Portfolio optimization for pension plans under hybrid stochastic and local volatility. (Q2343843)

This paper discusses an optimal portfolio selection problem for a defined contribution pension fund, where the price dynamics of a risky asset are governed by a combined constant elasticity of variance and stochastic volatility model, which is called the SVCEV model by the authors. The objective of the pension fund's member is to select an optimal composition of a portfolio consisting of a risk-free bond and a risky asset so as to maximize the expected power revenue on the terminal wealth. The model dynamics and the formulation of the optimal portfolio selection problem are presented in Section 2. In Section 3, the standard dynamic programming approach to stochastic optimal control is applied to discuss the optimal portfolio selection problem and the corresponding Hamilton-Jacobi-Bellman (HJB) equation governing the value of the control problem. Then a parametric form of the solution of the HJB equation is conjectured for the power utility with a view to separating some state variables. Consequently, the HJB equation is split into two equations. One equation is an ordinary differential equation whose closed form solution is obtained. Another equation is a nonlinear partial differential equation. Section 4 is devoted to discuss an approximation to the nonlinear partial differential equation based on asymptotic analysis. The key idea is to introduce a small perturbed parameter \(\epsilon\) which is the reciprocal of the fast mean-reversion parameter \(\alpha\). An asymptotic expansion to the solution of the nonlinear partial differential equation with respect to the small perturbed parameter \(\epsilon\) is then used. An explicit expression is obtained for the leading order term \(g_0\) of the asymptotic expansion. This result is presented in Theorem 1. With this explicit expression for the leading order term \(g_0\), the first correction term \(g_1\) is shown to be a solution of a linear partial differential equation. This result is presented in Theorem 2. Analytical expressions for the leading order term and the first correction term of an asymptotic expansion to the optimal portfolio strategy \(\pi^*\) are obtained. These expressions depend on \(g_0\) and \(g_1\) and are presented in Theorem 3. In Section 5, a particular situation, where the trading strategy does not depend on the unobserved stochastic volatility level, is considered with a view to simplifying the approximation results in Section 4. In Section 6, numerical results to the approximate solution to the problem are presented, and the behavior of the solution as well as the sensitivity analysis are discussed.