Online Estimation and Optimization of Utility-Based Shortfall Risk
Mathematics of Operations Research, INFORMS — 2023 · 48(4), 2444–2470
Utility-Based Shortfall Risk (UBSR) is a risk metric increasingly popular in financial applications due to properties it enjoys over classical VaR/CVaR measures. We consider estimating UBSR in a recursive online setting, casting the problem as root-finding and proposing stochastic approximation schemes. We derive non-asymptotic bounds on estimation error and extend to UBSR optimization over a parameterized class of random variables using stochastic gradient descent, with convergence guarantees.