Reliability-Aware ETF Tail-Risk Monitoring

Abstract

Daily ETF risk monitoring can become unreliable when market data quality degrades, market conditions shift, or predictive performance becomes unstable. This paper develops a reliability-aware risk monitoring service for next-day tail-risk surveillance. The proposed framework combines service-time quality checks, lower-tail prediction, uncertainty scoring, and risk-aware adjustment of the tail-risk estimate. We evaluate the system on a daily panel of multiple ETFs augmented with VIX and yield-curve information under a rolling walk-forward design. Empirically, the framework improves tail-risk monitoring, especially during stressed periods, while remaining reliable under simulated input degradation.

arXiv comment6 pages. Accepted as a regular paper at IEEE SMC 2026