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 conservative risk adjustment. We evaluate the system on a daily panel of multiple ETFs augmented with VIX and yield-curve information under a rolling walk-forward design. The empirical results suggest that the proposed framework improves tail-risk reliability during stressed periods while preserving full evaluable coverage. Under deliberate input corruption, the quality layer enhances service robustness.