Professional Documents
Culture Documents
=
o
(4)
Now, we extend the binomial model to two periods. Let
uu
f denote the call value at time t o 2 for two consecutive
upward stock movements,
ud
f for one downward and one
upward movement and
dd
f for two consecutive downward
movements of the stock price [9]. Then we have
) 0 , max( K Suu f
uu
= (5)
) 0 , max( K Sud f
ud
= (6)
) 0 , max( K Sdd f
dd
= (7)
The values of the call options at time t o are
] ) 1 ( [
ud uu
t r
u
f p pf e f + =
o
(8)
] ) 1 ( [
dd ud
t r
d
f p pf e f + =
o
(9)
Substituting (8) and (9) into (3), we have
[ (1 )
(1 ) ( (1 ) )]
r t r t
uu ud
r t
ud dd
f e pe f p f
p e pf p f
o o
o
= +
+ +
)] ) 1 ( ) 1 ( 2 [
2 2 2
dd ud uu
t r
f p f p p f p e f + + =
o
(10)
Equation (10) is called the current call value, where the
numbers
2
p , ) 1 ( 2 p p and
2
) 1 ( p are the risk
neutral probabilities for the underlying asset
prices Suu, Sud and Sdd respectively.
W generalize the result in (10) to value an option at
t N T o = as follows
=
N
j
d u
j N j
j
N t Nr
j N j
f p p C e f
0
) 1 (
o
=
=
N
j
j N j j N j
j
N t Nr
K d Su p p C e f
0
) 0 , max( ) 1 (
o
(11)
Where ) 0 , max( K d Su f
j N j
d u
j N j
=
and
! )! (
!
j j N
N
C
j
N
=
=
N
j
j N j j N j
j
N t Nr
d u p p C Se f
0
) 1 (
o
0
(1 )
N
Nr t N j N j
j
j
Ke C p p
o
=
(12)
Equation (12) gives us the present value of the call option.
The term
t Nr
e
o
is the discounting factor that reduces f to
its present value. We can see from the first term of (12)
that -1is the binomial probability of j upward movements
to occur after the first N trading periods and
j N j
d Su
is
the corresponding value of the asset after j upward
movements of the stock price. The second term of (12) is
the present value of the options strike price.
Let
t r
e Q
o
= , we substitute Qin the first term of (12) to
yield
International Journal of IT, Engineering and Applied Sciences Research (IJIEASR) ISSN: 2319-4413
Volume 2, No. 2, February 2013
i-Xplore International Research Journal Consortium www.irjcjournals.org
7
=
=
N
j
j N j j N j
j
N N
d u p p C SQ f
0
) 1 (
0
(1 )
N
Nr t N j N j
j
j
Ke C p p
o
=
=
=
N
j
j N j
j
N
d p Q pu Q C S f
0
1 1
] ) 1 ( [ ] [
0
(1 )
N
Nr t N j N j
j
j
Ke C p p
o
=
(13)
Now, let ) , ; ( p N m u be the binomial distribution
function given by
= u
N
j
j N j
j
N
p p C p N m
0
) 1 ( ) , ; ( (14)
Equation (14) is the probability of at least m success in
N independent trials, each resulting in a success with
probability p and in a failure with probability ) 1 ( p .
Then let pu Q p
1
= ' and d p Q p ) 1 ( ) 1 (
1
= '
.
Consequently, it follows that
) , ; ( ) , ; ( p N m Ke p N m S f
rT
u ' u =
(15)
The model in (15) was developed by Cox-Ross
Rubinstein [6], where
N
T
t = o and we will refer to it as
CRR model. The corresponding put value of the
European options can be obtained using call put
relationship of the form S P Ke C
E
rt
E
+ = +
as
) , ; ( ) , ; ( p N m S p N m Ke f
rT
' u u =
(16)
Where the risk free interest rate is denoted by r ,
E
C is
the European call,
E
P is the European put and S is the
initial stock price. European option can only be exercised
at expiration, while for an American option, we check at
each node to see whether early exercise is advisable to
holding the option for a further time period t o . When
early exercise is taken into consideration, the fair price
must be compared with the options intrinsic value.
2.1.2 NUMERICAL IMPLEMENTATION [8]
Now, we present the implementation of binomial model
for pricing vanilla options as follows.
When stock price movements are governed by a multi-step
binomial tree, we can treat each binomial step separately.
The multi-step binomial tree can be used for the American
and European style options.
Like the Black-Scholes, the CRR formula in (15) can only
be used in the pricing of European options and easily be
implemented in Matlab. To overcome this problem, we
use a different multi-period binomial model for the
American style options on both the dividend and non-
dividend paying stocks. Now we present the Matlab
implementation.
The stock price of the underlying asset for non-dividend
and dividend paying stocks are given respectively by
1 ,... 1 , 0 , ,..., 1 , 0 , = =
i N N j d Su
j N j
(17)
,... 1 , , ,..., 1 , 0 , ) 1 ( + = =
i i N N j d u S
j N j
(18)
Where the dividend is denoted by that reduces
underlying price of the asset. For the European call and
put options, the Matlab code takes into consideration on
the prices at the maturity date T and the stock prices for
non-dividend paying stocks in (17). The call and put prices
of European option are given by (15) and (16)
respectively.
For the American call and put options, the Matlab code
will incorporate the early exercise privilege and the
dateT , when the dividend will be paid. Then, it implies
that the stock prices will exhibit (17) and (18). The call
and put prices of American option for non-dividend
paying stock are given by
))] , ; ( ) , ; ( ( , max[ p N m Ke p N m S K S f
rT
T
u ' u =
and
))] , ; ( ) , ; ( ( , max[ p N m S p N m Ke S K f
rT
T
' u u =
respectively. For dividend paying stock, we replace (17)
with (18) in (12) and substitute in the last two equations to
get respectively the call and put prices of American
option.
2.2 FINITE DIFFERENCE METHOD
Many option contract values can be obtained by solving
partial differential equations with certain initial and
boundary conditions. The finite difference approach is one
of the premier mathematical tools employed to solve
partial differential equations. These methods were
pioneered for valuing derivative securities by Brennan and
Schwarz [5]. The most common finite difference methods
for solving the Black-Scholes partial differential equations
are the
- Explicit Method.
- Implicit Method.
- Crank Nicolson method.
These schemes are closely related but differ in stability,
accuracy and execution speed, but we shall only consider
International Journal of IT, Engineering and Applied Sciences Research (IJIEASR) ISSN: 2319-4413
Volume 2, No. 2, February 2013
i-Xplore International Research Journal Consortium www.irjcjournals.org
8
implicit and Crank Nicolson schemes. In the formulation
of a partial differential equation problem, there are three
components to be considered.
- The partial differential equation.
- The region of space time on which the partial
differential is required to be satisfied.
- The ancillary boundary and initial conditions to
be met.
2.2.1 DISCRETIZATION OF THE
EQUATION
The finite difference method consists of discretizing the
partial differential equation and the boundary conditions
using a forward or a backward difference approximation.
The Black-Scholes partial differential equation is given by
) , (
2
) , (
2 2
t S S
t
S t t t
S t rf f
S
f rS S t f
t t t
= + +
o
(19)
We discretize (1) with respect to time and to the
underlying price of the asset. Divide the ) , (
t
S t plane
into a sufficiently dense grid or mesh and approximate
the infinitesimal steps
t
A and
t
S
A by some small fixed
finite steps. Further, define an array of 1 + N equally
spaced grid points
N
t t ,...,
0
to discretize the time
derivative with
t n n
N
T
t t A = =
+1
. Using the same
procedures, we obtain for the underlying price of the
asset as follows:
t
S M M
M
S
S S A = =
+
max
1
.
This gives us a rectangular region on the ) , (
t
S t plane
with sides ) , 0 (
max
S and ) , 0 ( T . The grid coordinates
) , ( m n enables us to compute the solution at discrete
points. We will denote the value of the derivative at time
step
n
t when the underlying asset has value
m
S as
) , ( ) , ( ) , (
, t m n n m
S t f S t f S m t n f f = = A A =
(20)
where n and m are the numbers of discrete increments
in the time to maturity and stock price respectively.
2.2.2 FINITE DIFFERENCE
APPROXIMATIONS
In finite difference method, we replace the partial
derivative occurring in the partial differential equation by
approximations based on Taylor series expansions of
function near the points of interest [9]. Expanding
) , ( S S t f + A and ) , ( S S t f A in Taylor series we
have the forward and backward difference respectively
with ) , ( S t f represented in the grid by
m n
f
,
[1]:
t
m n m n
S
S
f f
f
t
A
~
+ , 1 ,
(21)
t
m n m n
S
S
f f
f
t
A
~
1 , ,
(22)
Also the first order partial derivative results in the
central difference given by
t
m n m n
S
S
f f
f
t
A
~
+
2
1 , 1 ,
(23)
And the second order partial derivative gives symmetric
central difference approximation of the form
2
1 , , 1 ,
2
t
m n m n m n
S S
S
f f f
f
t t
A
~
+
(24)
Similarly, we obtained forward difference approximation
for the maturity time given by
t
m n m n
t
S
f f
f
A
~
+ , , 1
(25)
Substituting equations (23), (24) and (25) into (19), we
have
m n m n m m n m m n m
f f f f
, 1 1 , 3 , 2 1 , 1 + +
= + + (26)
Where
t m t rm
m
A A =
2 2
1
2
1
2
1
o
t m t r
m
A + A + =
2 2
2
1 o ,
t m t rm
m
A A =
2 2
3
2
1
2
1
o ,
(26) is called a finite difference equation which gives
equation that we use to approximate the solution
of ) , ( S t f [4].
Similarly, we obtained for the explicit, implicit and Crank
Nicolson finite difference method as follows [7]:
Explicit case:
m n m n m m n m m n m
t
f f f f
r
, 1 , 1 3 , 1 2 1 , 1 1
) (
1
1
= + +
+
+ + + +
o o o
o
(27)
Where
t m
t t
m
m
rm m
o o o
o o o
o
2 2
2
2 2
1
1 ,
2 2
= =
and
2 2
2 2
3
t t
m
rm m o o o
o + = .
International Journal of IT, Engineering and Applied Sciences Research (IJIEASR) ISSN: 2319-4413
Volume 2, No. 2, February 2013
i-Xplore International Research Journal Consortium www.irjcjournals.org
9
This method is accurate to ) (
2
,
S
t
O o o .
For implicit case we have,
m n m n m m n m m n m
t
f f f f
r
, 1 1 , 3 , 2 1 , 1
) (
1
1
+ +
= + +
| | |
o
(28)
Where the parameters in (28) are given
by
t m
t t
m
m
rm m
o o |
o o o
|
2 2
2
2 2
1
1 ,
2 2
+ = + =
and
2 2
2 2
3
t t
m
rm m o o o
| = .
Similar to the explicit method, implicit method is accurate
to ) (
2
,
S
t
O o o .
Crank Nicolson method is obtained by taking the average
of the explicit and implicit methods in (27) and (28)
respectively. Then we have
1 , 1 2 , 3 , 1
1 1, 1 2 1, 3 1, 1
m n m m n m m n m
m n m m n m m n m
f f f
f f f
v v v
| | |
+
+ + + + +
+ +
= + +
(29)
Then the parameters are given by
4 4
2 2
1
t m t rm
m
A
A
=
o
v ,
2 2
1
2 2
2
t m t r
m
A
+
A
+ =
o
v ,
4 4
2 2
3
t m t rm
m
A
A
=
o
v
4 4
2 2
1
t m t rm
m
A
+
A
=
o
,
2 2
1
2 2
21
t m t r
m
A
A
=
o
,
.
4 4
2 2
3
t m t rm
m
A
+
A
=
o
1 ,..., 2 , 1 , 0 = N n and 1 ,..., 2 , 1 = M m [8].
2.2.3 STABILITY ANALYSIS
The two fundamental sources of error are the truncation
error in the stock price discretization and in the time
discretization. The importance of truncation error is that
the numerical scheme solves a problem that is not exactly
the same as the problem we are trying to solve.
The three fundamental factors that characterize a
numerical scheme are consistency, stability and
convergence [9, 10].
- Consistency: A finite difference of a partial
differential equation is consistent, if the
difference between partial differential equation
and finite differential equation vanishes as the
interval and time step size approach zero.
Consistency deals with how well the finite
difference equation approximates the partial
differential equation and it is the necessary
condition for convergence.
- Stability: For a stable numerical scheme, the
errors from any source will not grow
unboundedly with time.
- Convergence: It means that the solution to a finite
difference equation approaches the true solution
to the partial differential equation as both grid
interval and time step sizes are reduced. The
necessary and sufficient conditions for
convergent are consistency and stability.
These three factors that characterize a numerical
scheme are linked together by Lax equivalence
theorem [9] which states that given a well posed
linear initial value problem and a consistent finite
difference scheme, stability is the necessary and
sufficient condition for convergence.
In general, a problem is said to be well posed if:
- A solution to the problem exists.
- The solution is unique when it exists.
- The solution depends continuously on the
problem data.
2.2.3.1 A NECESSARY AND
SUFFICIENT CONDITION FOR
STABILITY
Let
n n
Af f =
+1
be a system of equations, where
Aand
1 + n
f are matrix and column vectors
respectively. Then
1
=
n n
Af f
2
2
=
n
f A
0
f A
n
= (30)
For N n ,..., 2 , 1 = and
0
f is the vector of initial
value. We are concerned with stability and we also
perturbed the vector of the initial value
0
f to
0
t . The
exact solution at the
th
n row will then be
0
t A t
n
n
= (31)
Let the perturbation or error vector e be denoted by
f t e =
and using the perturbation vectors (30) and (31), we
have
n n n
f t e =
0
f A
n
=
0
t A
n
= ) (
0 0
f t A
n
International Journal of IT, Engineering and Applied Sciences Research (IJIEASR) ISSN: 2319-4413
Volume 2, No. 2, February 2013
i-Xplore International Research Journal Consortium www.irjcjournals.org
10
Therefore,
0
e A e
n
n
= (32)
Hence for compatible matrix and vector norms [11]
n
n
A e s
0
e
Lax and Richmyer defined the difference scheme to be
stable when there exists a positive number L which is
independent of n ,
t
o and
s
o , then L A s . This limits
the amplification of any initial perturbation and therefore
of any arbitrary initial rounding errors, since
L e
n
s
0
e and
n
n
A A s , then the Lax-
Richmyer definition of stability is satisfied when
1 s A (33)
Hence (33) is the necessary and sufficient condition for
the finite difference equations to be stable [11]. Since
the spectral radius ( ) A satisfies A A s ) ( , it
follows from (33) that 1 ) ( s A .
By Lax equivalence theorem, the three finite difference
methods are consistent and convergent but in the analysis
of their stability, explicit method is quite stable, while the
implicit and Crank Nicolson methods are conditionally
and unconditionally stable finite difference methods
respectively because they calculate small change in the
option value for a small change of the initial conditions,
converge to the solution of the partial differential
equation and calculation error decreases when number of
time and price partitions increase.
3. NUMERICAL EXAMPLES
This section presents some numerical examples as
follows:
Example 1
We consider the convergence of the binomial model and
the Crank Nicolson finite difference method with relation
to the Black-Scholes value of the option.
We price the European call option on a non-dividend
paying stock with the following parameters:
1 , 2 . 0 , 05 . 0 , 60 , 50 = = = = = T r K S o
The Black-Scholes price for the call option is 1.6237.
Table 1and 2 below shows the illustrative result for the
performance of the two methods under consideration
when the value of M and N are the same and different
respectively.
Example 2
We shall consider the robustness of the two methods
against the true Black-Scholes price for a European put
with the parameters
50, 0.05, 0.25, 3 K r T o = = = =
The result obtained is shown in Table 3 below.
3.1 TABLE OF RESULTS
Table 1: The comparison of the convergence of the
Implicit method and the Crank Nicolson method as we
increase M and N
M N =
Binomial
Model
Crank Nicolson
Finite Difference
Method
10 1.6804 1.3113
20 1.5900 1.4957
30 1.6373 1.5423
40 1.6442 1.5603
50 1.6386 1.5692
60 1.6289 1.5743
70 1.6179 1.5776
80 1.6178 1.5798
90 1.6254 1.5814
100 1.6293 1.5826
Table 2: The Illustrative Result for the Performance of
Binomial method and the Crank Nicolson Finite
Difference Method for different values of M and N
M N
Binomial
Model
Crank Nicolson
Finite Difference
Method
10 20 1.6804 1.5731
20 40 1.5900 1.6108
30 60 1.6373 1.6180
40 80 1.6442 1.6205
50 100 1.6386 1.6216
60 120 1.6289 1.6222
70 140 1.6179 1.6225
80 160 1.6178 1.6227
90 180 1.6254 1.6229
100 200 1.6293 1.6230
Table 3: A Comparison with the Black-Scholes Price for
a European Put Option
S
Black-
Scholes
Binomial
Model
Finite Difference
Method
50 4.9564 4.9556 4.9563
60 2.7621 2.7640 2.7612
70 1.5328 1.5346 1.5325
80 0.8538 0.8549 0.8537
90 0.4797 0.4803 0.4794
International Journal of IT, Engineering and Applied Sciences Research (IJIEASR) ISSN: 2319-4413
Volume 2, No. 2, February 2013
i-Xplore International Research Journal Consortium www.irjcjournals.org
11
3.2 DISCUSSION OF RESULTS
Table 1 shows that the binomial model is closer to the
Black-Scholes values for small values of N than the
Crank Nicolson finite difference method as N ,
0
t
o and M , 0
s
o .
Table 2 shows that when N and M are of different
values, the finite difference method converges faster than
its counterpart binomial model when N is equal to M .
For Crank Nicolson scheme, the number of time steps
N initially set at 10and doubled with each grid
M refinement. We conclude that the Crank Nicolson
finite difference method has a higher accuracy than the
implicit method and therefore it converges faster. The
above results highlight that the two methods are stable.
Table 3shows the variation of the option price with the
underlying price S . The results demonstrate that the two
methods perform well, are mutually consistent and agree
with the Black-Scholes value. However, in general finite
difference method is far better suited for the pricing of this
contract.
4. CONCLUSION
Options can be classified into two flavours namely vanilla
and exotic. Binomial model and finite difference method
are suited to dealing with some of these option flavours.
In general, each numerical method has its strengths and
weaknesses of use. Binomial model is good for pricing
options with early exercise opportunities, accurate,
converges faster as we can see in Table 1 and it is
relatively easy to implement but can be quite hard to adapt
to more complex situations.
Finite difference methods require sophisticated algorithms
for solving large sparse linear systems of equations but
cannot be used in high dimensions. They are flexible in
handling different processes for the underlying state
variables and relatively difficult to code but these methods
are somewhat problematic for path dependent options.
From Tables 2 and 3, we conclude that finite difference
method is more stable, accurate, converges faster and it is
more robust than its counterpart binomial model when
pricing European options.
REFERENCES
[1] W. Ames, Numerical Methods for Partial Differential
Equations, Academic Press, New York, 1977.
[2] F. Black and M. Scholes, The Pricing of Options and
Corporate Liabilities, Journal of Political Economy,
Vol. 81, no. 3, pp. 637-654, 1973.
[3] P. Boyle, Options: A Monte Carlo Approach, Journal
of Financial Economics. Vol. 4, no. 3, pp. 323-338,
1977.
[4] P. Boyle, M. Broadie, and P. Glasserman, Monte
Carlo Methods for Security Pricing, Journal of
Economic Dynamics and Control, Vol. 21, no. 8-9,
1267-1321, 1997.
[5] M. Brennan and E. Schwartz, Finite Difference
Methods and Jump Processes Arising in the Pricing of
Contingent Claims, Journal of Financial and
Quantitative Analysis, Vol. 5, no. 4, pp. 461-474,
1978.
[6] J. Cox, S. Ross and M. Rubinstein, Option Pricing: A
Simplified Approach, Journal of Financial
Economics Vol. 7, pp. 229-263, 1979.
[7] S. E. Fadugba, F. H. Adefolaju and O. H. Akintudire,
On the Stability and Accuracy of Finite Difference
Method for Options Valuation, Journal of
Mathematical Theory and Modeling, USA. Vol. 2,
No. 6, 60-66, Jul. 2012.
[8] S. E. Fadugba, J. T. Okunlola and A. O. Adeyemo,
On the Strength and Weakness of Binomial Model
for Pricing Vanilla Options, International Journal of
Advanced Research in Engineering and Applied
Sciences, Vol. 1, no. 1, pp. 13-22, Jul. 2012.
[9] J. Hull, Options, Futures and other Derivatives,
Pearson Education Inc., 5th ed., Prentice Hall, New
Jersey, 2003.
[10] C. R. Nwozo and S. E. Fadugba, Some Numerical
Methods for Options Valuation, Communication in
Mathematical Finance, Scienpress Ltd, UK, Vol. 1,
no. 1, Aug. 2012.
[11] G. D. Smith, Numerical Solution of Partial
Differential Equations: Finite Difference Methods,
Clarendon Press, 3rd ed., Oxford, 1985.