Abstract/Summary

The thesis investigates topics on how to improve the estimation and forecasting for market risk measures (focusing on Value at Risk and Expected Shortfall, denoted by VaR and ES, jointly) by building superior models with extra information and by detecting structural changes in risk models (in a retrospective manner and a real-time manner). The first contribution is introducing a new framework by incorporating intraday information into dynamic semiparametric models to forecast VaR and ES. We consider the intraday measures including the realized variance and overnight returns. In the practical application, we apply the proposed models to international stock market indices, then evaluate the forecasting performance via various back-tests. Our results show that our models outperform the benchmarks consistently across all indices and various significance levels. Secondly, this thesis develops a test that can efficiently captures change points in the (VaR, ES) estimated by (semi)parametric models. We derive the asymp-totic distribution of the test statistic and adopt a stationary bootstrapping technique to obtain the p-values of the test statistic. Monte Carlo simulation results show that our proposed test has better size control and higher power than the alternative tests. An empirical study of risk measures based on the S&P 500 index illustrates that our proposed test can detect change points associated with well-known market events. The third main contribution is proposing a sequential monitoring method to detect changes in semiparametric risk models for (VaR, ES). We derive the asymptotic theorem for the monitoring scheme under the null hypothesis. Our Monte Carlo simulations with finite sample sizes show that this test has reasonable size control under the null hypothesis and high power under alternative hypotheses. Empirical applications based on the S&P 500 index and the GBP/EUR exchange rate illustrate that the detected change points often precede the actual market crashes.