On the Fortet-Mourier metric for the stability of Stochastic Optimization Problems, an example

Abstract

We consider the use of the Fortet-Mourier metric between two probability measures to bound the error term made by an approximated solution of a stochastic program. After a short analysis of usual stability arguments, we propose a simple example of stochastic program which enlightens the importance of the information structure. As a conclusion, we underline the need to take into account both the probability measure and the information structure in the discretization of a stochastic program.