Essays on multivariate volatility models: an application to emerging financial markets

Le, Trung Thanh (2012). Essays on multivariate volatility models: an application to emerging financial markets. University of Birmingham. Ph.D.

Preview

Le12PhD.pdf
PDF - Accepted Version
Download (2MB)

Abstract

This thesis is an empirical study of how multivariate models can be applied to analyze the dependence between emerging financial markets and the US financial market. This thesis comprises of 3 complete papers which will use this data set as follows.

The first paper is an comparative research on estimations and evaluations of 54 individual volatility models which belong to 10 different model classes being the Riskmetrics models, the Constant model (CCC), the Orthogonal-GARCH model (O-GARCH), the Dynamic Conditional Correlation model (DCC), the Asymmetric DCC model (ADCC), the Consistent DCC model (CDCC) and the Student’s t-DCC model (TDCC). All of these models were estimated and then ranked by using both in-sample and out of sample performances. This research is to emphasize the importance of model selection in modeling the volatility of financial time series from emerging financial markets.

The second paper uses the TDCC model which performed relatively well among the 54 volatility of financial time series from emerging financial markets. The second paper uses the TDCC model which performed relatively well among the 54 volatility models to analyze the volatilities and correlations of the emerging markets. Specifically, the pair-wise conditional correlations between each of the emerging markets and the US market, generated by the TDCC model, were used to perform empirical tests for the contagion of the 3 recent financial crises which are the Dotcom crisis in 2000, the Sub-prime in 2007-2008 and the Global financial crisis in 2008-2009. The use of the TDCC model which assumes a Student’s t-distribution is greatly meaningful for the empirical tests for contagion as it deals with the fat-tailed behaviours of the financial data.

The third paper is the application of multivariate copula, which provides a connection between the univariate distributions and the multivariate distribution inside the DCC model, to analyze the emerging data. The flexibility of the copula model that separates the multivariate distribution assumption from those univariate series allows us to have an efficient examination of the dependence structure of emerging financial markets. Following success of the copula models in recent studies, our research, which is the first to use the copula model to analyze high-dimensional data, confirms a significant improvement of the copula from the standard DCC model.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):