Volatility persistence across global equity markets: long-memory diagnostics versus horizon-specific VaR rankings

Lin, Zhixing, Lin, Shuoyan
Risk Management · Springer Nature · 2026-10-23 · DOI: 10.1057/s41283-026-00247-2

Аннотация

Statistical forecast loss and regulatory tail coverage can rank volatility models differently. We study 19 equity indices from 2000 to 2025 (seven developed and twelve emerging markets) under one univariate daily-close rolling design. We score three objects on the same origins: one-step QLIKE, the 95% Kupiec–Christoffersen dual test, and 99% Value-at-Risk coverage at the 10-day horizon used in FRTB internal-model backtesting ( $$h=10$$ ). GJR-GARCH records the lowest QLIKE in 16–19 indices. At that horizon, exponentially weighted moving average and filtered historical simulation lead the 95% dual test in developed markets, whereas GJR-GARCH leads the 99% joint test. Both rankings move with the estimation window, and pass rates are descriptive shares on this sample.

Читать у издателя →

Похожие статьи