Paper 4, Section II, J

Stochastic Financial Models
Part II, 2009

An agent with utility U(x)=exp(γx)U(x)=-\exp (-\gamma x), where γ>0\gamma>0 is a constant, may select at time 0 a portfolio of nn assets, which he then holds to time 1 . The values X=(X1,,Xn)TX=\left(X_{1}, \ldots, X_{n}\right)^{T} of the assets at time 1 have a multivariate normal distribution with mean μ\mu and nonsingular covariance matrix VV. Prove that the agent will prefer portfolio ψRn\psi \in \mathbb{R}^{n} to portfolio θRn\theta \in \mathbb{R}^{n} if and only if q(ψ)>q(θ)q(\psi)>q(\theta), where

q(x)=xμγ2xVxq(x)=x \cdot \mu-\frac{\gamma}{2} x \cdot V x

Determine his optimal portfolio.

The agent initially holds portfolio θ\theta, which he may change to portfolio θ+z\theta+z at cost εi=1nzi\varepsilon \sum_{i=1}^{n}\left|z_{i}\right|, where ε\varepsilon is some positive transaction cost. By considering the function tq(θ+tz)t \mapsto q(\theta+t z) for 0t10 \leqslant t \leqslant 1, or otherwise, prove that the agent will have no reason to change his initial portfolio θ\theta if and only if, for every i=1,,ni=1, \ldots, n,

μiγ(Vθ)iε\left|\mu_{i}-\gamma(V \theta)_{i}\right| \leqslant \varepsilon