An intraday-return-based Value-at-Risk model driven by dynamic conditional score with censored generalized Pareto distribution

Encouraged by the literary fact that high-frequency data such as intraday returns contribute to estimating the tail risk of daily returns, we propose an intraday-return-based Value-at-Risk (VaR) model driven by dynamic conditional score with censored generalized Pareto distribution (hence, Censored GP-DCS-VaR model), which is a novel parametric VaR approach based on dynamic score-driven model and can incorporate intraday information into daily VaR forecast. This model helps present the dynamic evolution of intraday return distribution and well capture its tail feature. Applying bootstrap or a parametric method, we are allowed to form the daily return distribution in light of intraday data and thus can calculate VaR directly. Empirical analysis using the data of the Chinese stock market shows that our model gain an advantage in the risk estimation of extreme returns, proved by the comparison of out-of-sample forecasts between the Censored GP-DCS-VaR and the realized-GARCH-VaR.