Parallel Multistage Stochastic Programming Model Used in Portfolio Management

Problems of portfolio management can be viewed as multi-period dynamic decision problems. We present a model for allocation of financial resources to bond indices in different currencies. The future uncertainty is represented by stochastic scenario trees, where stochastic properties of possible future development of exchange rates and interest rates are incorporated. The objective is to maximize the expected value of the portfolio at a defined time horizon. Mathematically such models lead to multistage stochastic programming. The prob-abilistic scenario trees are ”bushy ” and the number of scenarios grows exponen-tially with the number of stages. We have developed and compared two-stage and three-stage models that were tested by using historical data of interest rates and exchange rates. The portfolios were composed of 4 indices of government bonds in different currencies (USD, EUR, GBP and CHF). The scenarios of future de-velopment of interest rates and exchange rates were generated using Monte Carlo simulations