Applied Mathematics, 2011, 2, 912-913
doi:10.4236/am.2011.27123 Published Online July 2011 (http://www.SciRP.org/journal/am)
Copyright © 2011 SciRes. AM
An Alternative Method of Stochastic Optimization: The
Portfolio Model
Moawia Alghalith
University of the West India, Saint Augustine, Trinidad and Tobago
E-mail: malghalith@gmail.com
Received May 19, 2011; revised May 26, 2011; accep te d Ma y 29, 2011
Abstract
We provide a new simple approach to stochastic dynamic optimization. In doing so, we derive the existing
(standard) results using a far simpler technique than the duality and the variational methods.
Keywords: Stochastic Optimization, Investment, Portfolio
1. Introduction
Previous studies in stochastic optimization relied on the
duality approach and/or variational techniques such as
using the Feynman Kac formula and the Hamilton-
Jacobi-Bellman partial differential equations. Examples
include [1-3], among many others.
In this paper, we offer a new simple approach to
stochastic dynamic optimization. That is, we prove the
previous results using a simpler method than the duality
or the Hamilton-Jacobi-Bellman partial differential
equations methods. We apply our method to the standard
investment model. Our approach is based on dividing the
time horizon into sub-horizons and applying Stein’s
lemma.
2. The Portfolio M odel
We use the standard investment model (see, for example,
[3], among many others). Similar to previous models, we
consider a risky asset and a risk-free asset. The risk-free
asset price process is given by where
is the rate of return.
0=
T
rds
t
Se
,
2
b
rCR
The dynamics of the risky asset price are given by
d=dd ,
ss s
SS s W
(1)
where
and
are the deterministic rate of return
and the volatility, respectively, and
W is a standard
Brownian motion.
The wealth process is given by
ππ
=πdπd,
TT
Tssssss
tt
s
xrXr sW
(2)
where
is the initial wealth,
is the risky πstsT
portfolio process with . The trading strategy
2
πd<
T
s
t
Es
s
is admissible (that is, ).
π0
s
X
The investor’s objective is to maximize the expected
utility of the terminal wealth
π
,=Sup= π,
Tt
t
Vtx EUXEU
(3)
where
.V is the (smooth) value function,
.U is
continuous, bounded and strictly concave utility function,
and is the filtration.
We rewrite (2) as
ππ
π
π
=ππ
πdπd
πdπd;
<<,,, ,, .
Tuuuuuuuu
TT
ssss ss
tt
TT
s
ssss ss
tt
XxrX rW
rXr sW
rXr sW
tututT tTutT
s
(4)
Substituting the above equation into (3) and dif-
ferentiating with respect to (and setting the
derivative equal to zero) yields
πu
..
uut uut
rEUEU W
=0.
(5)
By Stein’s lemma
.=, .
=..
utuut
uu t
EUWCovX WEU
EU
(6)