Risk assessment for uncertain cash flows: model ambiguity, discounting ambiguity, and the role of bubbles (Q693031)

The paper studies risk assessment of random cashflows arising in a discrete time setup under both model and discounting uncertainty. Following \textit{P. Cheridito, F. Delbaen} and \textit{M. Kupper} [Electron. J. Probab. 11, Paper No. 3, 57--106 (2006; Zbl 1184.91109)], the authors consider conditional convex risk measures for processes but propose to think of them as conditional convex risk measures for random variables on \(\bar{\Omega}\times \mathbb{T}\) endowed with the optional \(\sigma\)--field \(\bar{\mathcal{F}}\). They then show that an absolutely continuous probability measure \(\bar{Q}\) on \(\bar{\mathcal{F}}\) can be represented as \(\bar{Q}=Q\times D\) for a locally absolutely continuous probability measure \(Q\) on the original space and a predictable discount factor \(D\). This leads directly, following classical results, to a robust representation of the risk measures in (3.13) as a supremum of suitably penalised expectations of discounted cashflows. This representation makes explicit the role of model uncertainty and discounting uncertainty. The authors then use the above representation to study time consistency focusing on a supermartingale characterisation. They derive Doob and Riesz decomposition of the penalty process. Appearance of ``bubbles'' in the latter is linked to asymptotic properties. Finally, the authors discuss cash subadditivity and the issues of calibration. As expected, if enough assets (zero coupon bounds) trade to specify uniquely the term structure then the discounting ambiguity disappears.