Scholes formula, where T = 5/12 yr,
S0=$50, K = $50, σ=30%, r = 10%.

EFD Method with Smax=$100, ∆S=1.5, ∆t=5/1200: $3.1414
EFD Method with Smax=$100, ∆S=1, ∆t=5/1200: -$2.8271E22
19
Methods
5.2.2 Numerical Stability

Numerical Accuracy
• The problem itself

• The discretization scheme used
• The numerical algorithm used
21
Conditioning Issue
Suppose we have mathematically posed problem:

y = f ( x)
where y is to evaluated given an input x.
Let x* = x + δ x for small change δ x.
( )
If hen f x* is near f ( x ) , then we call the problem is
well - conditioned . Otherwise, it is ill-posed/ill-conditioned.
22
Conditioning Issue
• Conditioning issue is related to the problem itself, not to the

specific numerical algorithm; Stability issue is related to the
numerical algorithm
• One can not expect a good numerical algorithm to solve an ill-
conditioned problem any more accurately than the data
warrant
• But a bad numerical algorithm can produce poor solutions
even to well-conditioned problems
23
Conditional Issue
The concept "near" can be measured by further information about

the particular problem:
( )
f ( x ) − f x* δx
f ( x)
≤C
x
( f ( x ) ≠ 0)
where C is called condition number of this problem. If C is large,
the problem is ill-conditioned.
24
Floating Point Number & Error
Let x be any real number.

Infinite decimal expansion : x = ± .x1 x2 xd 10 e
Truncated floating point number : x ≈ fl ( x ) = ± .x1 x2 xd 10 e
where x1 ≠ 0, 0 ≤ xi ≤ 9,
d : an integer, precision of the floating point system
e : an bounded integer
Floating point or roundoff error : fl ( x ) − x
25
Error Propagation
When additional calculations are done, there is an

accumulation of these floating point errors.
Example : Let x = − 0.6667 and fl ( x ) = −0.667 10 0
where d = 3.
Floating point error : fl ( x ) − x = −0.0003
Error propagatio n : fl ( x ) − x = 0.00040011
2 2
26
Numerical Stability or Instability
Stability ensures if the error between the numerical

soltuion and the exact solution remains bounded
as numerical computation progresses.
( )
That is, f x* (the solution of a slightly perturbed problem)
is near f * ( x )( the computed solution) .
Stability concerns about the behavior of fi , j − f ( i∆t, j∆S )

as numerical computation progresses for fixed discretization
steps ∆t and ∆S. 27
Convergence issue
Convergence of the numerical algorithm concerns about

the behavior of fi , j − f ( i∆t, j∆S ) as ∆t, ∆S → 0 for fixed
values ( i∆t, j∆S ) .
For well-posed linear initial value problem,

Stability ⇔ Convergence
(Lax's equivalence theorem, Richtmyer and Morton, "Difference
Methods for Initial Value Problems" (2nd) 1967)
28
Numerical Accuracy
• These factors contribute the accuracy of a numerical solution.

We can find a “good” estimate if our problem is well-
conditioned and the algorithm is stable
( )
Stable: f * ( x ) ≈ f x* ⎫
⎪
⎬ f ( x) ≈ f ( x)
*
Well-conditioned: f ( x ) ≈ f ( x ) ⎪⎭
*
29
Methods
5.2.3 Financial Interpretation of Numerical Instability

Financial Interpretation of instability
(Hall, page 423-4)
If ∂ f ∂ S and ∂ 2 f ∂ S 2 are assum ed to be the sam e at ( i + 1, j )

as they are at ( i, j ), we obtain equations of the form :
f = aˆ f
i,j j + bˆ f
i +1 , j −1+ cˆ f
j i +1 , j (5.3)
j i +1 , j +1
where
1 ⎛1 1 ⎞ 1
â j = ⎜ ∆ tσ j − ∆ t r j ⎟ = πd
2 2
1 + r ∆t ⎝ 2 2 ⎠ 1 + r ∆t
b̂ j =
1
1 + r ∆t
( 1 − σ j ∆t =
2 2
)
1
1 + r ∆t
π0
1 ⎛1 1 ⎞ 1
ĉ j = ⎜ ∆ tσ j + ∆ t r j ⎟ = πu
2 2
1 + r ∆t ⎝ 2 2 ⎠ 1 + r ∆t
for i = N - 1, N - 2 , ..., 1, 0 and j = 1, 2 , ..., M - 1 .
ƒi +1, j +1
πu
π0
ƒi , j ƒi +1, j
πd
ƒi +1, j –1
These coefficients can be interpreted as probabilities times
a discount factor. If one of these probability < 0, instability
occurs.
Explicit Finite Difference Method as
Trinomial Tree
Check if the mean and variance of the

Expected value of the increase in asset price during ∆t:
E [ ∆ ] = −∆S π d + 0π 0 + ∆S π u = rj∆S ∆t = rS ∆t
Variance of the increment:
E ⎡⎣ ∆ 2 ⎤⎦ = − ( ∆S ) π d + 0π 0 + ( ∆S ) π u = σ 2 j 2 ( ∆S ) 2 ∆t = σ 2 S 2 ∆t
2 2
Var [ ∆ ] = E ⎡⎣ ∆ 2 ⎤⎦ − E 2 [ ∆ ] = σ 2 S 2 ∆t − r 2 S 2 ( ∆t ) ≈ σ 2 S 2 ∆t
2
which is coherent with geometric Brownian motion in a risk-neutral world

Change of Variable
Define Z = ln S. The B-S equation becomes

∂f ⎛ σ 2 ⎞ ∂f σ 2 ∂ 2 f
+⎜r − ⎟ + = rf
∂t ⎝ 2 ⎠ ∂Z 2 ∂Z 2
The corresponding difference equation is

fi +1, j − fi , j ⎛ σ 2 ⎞ fi +1, j +1 − f i +1, j −1 σ 2 f i +1, j +1 − 2 fi +1, j −1 + fi +1, j +1
+⎜r − ⎟ + = rfi , j
∆t ⎝ 2 ⎠ 2 ∆Z 2 ∆Z 2
or
fi , j = α *j f i +1, j −1 + β *j fi +1, j + γ *j f i +1, j +1 (5.4)
34
Change of Variable
where
1 ⎡ ⎛ σ 2
⎞ ∆t σ 2
∆t ⎤
αj =
*
⎢− ⎜ r − ⎟ + ⎥
1 + r ∆t ⎣ ⎝ 2 ⎠ 2∆Z 2 ∆Z 2 ⎦
1 ⎛ ∆t 2 ⎞
β *j = ⎜ 1 − σ ⎟
1 + r ∆t ⎝ ∆Z 2
⎠
1 ⎡ ⎛ σ 2
⎞ ∆ t σ 2
∆t ⎤
γj =
*
⎢⎜ r − ⎟ + 2⎥
1 + r ∆t ⎣⎝ 2 ⎠ 2 ∆ Z 2 ∆Z ⎦
It can be shown that it is numerically most efficient if ∆Z = σ 3∆t .
35
Reduced to Heat Equation
Get rid of the varying coefficients S and S² by using

change of variables:
− ( k −1) x − ( k +1)2 τ
1 1
S = Ee x , t = T − 2τ σ 2 , V (S , t ) = E e 2 4
u ( x, τ )
⎛1 2⎞
k =r ⎜ σ ⎟
⎝2 ⎠
Equation (5.1) becomes heat equation (5.5):
∂ u ∂ 2u
= (5.5)
∂ τ ∂ x2
for − ∞ < x < +∞ and τ > 0 36
Explicit Finite Difference Method
With u nm = u (n δx,mδ τ ), this involves solving a system of

finite difference equations of the form :
u nm +1 − u nm
δτ
+ O (δτ ) =
u nm+1 − 2u nm + u nm−1
(δx ) 2
(
+ O (δx )
2
)
( )
Ignoring terms of O (δτ ) and O (δx ) , we can approximat e
2
this by
u nm +1 = α u nm+1 + (1 − 2α )u nm + α u nm−1
⎛ 1 2 ⎞
δτ ⎜ σ T ⎟
where α = , for -N −
≤ n ≤ N +
and m = 0 ,
1 ,...M ⎜ = 2 ⎟
(δx )2 ⎜
⎜
δτ ⎟
⎟
Math6911, S08, HM ZHU ⎝ ⎠
Stability and Convergence
(P. Wilmott, et al, Option Pricing)
Stability:
The solution of Eqn (5.5) is
δτ 1 1
i) Stable if 0 < α = ≤ ; ii) Unstable if α >
(δ x ) 2
2
2
Convergence:
1
If 0 < α ≤ , then the explicit finite-difference approximation
2
converges to the exact solution as δτ ,δ x → 0
(in the sense that unm → u ( nδ x, mδτ ) as δτ ,δ x → 0)
Rate of Convergence is O (δτ ) 38
Methods
5.3.1 Implicit Finite-Difference Method

Implicit Finite Difference Methods
∂ f ∂f 1 2 2∂2f
In + rS + σ S = r f , we use
∂t ∂S 2 ∂S 2
∂ f f i +1,j − f i ,j
forw ard difference: ≈
∂t ∆t
∂ f f i , j + 1 − f i , j −1
central difference: ≈ ,
∂ S 2∆S
and
∂ 2f f i , j + 1 + f i , j −1 − 2 f i , j
≈ , rf = rf i , j
∂ S 2
∆S 2

Rewriting the equation, we get an implicit scheme:

a j f i,j −1 + b j f i,j + c j f i,j +1 = f i +1, j (5.6)
where
1
(
a j = ∆ t −σ 2 j 2 + rj
2
)
(
b j = 1 + ∆t σ j + r
2 2
)
= − ∆ t (σ + rj )
1
cj 2
j 2
2
for i = N - 1, N - 2 , ..., 1, 0 and j = 1, 2 , ..., M - 1 .
S
Smax=M∆S
(j+1)∆S x
j∆S x x
(j-1)∆S x
0
0 (i-1)∆t i∆t (i+1)∆t T=N ∆t t
Implementation
Equation (5.6) can be rewritten in matrix form:

Cfi = fi +1 + bi ( 5.7 )
where fi and bi are (M − 1) dimensional vectors
T T
fi = ⎡⎣ fi ,1 , fi ,2 , fi ,3 , fi ,M −1 ⎤⎦ ,bi = ⎡⎣ − a1 fi ,0 ,0 , 0 , , 0 , −cM −1 fi ,M ⎤⎦
and C is (M − 1) × (M − 1) symmetric matrices
⎡ b1 c1 0 0 ⎤
⎢a b2 c2 0 ⎥⎥
⎢ 2
C=⎢0 a3 b3 ⎥
⎢ ⎥
⎢ cM − 2 ⎥
⎢⎣ 0 0 aM −1 bM −1 ⎥⎦
Implementation
1. Starting with the final values f N , j , we need to solve a

linear system (5.7) to obtain f N −1, j for 1 ≤ j ≤ M − 1
using LU factorization or iterative methods. We use the
boundary condition to determine f N −1,0 and f N -1,M .
2. Repeat the process to determine f N − 2 , j and so on
44
Example
We compare implicit finite difference solution for a

European put with the exact Black-Scholes formula, where
T = 5/12 yr, S0=$50, K = $50, σ=30%, r = 10%.

IFD Method with Smax=$100, ∆S=2, ∆t=5/1200: $2.8194
45
Example (Stability)
We compare implicit finite difference solution for a European

put with the exact Black-Scholes formula, where T = 5/12 yr,
S0=$50, K = $50, σ=30%, r = 10%.

IFD Method with Smax=$100, ∆S=1.5, ∆t=5/1200: $3.1325
46
Implicit vs Explicit Finite Difference Methods
ƒi +1, j +1 ƒi , j +1
ƒi , j ƒi +1, j
ƒi , j ƒi +1, j
ƒi , j –1
ƒi +1, j –1
Implicit Method
Explicit Method
(always stable)
Implicit vs Explicit Finite Difference
Method
• The explicit finite difference method is

equivalent to the trinomial tree approach:
– Truncation error: O(∆t)
– Stability: not always
• The implicit finite difference method is
equivalent to a multinomial tree approach:
– Truncation error: O(∆t)
– Stability: always
48
Other Points on Finite Difference
Methods
• It is better to have ln S rather than S as

the underlying variable in general
• Improvements over the basic implicit and
explicit methods:
– Crank-Nicolson method, average of
explicit and implicit FD methods, trying
to achieve
• Truncation error: O((∆t)2 )
• Stability: always

Methods
5.3.2 Solving a linear system using direct methods

Solve Ax=b
A x=b
Various shapes of matrix A
Lower triangular Upper triangular General
51
5.3.2.A Triangular Solvers

Example: 3 x 3 upper triangular system
⎡4 6 1 ⎤⎡ x1 ⎤ ⎡100⎤
⎢0 1 1 ⎥⎢ x ⎥= ⎢ 10 ⎥
⎢ ⎥⎢ 2 ⎥ ⎢ ⎥
⎢⎣0 0 4⎥⎦⎢⎣ x3 ⎥⎦ ⎢⎣ 20 ⎥⎦
⇒ x3 = 20 / 4 = 5
⇒ x2 = 10 − x3 = 5
⇒ 4 x1 = 100 − x3 − 6 * x2 = 65
∴ x1 = 65 4
53
Solve an upper triangular system Ax=b
⎛ x1 ⎞
⎜ ⎟
⎜ x2 ⎟
(0 0 aii ain )⎜ ⎟ = aii xi + ∑ aij x j = bi ,i = 1, … , n
⎜ ⎟ j >i
⎜x ⎟
⎝ n⎠
⇒ xn = bn ann
⎛ ⎞
xi = ⎜⎜ bi − ∑ aij x j ⎟⎟ aii ,i = 1, n −1
⎝ j >i ⎠
54
Implementation
Function x = UpperTriSolve(A, b)
n = length(b);
x(n) = b(n)/A(n,n);
for i = n-1:-1:1
sum = b(i);
for j = i+1:n
sum = sum - A(i,j)*x(j);
end
x(i) = sum/A(i,i);
end
55
5.3.2.B Gauss Elimination

To solve Ax=b for general form A
= *
To solve Ax = b : Solve two triangular systems :

Suppose A = LU. Then 1) solve z from Lz = b
Ax = LUx = b 2) solve x from Ux = z
⇓
z
57
Gauss Elimination = *
Goal: Make A an upper triangular matrix through using

fundamental row operations
For example, to zero the elements in the lower triangular part of A

1) Zero a21
⎡ 1 0 0⎤⎡ a11 a12 a13 ⎤ ⎡ a11 a12 a13 ⎤
⎢ a ⎥⎢ ⎥ ⎢ a21 a12 a21 a13 ⎥
⎢− 21 1 0⎥⎢a21 a22 a23 ⎥ = ⎢ 0 a22 − a23 − ⎥
⎢ a11 ⎥⎢ a a32 a33 ⎥⎦ ⎢⎢a
a11 a11 ⎥
⎢⎣ 0 0 1⎥⎦⎣ 31 ⎣ 31 a32 a33 ⎥⎦
⇓
58
Math6911, S08, HM ZHU E 21 A
2) Zero a31
⎡ ⎤
⎡ ⎤
⎢ 1 ⎥ ⎡ a11 a12 a13 ⎤ ⎢ a11 a a
⎥
⎥ ⎢ ⎥
12 13
0 0 ⎢
⎢ ⎥ a21 a12 a21 a13 ⎥ ⎢ a21 a12 a21 a13 ⎥
⎢ 0 1 0⎥ ⎢ 0 a22 − a23 − = ⎢ 0 a22 − a23 − ⎥
⎢ ⎥⎢ a11 a11 ⎥ ⎢ a11 a11 ⎥
⎢ − a31 ⎢ ⎥
0 1 ⎥ ⎢⎣ a31 a32 a33 ⎥⎦ ⎢ a31 a12 a31 a13 ⎥
⎢⎣ a11 ⎥⎦ ⎢ 0 a32 − a33 − ⎥
⎣ a11 a11 ⎦
⇓ ⎡ a11 a12 a13 ⎤
E 31 E 21 A = ⎢⎢ 0 a22 a23 ⎥⎥
⎢⎣ 0 a32 a33 ⎥⎦
59
3) Zero a32
⎡ ⎤ ⎡ ⎤
⎢1 0 0 ⎥ ⎡ a11 a12 a13 ⎤ ⎢ a11 a12 a13 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 1 0⎥ ⎢ 0 a22 a23 ⎥ = ⎢ 0 a22 a23 ⎥ ≡U
⎢ ⎥⎢ ⎢ ⎥
⎢0 − a32 ⎥ ⎣ 0 a32 a33 ⎥⎦ ⎢ a32 a23 ⎥
1 0 0 a33 −
⎢⎣ a22 ⎥⎦ ⎢⎣ a22 ⎥⎦
⇓
E 32 E 31 E 21 A = U
⇓
lower triangular 60
⎡ ⎤
⎢ ⎥
Claim 1: ⎢ 1 0 0⎥
⎢ a 21 ⎥
E 32 E 31 E 21 = ⎢− 1 0⎥
⎢ a1 1 ⎥
⎢ a 31 a 32 ⎥
⎢− − 1⎥
⎣ a1 1 a 22 ⎦
Claim 2: ⎡ ⎤
⎢ ⎥
⎢ 1 0 0⎥
⎢ a 21 ⎥
( E 32 E 31 E 21 )
−1
=⎢ 1 0⎥
⎢ a11 ⎥
⎢ a 31 a 32 ⎥
⎢ 1⎥ 61
⎣ a11 a 22 ⎦
LU Factorization = *
Therefore, through Gauss Elimination, we have
E 32 E 31 E 21A = U
A = (E 32 E 31 E 21 ) U
−1
A = LU
It is called LU factorization. When A is symmetric,
A = LLT which is called Cholesky Decomposition

62
To solve Ax=b for general form A
To solve Ax = b becomes
1) Use Gauss elimination to make A upper triangular, i.e.
L−1Ax = L−1b ⇒ Ux = z
2) Solve x from Ux = z
This suggests that when doing Gauss elimination, we can do it

to the augmented matrix [A b] associated with the linear system.
63
An exercise
% build the matrix A

A = [2, -1, 0; -1, 2, -1; 0, -1, 2]
% build the vector b

⎡ 2 −1 ⎤⎡ x1 ⎤ ⎡0 ⎤ x_true = [1:3]';
⎢− 1 2 − 1⎥⎢ x ⎥ = ⎢0 ⎥ b = A * x_true;
⎢ ⎥⎢ 2 ⎥ ⎢ ⎥ % lu decomposition of A
⎢⎣ − 1 2 ⎥⎦⎢⎣ x3 ⎥⎦ ⎢⎣4⎥⎦ [l, u] = lu(A)
% solve z from lz = b where z = ux

z = l\b;
% solve x from ux = z
x = u\z
64
General Gauss Elimination
row i
a ji
row j - ∗ row i
aii
65
What if we have zero or very small
diagonal elements?
Somtimes, we need to permute rows of A so that Gauss eliminatio n

can be computed or computed stably, i.e., PA = LU. This is called
partial pivoting.
For example
⎡0 1⎤
A=⎢
⎣2 3 ⎥⎦
⎡ 0 1 ⎤ ⎡ 0 1 ⎤ ⎡ 2 3⎤
PA = ⎢ ⎥ ⎢ ⎥ =⎢ ⎥
⎣ 1 0 ⎦⎣ 2 3 ⎦ ⎣ 0 1 ⎦
⎡ 0.0001 1⎤ ⎡ 1 0 ⎤ ⎡ 0.0001 1 ⎤
A=⎢ ⎥ =⎢ ⎥ ⎢ ⎥ = LU
⎣ 1 1⎦ ⎣10,000 1 ⎦ ⎣ 0 - 9999 ⎦
⎡ 0 1 ⎤ ⎡ 0.0001 1⎤ ⎡ 1 0 ⎤ ⎡1 1 ⎤
PA = ⎢ ⎥ ⎢ ⎥ =⎢ ⎥ ⎢ ⎥ = LU 66
Math6911, S08, HM⎣ 1
ZHU 0 ⎦ ⎣ 1 1⎦ ⎣ 0.0001 1 ⎦ ⎣ 0 .9999 ⎦
Can we always have A = LU?
No!
If det ( A (1:k,1:k ) ) ≠ 0 for k =1,...,n-1,
then A ∈ Rnxn has an LU factorization.
Proof:
See Golub and Loan, Matrix Computation, 3rd edition
67
Methods
5.4.1 Crank-Nicolson Method

∂ f ∂f 1 2 2∂2f
W ith + rS + σ S = rf ,
∂t ∂S 2 ∂S 2
we can obtain an explicite form :

f i +1 , j − f i , j
+ O ( ∆t ) =
∆t
f i +1 , j +1 − f i +1 , j −1 1 2 2 2 f i + 1 , j + 1 + f i + 1 , j −1 − 2 f i + 1 , j
− rj ∆ S − σ j ( ∆S )
2∆S 2 ∆ S2
+ rf i +1 , j + O (( ∆S ) ) 2

∂ f ∂f 1 2 2∂2f
W ith + rS + σ S = r f,
∂t ∂S 2 ∂S 2
we can obtain an im plicit form :

f i +1 , j − f i , j
+ O ( ∆t ) =
∆t
f i , j +1 − f i , j −1 1 2 2 2 f i , j + 1 + f i , j −1 − 2 f i , j
− rj ∆ S − σ j ( ∆S )
2∆S 2 ∆ S2
+ rf i , j + O (( ∆S ) )2

Crank-Nicolson Methods: average of explicit
and implicit methods
(Brandmarte, p485)
fi+1, j − fi, j
∆t
(
+ O ( ∆t ) =
2
)
rj∆S ⎛ fi+1, j +1 − fi+1, j −1 fi, j +1 − fi, j −1 ⎞
− ⎜ + ⎟
2 ⎝ 2∆S 2∆S ⎠
1 2 2 2 ⎛ fi +1, j +1 + fi +1, j −1 − 2 fi +1, j fi, j +1 + fi, j −1 − 2 fi, j ⎞
− σ j ( ∆S ) ⎜ + ⎟
4 ⎝ ∆S 2
∆S 2
⎠
r
+ ( fi+1, j + fi, j ) + O ( ∆S )
2
(2
)
Crank-Nicolson Methods
R e w ritin g th e e q u a tio n , w e g e t a n C ra n k -N ic o ls o n s c h e m e :
(
− α j f i , j −1 + 1 − β )fj i,j − γ j f i , j +1 =
α j f i + 1 , j −1 + (1 + β ) f
j i +1 , j + γ j f i +1 , j +1 (5 .5 )
w h e re
∆t
α j=
4
σ 2 j 2 − rj( )
∆t
βj = −
2
σ 2 j2 + r ( )
∆t
γj=
4
σ 2 j 2 + rj( )
fo r i = N - 1 , N - 2 , ..., 1 , 0 a n d j = 1 , 2 , ..., M - 1 .
S
Smax=M∆S
(j+1)∆S x x
j∆S x x
(j-1)∆S x x
0
0 (i-1)∆t i∆t (i+1)∆t T=N ∆t t
Implementation
Equation (5.5) can be rewritten in matrix form:
M1fi = M 2fi +1 + b
where fi and bi are (M − 1) dimensional vectors
T
fi = ⎡⎣ fi ,1 , fi ,2 , fi ,3 , fi ,M −1 ⎤⎦ ,
b = ⎡⎣α1 ( fi ,0 + fi +1,0 ) , 0 ,γ M −1 ( fi ,M + f i +1,M ) ⎤⎦

T
and M1 and M 2 are (M − 1) × (M − 1) symmetric matrices

⎡1 − β1 −γ 1 0 0 ⎤
⎢ −α 1 − β γ2 0 ⎥
⎢ 2 2 ⎥
M1 = ⎢ 0 α3 ⎥
⎢ ⎥
⎢ −γ M − 2 ⎥
⎢⎣ 0 0 −α M −1 1 − β M −1 ⎥⎦
Implementation
and
⎡1 + β1 γ1 0 0 ⎤
⎢ α 1 + β γ 0 ⎥
⎢ 2 2 2 ⎥
M2 = ⎢ 0 α3 ⎥
⎢ ⎥
⎢ γ M −2 ⎥
⎢⎣ 0 0 α M −1 1 + β M −1 ⎥⎦
75
Example
We compare Crank-Nicolson Methods for a European put

with the exact Black-Scholes formula, where T = 5/12 yr,
S0=$50, K = $50, σ=30%, r = 10%.

CN Method with Smax=$100, ∆S=2, ∆t=5/1200: $2.8241
76
Example (Stability)
We compare Crank-Nicolson Method for a European put with

the exact Black-Scholes formula, where T = 5/12 yr, S0=$50,
K = $50, σ=30%, r = 10%.

CN Method with Smax=$100, ∆S=1.5, ∆t=5/1200: $3.1370
77
Example: Barrier Options
Barrier options are options where the payoff depends on whether

the underlying asset’s price reaches a certain level during a certain
period of time.
Types:
knock-out: option ceases to exist when the asset price reaches a barrier
Knock-in: option comes into existence when the asset price reaches a barrier
78
Example: Barrier Option
We compare Crank-Nicolson Method for a European down-

and-out put, where T = 5/12 yr, S0=$50, K = $50, Sb=$50
σ=40%, r = 10%.
What are the boundary conditions for S?
79
Example: Barrier Option
We compare Crank-Nicolson Method for a European down-

and-out put, where T = 5/12 yr, S0=$50, K = $50, Sb=$50
σ=40%, r = 10%.
Boundary conditions are f ( t,S max ) = 0 and f ( t,Sb ) = 0
Exact Price (Hall, p533-535): $0.5424

CN Method with Smax=$100, ∆S=0.5, ∆t=1/1200: $0.5414
80
Appendix A.
Matrix Norms

Vector Norms
- Norms serve as a way to measure the length of a vector or a matrix

- A vector norm is a function mapping x ∈ ℜ n to a real number x s.t.
• x > 0 for any x ≠ 0; x = 0 iff x = 0
• c x = c x for any c ∈ ℜ
• x + y ≤ x + y for any x, y ∈ ℜ n
- There are various ways to define a norm
1
⎛ n
p⎞
p
x p
≡ ⎜ ∑ xi ⎟ (p = 2 is the Euclidean norm )
⎝ i =1 ⎠
x ∞
≡ max xi
1≤ i ≤ n
− For example, v = [2 4 - 1 3]. v 1 = ?, v ∞

= ?, v 2 = ? 82
Matrix Norms
- Similarly, a matrix norm is a function mapping A ∈ ℜ m×n to a real number A s.t.
• A > 0 for any A ≠ 0; A = 0 iff A = 0
• c A = c A for any c ∈ ℜ
• A + B ≤ A + B for any A, B ∈ ℜ m×n
- Various commonly used matrix norms
1
Ax ⎛ m n ⎞ 2
≡ ⎜⎜ ∑∑ aij
2
A ≡ sup
p
A ⎟
p F ⎟
x≠0 x p ⎝ i =1 j =1 ⎠
m n
A 1 ≡ max ∑a
1≤ j ≤ n i =1
ij A ∞
≡ max
1≤ i ≤ m
∑a
j =1
ij
( )
A 2 ≡ ρ A T A , the spectral norm, where
ρ (B ) ≡ max{λk : λk is an eigenvalue of B}
An Example
⎡ 2 4 − 1⎤
⎢
A=⎢ 3 1 5⎥ ⎥
⎢⎣− 2 3 − 1⎥⎦
A ∞
=? A 2 =?
A1 =? A F
=?
84
Basic Properties of Norms
n×n
Let A, B ∈ ℜ and x, y ∈ ℜ . Then
n
1. x ≥ 0; and x = 0 ⇔ x = 0
2. x + y ≤ x + y
3. α x = α x where α is a real number
4. Ax ≤ A x
5. AB ≤ A B
85
Condition number of a square matrix
( )
All norms in ℜn ℜm×n are equivalent. That is, if • α and • β are norms
on ℜn , then ∃ c1 ,c2 > 0 such that for all x ∈ℜn , we have
c1 x α ≤ x β
≤ c2 x α
Condition Number of A Matrix: C ≡ A A−1 , where A ∈ℜn×n .

The condition number gives a measure of how close a matrix is
close to singular. The bigger the C, the harder it is to solve Ax = b.
86
Convergence
- vectors x k converges to x ⇔ x k − x converges to 0
- matrix A k → 0 ⇔ A k − 0 → 0
87
Appendix B.
Basic Row Operations

Basic row operations
= *
Three kinds of basic row operations:

1) Interchange the order of two rows or (equations)
⎡0 1 0⎤⎡a11 a12 a13 ⎤ ⎡a21 a22 a23 ⎤
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢1 0 0⎥⎢a21 a22 a23 ⎥ = ⎢a11 a12 a13 ⎥
⎢⎣0 0 1⎥⎦⎢⎣a31 a32 a33 ⎥⎦ ⎢⎣a31 a32 a33 ⎥⎦
89
Basic row operations = *
2) Multiply a row by a nonzero constant

⎡c 0 0⎤⎡a11 a12 a13 ⎤ ⎡ca11 ca12 ca13 ⎤
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢0 1 0⎥⎢a21 a22 a23 ⎥ = ⎢ a21 a22 a23 ⎥
⎢⎣0 0 1⎥⎦⎢⎣a31 a32 a33 ⎥⎦ ⎢⎣ a31 a32 a33 ⎥⎦
3) Add or subtract rows

⎡ 1 0 0⎤⎡ a11 a12 a13 ⎤ ⎡ a11 a12 a13 ⎤
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ −1 1 0 a
⎥⎢ 21 a 22 a23 ⎥ = ⎢a21 − a11 a22 −a12 a23 −a13 ⎥
⎢⎣ 0 0 1⎥⎦⎢⎣a31 a32 a33 ⎥⎦ ⎢⎣ a31 a32 a33 ⎥⎦
90

Good Notes On FDM For Option Pricing PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Good Notes On FDM For Option Pricing PDF

Uploaded by

Copyright:

Available Formats

Chapter 5 Finite Difference

Math6911 S08, HM Zhu

1. Chapters 5 and 9, Brandimarte

2. Section 17.8, Hull

3. Chapter 7, “Numerical analysis”, Burden and Faires

• Finite difference (FD) approximation to the derivatives

5.1 Finite difference approximations

Math6911 S08, HM Zhu

• Aim to approximate the values of the continuous function

Math6911, S08, HM ZHU

The basic idea of FDM is to replace the partial derivatives

∂ 2 f ( t,S ) f ( t,S + ∆S ) − 2 f ( t,S ) + f ( t,S − ∆S )

• Depending on which combination of schemes we use in

5.2.1 Explicit Finite-Difference Method

Math6911 S08, HM Zhu

Math6911, S08, HM ZHU

Rewriting the equation, we get an explicit schem e:

1. Starting with the final values f N , j , we apply (5.2) to solve

2. Repeat the process to determine f N − 2 , j and so on

We compare explicit finite difference solution for a

Black-Scholes Price: $2.8446

We compare explicit finite difference solution for a European

Black-Scholes Price: $2.8446

5.2.2 Numerical Stability

Math6911 S08, HM Zhu

• The problem itself

Suppose we have mathematically posed problem:

• Conditioning issue is related to the problem itself, not to the

The concept "near" can be measured by further information about

Let x be any real number.

When additional calculations are done, there is an

Stability ensures if the error between the numerical

Stability concerns about the behavior of fi , j − f ( i∆t, j∆S )

Convergence of the numerical algorithm concerns about

For well-posed linear initial value problem,

• These factors contribute the accuracy of a numerical solution.

5.2.3 Financial Interpretation of Numerical Instability

Math6911 S08, HM Zhu

If ∂ f ∂ S and ∂ 2 f ∂ S 2 are assum ed to be the sam e at ( i + 1, j )

Check if the mean and variance of the

which is coherent with geometric Brownian motion in a risk-neutral world

Math6911, S08, HM ZHU

Define Z = ln S. The B-S equation becomes

The corresponding difference equation is

Get rid of the varying coefficients S and S² by using

With u nm = u (n δx,mδ τ ), this involves solving a system of

5.3.1 Implicit Finite-Difference Method

Math6911 S08, HM Zhu

Math6911, S08, HM ZHU

Rewriting the equation, we get an implicit scheme:

Equation (5.6) can be rewritten in matrix form:

1. Starting with the final values f N , j , we need to solve a

2. Repeat the process to determine f N − 2 , j and so on

We compare implicit finite difference solution for a

Black-Scholes Price: $2.8446

We compare implicit finite difference solution for a European

Black-Scholes Price: $2.8846

• The explicit finite difference method is

• It is better to have ln S rather than S as

Math6911, S08, HM ZHU

5.3.2 Solving a linear system using direct methods

Math6911 S08, HM Zhu