A Study of Value-At-Risk Models and Their Prediction Power

This dissertation, "A Study of Value-at-risk Models and Their Prediction Power" by Gang, Li, 李剛, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of thesis entitled A Study of Value-at-Risk Models and Their Prediction Power Submitted By Li Gang For the Degree of Master of Philosophy at the University of Hong Kong in August 2005 Abstract Market risk refers to the risks of financial losses arising from adverse movements in market prices. The interest in managing market risk has grown during the last two decades among financial institutions and other market participants. A market risk measurement called Value-at-Risk (VaR) has been adopted as the most important measure of market risk. Despite its recent advent, value at risk (VaR) became the most widely used technique for measuring future expected risk for both financial and non-financial institutions. VaR, the measure of the worst expected loss over a given horizon at a given confidence level, depends crucially on the distributional aspects of portfolio returns. Existing VaR models do not capture adequately some empirical aspects of financial data such as the tail thickness, which is vital in VaR calculations. Tail thickness in financial variables results basically from stochastic volatility. Those two sources are not totally separated; under event risk, volatility updates faster than under normal market conditions. There are several alternative techniques to estimate VaR measures, categorized as parametric and non-parametric methods. However, accurate estimation of daily volatility as well as the data availability limitation make the traditional VaR measurements greatly challenged. Besides, most of these models do not include volatility updating feature, which make the VaR estimation not quite accurate. In this dissertation, I shall focus on a powerful model - stochastic volatility model (SVM). This model can solve the above deficiency. I shall state possible outcomes and test validation and efficiency of the model. Besides, I shall also compare its predicted VaR with other commonly used VaR forecast models. To estimate the corresponding parameters in the stochastic volatility model, I shall use Gibb Sampling approach and Particle Filter in the analysis. Bayesian analysis is becoming a more and more popular tool in statistical inference nowadays. It is widely used in quantitative finance area. After obtaining parameter estimation, I shall use simulated and empirical financial data to test the validity and efficiency of this model. Finally, I shall compare the model with other commonly used VaR models. The SVM is expected to perform significantly better in predicting future VaR than other model. DOI: 10.5353/th_b3202971 Subjects: Nonparametric statistics Risk - Econometric models