Stochastic Programming Approach For Dynamic Portfolio Selection Of Pension Funds

This dissertation is devoted to dynamic portfolio selection of pension funds based on stochastic programming.First, we show classical stochastic programming models for dynamic portfolio selection and general formulation is given. We found the stochastic programming approach has more flexibility because realistic considerations can be incorporated.Then, we introduce scenario generation and scenario tree construction based on VAR models. When a stochastic programming model is applied, scenario analysis is very importance, so we focus on algorithm of random sampling using MC Simulation.As an application of multi-stage stochastic programming, we propose a dynamic model for pension funds portfolio selection. We focus on the multi-period optimal investment allocation in the sense of minimizing the CVaR function. In order to control the risk better, we develop the objective function penalizes CVaR through the whole horizon. We generate the future scenarios of financial market which are illustrated via applying the VAR method. Furthermore, we obtain the simulate solution of the optimal strategies combining the history data. Finally, we analyze the sensitivity of the parameter to the optimal portfolio strategies.At last, policy advices about management of pension funds investment are given.