On a time consistency concept in risk averse multistage stochastic programming

In this paper we discuss time consistency of multi-stage risk averse stochastic programming problems. We approach the concept of time consistency from an optimization point of view. That is, at each state of the system optimality of a decision policy should not involve states which cannot happen in the future. We also discuss a relation of this concept of time consistency to deriving dynamic programming equations. Finally, we argue that some risk averse approaches to multi-stage programming are time consistent while some others are not.