Вы находитесь на странице: 1из 3

Vasicek PDE Solution

Mauricio Bedoya
javierma36@gmail.com
July 2014
Here Im going to show a procedure that allow to get the nal expression of Vasicek interest
rate model:
dr
(t)
= a(
b
a
r
(t)
)dt + dw
(t)
(1)
The procedure can be implemented with other interest rate models, and believe me, it works.
More information about the properties of Vasicek interest rate model, can be found in: Monte
Carlo Methods in Financial Engineering - P. Glasserman and Stochastic Calculus for Finance
II - Shreve R. E.
Equation 1 is a linear rst order PDE. In Elementary Dierential Equations - Edwards and
Penney; show the procedure to follow that solve the PDE. Edwards and Penney book is the
one used to teach Dierential Equations in MIT. Take a look, is really nice.
This kind of PDE can be solved using the integrator factor method. To understand better the
method, I will show you rst an example.
dy
dx
= y
11
8
e
x
3
(2)
1. Step 1 Write the equation in standard form
dy
dx
y =
11
8
e
x
3
2. Find the integrator factor
Q(x) = e

1dx
The -1 is the term that multiplies y.
Q(x) = e
x
1
3. Multiply booth sides by the integrator factor
e
x

dy
dx
e
x
y = e
x

11
8
e
x
3
d(e
x
y) =
11
8
e
4x
3
4. Integrate booth sides

d(e
x
y) =
11
8

e
4x
3
dx
e
x
y =
33
32
e
4x
3
+ C
If initial conditions are dened, C value has an expression. To prove that this is the
solution to the original PDE, implement the derivative and compare.
In Edwards and Penney book you can nd a lot of exercise. When you search the solution of
this procedure in the web, you may nd: lets multiply both sides by blablabla. They dont
show you why that magical expression allow to solve the PDE. Now you know the reasoning
behind.
To nd the integrator factor in Vasicek model, write the model in the standard form (form used
in PDE):
dr
(t)
+ a r
(t)
dt = a dt + dw
(t)
(3)
then the integrator factor is: e

adt
= e
at
. Multiplying both sides of equation 3 by the integrator
factor, to have
e
at
dr
(t)
+ e
at
a r
(t)
dt = e
at
b dt + e
at
dw
(t)
(4)
rewriting 4,
d(e
at
r
(t)
) = e
at
b dt + e
at
dw
(t)
(5)
now integrating in the interval [0,T]

T
0
d(e
at
r
(t)
) =

T
0
e
at
b dt +

T
0
e
at
dw
(t)
(6)
e
aT
r
(T)
r
(0)
=
b
a
(e
aT
1) +

T
0
e
at
dw
(t)
nally,
r
(T)
= r
(0)
e
aT
+
b
a
(1 e
aT
) + e
aT

T
0
e
at
dw
(t)
(7)
The expectation of r
(t)
is:
2
E[r
(T)
] = r
(0)
e
aT
+
b
a
(1 e
aT
) (8)
because e
aT

T
0
e
at
E[dw
(t)
] = 0. Remember that dw
(t)
N[0, dt].
The variance of r
(T)
is:
V [r
(T)
] = E[r
2
(T)
] (E[r
(T)
])
2
We already have E[r
(T)
]. What we need is E[r
2
(T)
]. Then:
E[r
2
(T)
] = E[(r
(0)
e
aT
+
b
a
(1 e
aT
) + e
aT

T
0
e
at
dw
(t)
)
2
]
E[r
2
(T)
] = E[(r
(0)
e
aT
+
b
a
(1 e
aT
))
2
+ 2 (r
(0)
e
aT
+
b
a
(1 e
aT
)) (e
aT

T
0
e
at
dw
(t)
) + (e
aT

T
0
e
at
dw
(t)
)
2
]
Taking the expectation and knowing that dw
(t)
N[0, dt], we get:
E[r
2
(T)
] = [(r
(0)
e
aT
+
b
a
(1 e
aT
))
2
] + (e
2aT

T
0
e
2at

2
dt) (9)
replacing E[r
(T)
] and E[r
2
(T)
] in V [r
(T)
], we nally get
V [r
(T)
] = (e
2aT

T
0
e
2at

2
dt) (10)
or
V [r
(T)
] = (e
2aT


2
2 a
[e
2aT
1]) =

2
2 a
[1 e
2aT
] (11)
3

Вам также может понравиться