Insights on Time-consistent Deep Hedging under Elicitable Dynamic Risk Measures

Abstract

We study deep hedging in the context of dynamics risk measures, where sequential decisions are time-consistent. Whereas the literature in such context mainly considers low-dimensional problems with simple environment dynamics, we tackle the high-dimensional problem of basket option hedging; we show that the approach is feasible and can be used conveniently in the presence of more complex state spaces. We rely on the conditional elicitability of spectral risk measures to represent the optimization objective. We provide insights on how the choice of scoring function impacts the training of the hedging agent. Lastly, the time-consistent hedging strategies are benchmark against deep hedging approaches relying on static risk measures leading to precommitment.