Value-at-Risk calculations need a volatility estimate. The standard approach: use a proxy like the VIX, scale your forecast by its current value, and recalibrate. The more responsive the calibration is to the proxy, the better it tracks changing market conditions. Full proxy reliance seems optimal.
Zhong (arXiv:2603.22569) introduces a continuous parameter controlling how much the VaR recalibration depends on the volatility proxy, from minimal reliance to full scaling. The parameter makes explicit a design choice that most frameworks leave implicit.
The theoretical prediction is clean: high proxy reliance increases responsiveness to volatility changes. The empirical finding is messier. Across six ETFs using VIX-related proxies, full proxy reliance does not uniformly dominate. The advantage of the moderate or reduced approaches appears specifically during market stress — exactly when the proxy itself becomes least reliable, when VIX spikes don't cleanly map to the specific asset's tail risk.
The practical value is “improved stressed-state robustness rather than uniform overall dominance.” The moderate approach sacrifices responsiveness in normal times to avoid catastrophic dependence on a proxy that degrades under the conditions where accuracy matters most.
This is the general structure of proxy dependence in any system. The proxy is most useful when conditions are stable (because the mapping from proxy to target is well-calibrated) and least useful when conditions are extreme (because the calibration was learned from non-extreme data). The system that relies fully on the proxy inherits its failure mode. The system that partially ignores the proxy trades average performance for tail robustness — the same tradeoff that makes diversification work.