EVT-Based Rate-Preserving Distributional Robustness for Tail Risk Functionals

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Deo, Anand
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914270145413120
author Deo, Anand
author_facet Deo, Anand
contents Risk measures such as Conditional Value-at-Risk (CVaR) focus on extreme losses, where scarce tail data makes model error unavoidable. To hedge misspecification, one evaluates worst-case tail risk over an ambiguity set. Using Extreme Value Theory (EVT), we derive first-order asymptotics for worst-case tail risk for a broad class of tail-risk measures under standard ambiguity sets, including Wasserstein balls and $ϕ$-divergence neighborhoods. We show that robustification can alter the nominal tail asymptotic scaling as the tail level $β\to0$, leading to excess risk inflation. Motivated by this diagnostic, we propose a tail-calibrated ambiguity design that preserves the nominal tail asymptotic scaling while still guarding against misspecification. Under standard domain of attraction assumptions, we prove that the resulting worst-case risk preserves the baseline first-order scaling as $β\to0$, uniformly over key tuning parameters, and that a plug-in implementation based on consistent tail-index estimation inherits these guarantees. Synthetic and real-data experiments show that the proposed design avoids the severe inflation often induced by standard ambiguity sets.
format Preprint
id arxiv_https___arxiv_org_abs_2506_16230
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle EVT-Based Rate-Preserving Distributional Robustness for Tail Risk Functionals
Deo, Anand
Risk Management
Probability
Methodology
Machine Learning
Risk measures such as Conditional Value-at-Risk (CVaR) focus on extreme losses, where scarce tail data makes model error unavoidable. To hedge misspecification, one evaluates worst-case tail risk over an ambiguity set. Using Extreme Value Theory (EVT), we derive first-order asymptotics for worst-case tail risk for a broad class of tail-risk measures under standard ambiguity sets, including Wasserstein balls and $ϕ$-divergence neighborhoods. We show that robustification can alter the nominal tail asymptotic scaling as the tail level $β\to0$, leading to excess risk inflation. Motivated by this diagnostic, we propose a tail-calibrated ambiguity design that preserves the nominal tail asymptotic scaling while still guarding against misspecification. Under standard domain of attraction assumptions, we prove that the resulting worst-case risk preserves the baseline first-order scaling as $β\to0$, uniformly over key tuning parameters, and that a plug-in implementation based on consistent tail-index estimation inherits these guarantees. Synthetic and real-data experiments show that the proposed design avoids the severe inflation often induced by standard ambiguity sets.
title EVT-Based Rate-Preserving Distributional Robustness for Tail Risk Functionals
topic Risk Management
Probability
Methodology
Machine Learning
url https://arxiv.org/abs/2506.16230