Time consistent convex Feller processes and non linear second order partial differential equations

Abstract

This paper shows how the theory of dynamic risk measures provides viscosity solutions to a family of second-order parabolic partial differential equations, even in the degenerate case. First, motivated by the martingale problem approach of Stroock and Varadhan, we make use of the time consistency characterization for dynamic risk measures, to construct time consistent convex Markov processes. This is done in a general setting in which compacity arguments cannot be used, and for which there does not always exist an optimal control. Second, we prove that these processes lead to viscosity solutions for semi linear second-order partial differential equations with convex generator. Finally we give an application to mathematical finance. We show that our results allow for the construction of a No Arbitrage Pricing Procedure in the context of stochastic volatility. Within this approach, convexity takes into account liquidy risk.