Do Long-memory GARCH-type-Value-at-Risk Models Outperform None-and Semi-parametric Value-at-Risk Models?

As a result of the 2007-2008 global financial crisis, traditional value-at-risk (VaR) models used to measure the market risk have been criticised for their inaccuracy. Therefore, alternative models such as long-memory GARCH-type based VaR models have been receiving increased attention in recent literature. In this regard, this study compares the one-day-ahead out-of-sample VaR forecasting performances of FIGARCH, HYGARCH, and FIAPARCH models under normal, student t, and skewed student t distribution assumptions with FHS and HS model performances, which are the most commonly applied models especially by commercial banks in practice, for eight different financial variables including energy commodities (West intermediate crude oil (WTI) and New York Harbour Conventional Gasoline regular (NYHCGR)), stock indices (NIKKEI 225 stock market index and TSEC weighted stock index), foreign exchange rates (Euro/US Dollar (EUR/USD) and Japanese Yen/USD (JPY/USD)), and precious metals (gold and copper). Results clearly show that the FHS model is the most appropriate model for long trading positions, to which the relevant literature has paid more attention, whereas for short trading positions the HYGARCH model under skewed student t distribution assumption should be preferred.