Click to view a PDF (Adobe Acrobat PDF) file.
January 2003
In this paper, a notion of risk measure is defined for dynamic models. Three axioms, coherence, relevance and dynamic consistence, are postulated. It is shown that every dynamic risk measure that satisfies the axioms can be represented as the maximal expected present value of future losses where expectations are taken with respect to a set of probability measures. As new information arrives, this set of probability measures is updated in the Bayesian way. Moreover, dynamic consistency implies that this set satisfies a certain consistency condition.