This dissertation proposes a dynamic programming model for a fixed income optimization problem that involves multiple scenarios. The result is a model that can be used by investors that are actively managing a fixed income portfolio. The model was originally conceived to be applied to Mexican fixed income portfolio management, and the testing of the computational algorithm has been performed using portfolios of Mexican fixed income securities. However the flexibility of the model makes it applicable to any fixed income market.; The proposed model is a multi-period, multi-scenario, dynamic portfolio model that specifies a sequence of investment decisions over a finite planning period. It aims to compose a portfolio that achieves a target return under different interest-rate environments. The model is multi-period in that the horizon is divided into several decision periods. A multi-scenario approach is applied to account for interest rate risk. A detailed scenario analysis is performed in which the particular securities and the portfolio as a whole are examined over the multiple periods within the planning horizon under different interest rate environments. A “rolling horizon” approach allows the portfolio manager to decide on a strategy to be implemented today and revised in the subsequent periodic reviews, given the projections of interest rates current at those times. The dynamics of the portfolio result from the dependence of the results in one period upon the results obtained in the previous period.; The mathematical formulation of the model belongs to the class of dynamic programming models with Quadratic Criterion and Linear Dynamics. Both state and decision spaces at each stage are continuous and multidimensional. This model has been modified to include several scenarios. The mathematical model is also altered to include barrier functions that enforce sign restrictions on specified variables. These barrier functions are applied only to the first stage of the horizon, since the portfolio is expected to be rebalanced in each subsequent period.; The results of the computational tests demonstrate that the model can provide reasonable and attainable results given the inputs provided. The resulting algorithm permits to reach not only the target return but the highest return.
展开▼
