Dynamic Conditional Correlation Model-an Analysis Between Taiwan and US Stock Market

資產與資產之間的連動性在風險管理上，已扮演了愈來愈不可或缺的角色。這篇研究利用了Engle所提出的動態條件相關係數模型，估計台股與美股1998年至2006年共九年每天的動態相關係數。當利用不同參數估計時，此模型顯示相當大的敏感度-不同參數下的相關係數大小估計值相差甚多。對此一序列做了相關係數為常數的檢定後，接受了此一序列的相關係數為固定。然而九年中的相關係數皆為一固定常數不符合一般經濟直覺，因此將此九年的序列分割成九份序列，分別估計各別的動態相關係數。除了當期的相關係數外，交錯相關係數更在風險管理上提供一個更全面性的分析，而實證顯示美股落後台股一期的相關係數為顯著。最後做了一個最小變異數的投資組合，並比較不同相關係數估計方法下所得到的投資績效，結果顯示在此台股與美股的分析中，固定相關係數在配置最小變異數的投資組合中即已足夠。The correlation between different assets is gaining more and more importance toward the area of risk management. In this research, the dynamic conditional correlation model proposed by Engle is used to conduct an empirical analysis between the NASDSQ and the Taiwan stock market. The DCC model shows strong sensitivity when using different correlation parameters. The sample is accepted under the test of constant, but a nine-year constant correlation is economically insensible, so the whole sample of nine-year has been divided into nine sub-periods for model estimation. Aside from instantaneous correlation, the cross correlation also plays a role in risk management and thus the cross correlation function is conducted. A minimum variance portfolio was built to measure the performance based on different correlation estimates, the results show that a constant correlation based performance is even better than a dynamic correlation based performance.Contents Chapter 1.Introduction 9 1.1 Motivation 9 1.2 Literature Review 9 1.3 Research Structure 11 Chapter 2.Methodology 12 2.1 The Dynamic and Constant Correlation Model 12 2.2 Testing For Constant Correlation 16 2.3 Specification Testing 17 2.4 Forecasting 18 2.5 Cross Correlation 19 Chapter 3.Empirical Analysis 20 3.1 Data Statistics 21 3.2.Model Estimation 23 3.3.Forecasting 38 3.4 Cross Correlation Function 40 Chapter 4.Correlation Estimates for Assets Allocation 46 Chapter 5.Conclusion 48 5.1 Conclusion 49 5.2 Further Research 4