Risk measures and progressive enlargement of filtration: a BSDE approach

Abstract

We consider dynamic risk measures induced by Backward Stochastic Differential Equations (BSDEs) in enlargement of filtration setting. On a fixed probability space, we are given a standard Brownian motion and a pair of random variables $(τ, ζ) \in (0,+\infty) \times E$, with $E \subset \mathbb{R}^m$, that enlarge the reference filtration, i.e., the one generated by the Brownian motion. These random variables can be interpreted financially as a default time and an associated mark. After introducing a BSDE driven by the Brownian motion and the random measure associated to $(τ, ζ)$, we define the dynamic risk measure $(ρ_t)_{t \in [0,T]}$, for a fixed time $T > 0$, induced by its solution. We prove that $(ρ_t)_{t \in [0,T]}$ can be decomposed in a pair of risk measures, acting before and after $τ$ and we characterize its properties giving suitable assumptions on the driver of the BSDE. Furthermore, we prove an inequality satisfied by the penalty term associated to the robust representation of $(ρ_t)_{t \in [0,T]}$ and we discuss the dynamic entropic risk measure case, providing examples where it is possible to write explicitly its decomposition and simulate it numerically.