B0324
Title: Using quantile time series and historical simulation to forecast financial risk multiple steps ahead
Authors: Richard Gerlach - University of Sydney (Australia) [presenting]
Giuseppe Storti - University of Salerno (Italy)
Antonio Naimoli - University of Salerno (Italy)
Abstract: In financial time series, historical simulation is employed to standardize the financial return data distribution, allowing bootstrap methods to forecast quantities of interest, e.g. value at risk (VaR) and expected shortfall (ES), without assuming a parametric error distribution. Instead of standardizing by the volatility, in a semi-parametric quantile time series setting, the data is standardized by the estimated quantile time series, using the quantile model to allow bootstrapped single and multi-step ahead forecasts of VaR. It is further illustrated that the distribution of returns standardized by the quantile estimates can be bootstrapped to estimate and forecast ES (single and) multiple steps ahead. The methods can be applied to all time series settings where quantiles are directly modelled. A simulation study using the well-known CaViaR quantile time series model illustrates favourable performance, compared to standard GARCH-historical simulation, for volatility, VaR and ES forecasting. Empirical studies highlight the favourable performance of the methods applied to semi-parametric financial time series settings incorporating realized measures of volatility, e.g. Realized (E-)GARCH.