A Stochastic Programming Approach For Multi-Period Portfolio Optimization With Transaction Costs

This paper uses stochastic programming to solve multi-period investment problems. We combine the feature of asset return predictability with practically relevant constraints arising in a multi-period investment context. The objective is to maximize the expected utility of the returns the periods to balance the liabilities. Asset returns and state variables follow a first-order vector auto-regression and the associated uncertainty is described by discrete scenario trees. To deal with the long time intervals involved in multi-period problems, we consider short-term decisions, and incorporate a solution for the long, subsequent steady-state period to account for end effects.