Bond portfolio management via stochastic programming

Stochastic programming is a tool to support bond portfolio management decisions.For a successful application of the stochastic programming methodology, one must choose an adequate model, asses its parameters, generate sensible input scenarios or scenario tree, solve the scenario-based problem using an optimization software and validate the results. Formulation of two-stage and three-stage models for bond portfolio optimization are discussed and computational results displayed. Comparisons of monthly and quarterly time discretizations and for various topologies of the input scenario trees are presented. Black-Derman-Toy binomial lattice calibrated from market data is used to generate scenarios and scenario reduction and scenario construction methods are applied using GAMS. The contamination technique is exploited to quantify the impact of including additonal (stress or out-of-sample) scenarios and/or additional stages to an already selected scenario tree.