Risk measures based on weak optimal transport

Michael Kupper,Max Nendel,Alessandro Sgarabottolo

2023-12-11

Abstract:In this paper, we study convex risk measures with weak optimal transport penalties. In a first step, we show that these risk measures allow for an explicit representation via a nonlinear transform of the loss function. In a second step, we discuss computational aspects related to the nonlinear transform as well as approximations of the risk measures using, for example, neural networks. Our setup comprises a variety of examples, such as classical optimal transport penalties, parametric families of models, uncertainty on path spaces, moment constrains, and martingale constraints. In a last step, we show how to use the theoretical results for the numerical computation of worst-case losses in an insurance context and no-arbitrage prices of European contingent claims after quoted maturities in a model-free setting.

Mathematical Finance,Risk Management

What problem does this paper attempt to address?

This paper investigates convex risk measures based on weakly optimal transport. In financial modeling, describing the distribution of underlying risk factors often involves model uncertainty. The paper proposes a convex risk measure approach using weakly optimal transport cost as a penalty function, allowing for the consideration of model uncertainty in evaluating financial portfolios. The weakly optimal transport problem was initially introduced in [20, 1] as a way to measure the distance between probability distributions, not limited to standard optimal transport problems.

The paper first demonstrates that this risk measure can be represented through a nonlinear transformation of the loss function, which is independent of the reference measure. Then, it discusses how to solve these measures more specifically when the extreme point set is known, particularly when the cost function is associated with certain special extreme point constraints. Furthermore, the paper covers various examples from classical optimal transport penalties to uncertainty in path space, and proposes numerical methods for computing worst-case losses and no-arbitrage prices in insurance contexts, especially in European option pricing.

In the finance and insurance industries, this method can be used to handle model uncertainty, estimate risks, and calculate no-arbitrage price ranges. Through neural network approximation, the paper provides an efficient tool for computing risk measures, especially when the reference model changes and risk needs to be reassessed quickly.