Postoptimality for Scenario Based Financial Planning Models with an Application to Bond Portfolio Management

The contamination technique is presented as a numerically
tractable technique for postoptimality analysis and analysis of the
robustness of the optimal value of various scenario based stochastic
programs with respect to inclusion of additional "out-of-sample" scenarios.
Using results based on the initial selection of scenarios and those based
on the alternative out-of-sample scenarios it provides bounds for the
optimal value based on the pooled sample of scenarios of these groups.
The application of the method to models supporting financial decision
making is detailed for bond portfolio management and tracking models.
Numerical experience is presented for a bond portfolio management model
using data from the Italian bond market.