Citation

Aridi, Nor Azliana (2023) High frequency extreme value theory and market risk analysis for global stock markets. PhD thesis, Multimedia University.

Full text not available from this repository.

Abstract

These studies aim to assess whether the availability of high-frequency data enhances the accuracy of market volatility estimation when compared to lowfrequency data. In this study, an empirical analysis has employed the Heterogeneous Market Hypothesis (HMH), an extension of the Efficient Market Hypothesis (EMH), as a framework to forecast stock market risk. The data used for analysis comprised the daily closing stock prices and 5-minute stock prices of Dow DJIA, FTSE100, BVSP, and MERVAL Index from 2014 to 2022. The initial phase of the analysis involved examining the stylized facts present in the daily and intraday returns. The subsequent stage of the analysis focused on an in-sample empirical analysis to compare the performance of low-frequency volatility models, based on standard GARCH specifications, with high-frequency volatility models, specifically realized volatility (RV) specification and Heterogeneous Autoregressive (HAR) model. Finally, subsequent analysis was done to compare the performance of the two stage hybrid methods that combined the standard GARCH, RV and HAR models with the EVT to see whether this hybrid method can improve the volatility forecasting. To assess the accuracy of the VaR forecasts, out-of-sample testing was conducted using the unconditional coverage (UC), conditional coverage (CC), and dynamic quantile (DQ) tests. The backtesting of VaR also incorporated the utilization of both the regulatory loss function (RLF) and the firm's loss function (FLF). Moreover, the backtesting of ES in this study utilized the generalized breach indicator (GBI) method. This research findings revealed that, during financial crisis periods, high-frequency data provides investors with more information than low-frequency data. The findings also suggest that high-frequency models outperform low-frequency models in predicting stock market risk. Specifically, the study found that the HAR model was the most accurate in forecasting market risk, both during normal and financial crisis periods. Furthermore, the analysis emphasized that high-frequency conditional EVT models can outperform their low-frequency counterparts in risk forecasting by capturing shortterm market fluctuations and providing more timely information on extreme events.

Item Type:	Thesis (PhD)
Additional Information:	Call No.: QA273.6 .N67 2023
Uncontrolled Keywords:	Extreme value theory
Subjects:	Q Science > QA Mathematics Q Science > QA Mathematics > QA273-280 Probabilities. Mathematical statistics
Divisions:	Faculty of Management (FOM)
Depositing User:	Ms Nurul Iqtiani Ahmad
Date Deposited:	01 Oct 2025 07:25
Last Modified:	01 Oct 2025 07:25
URII:	http://shdl.mmu.edu.my/id/eprint/14649

Downloads

Edit (login required)