Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.498914
Title: A study of stock volatility in the context of factor volatility models for large datasets : factor analysis and forecasting
Author: Lui, Sze Wai Silvia
Awarding Body: Queen Mary, University of London
Current Institution: Queen Mary, University of London
Date of Award: 2007
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
This thesis is a study of stock volatility adopting two factor volatility models for large datasets: the orthogonal GARCH model and the stochastic volatility factor model. An application is made to the constituent stocks of five Asian indexes. Factor analysis and volatility forecasting exercise are carried out. Chapter 1 is an empirical application of the orthogonal GARCH model to Asian stock returns. Correlation analysis, eigenvalue and eigenvector analysis and several diagnostic tests are carried out. Our results show that using large number of principal components cannot guarantee an improvement in capturing dynamics. Moreover, GARCH(l,l) is the appropriate specification for the principal components of stock returns of some datasets. An empirical example of how GARCH analysis of all series in the entire dataset can be summarised by a univariate GARCH analysis of the first principal component is also provided. Chapter 2 is a factor analysis using stochastic volatility factor models. In contrast to the first part of the study, common factors are estimated from large datasets of Asian stock volatilities via principal components. Correlation analysis of stock volatilities is performed. Examinations of the dynamics of factor estimates and their explanatory power are also carried out. Our results 3 confirm that large dataset with many cross-sectional series from the same category may not always be desirable for factor analysis. Evidence of long memory is found in the first principal component of some datasets but not all of them. Chapter 3 is a volatility forecasting exercise. In-sample analysis is implemented using the stochastic volatility factor models and the orthogonal GARCH model. Moreover, we propose an extension of a local-factor model to a multi-factor model. Testing of factor significance is scrutinised. A comparison of forecasting performance shows that the stochastic volatility factor models outperform the orthogonal GARCH model in forecasting Asian volatilities.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.498914  DOI: Not available
Keywords: Economics
Share: