54 3 Utility Maximisation on Finite Probability Spaces
investment as theFN− 1 -measurable function̂hNdefined bŷhN=
̂h
SN− 1 .Eco-
nomically speaking, the valuêhNSN− 1 of the proportion of the investment in
the stock is constant, while the number̂hNof stocks in the portfolio depends
on the current stock priceSN− 1 : it is simply inversely proportional toSN− 1.
To computeuN− 2 (x), note that this step again is reduced to the analysis of
the one step problem at times{N− 2 ,N− 1 }, where we now have to replace
the original utility functionU(x) by the conditional utility functionuN− 1 (x).
This is just the principle of dynamic programming which reduces to an obvious
fact in the present context. Hence
uN− 2 (x)=sup{E[uN− 1 (x+h∆SN− 1 )]|h∈R}
=c^2 UU(x).
By induction we conclude that
ut(x)=cNU−tU(x),t=0,···,N,
so that we obtain in particular for the value functionu(x)=u 0 (x)
u(x)=cNUU(x). (3.46)
In order to compute the parameters of the optimal investment strategy,
we now assume that there is a fixed horizonT>0 such thatN∆t=Tand
we letN→∞,sothat∆t=TN→0. As above we define ̃u=σ∆t
(^12)
+ν∆t,
d ̃=−σ∆t^12 +μ∆t. We have found in (3.40) that
cU=1−
βν^2
2 σ^2
∆t+o(∆t)
so that
cNU=c
∆Tt
U =exp
(
−
βν^2 T
2 σ^2
)
+o(1).
The optimal investment strategy
(
X̂t(x)
)T
t=0
for initial wealthx>0is
given by
X̂t(x)=x+
∑t
n=1
̂ht∆St
where
̂ht=
X̂t− 1 (x)
St− 1
̂k
and as in (3.42)
̂k=
̂h
x
+O(∆t
(^12)
)=
(1−β)ν
σ^2
(3.47)
is the ratio of the current wealthX̂t− 1 (x) invested into the stock. We thus
find the discrete version of the well-known “Merton-line” investment strategy
[M 90]; the latter applies to the continuous time limit, i.e., the Black-Scholes
model.