Applications of dynamic conditional correlation based models to financial and commodity asset data

Abstract

The ongoing global crises like the US-China trade and COVID-19 pandemic are the phenomena caused significant impact on several countries. To understand the economic phenomena on the investment and financial market, we take advantages by using the theories that explain it. However, the conventional concepts of correlation which it explained relationship between variables limited to a pair of variables and the correlation assumed to be constant, those assumptions are not useful to help understand the economic phenomena which it has more complex due to the more integration of financial markets. We want to investigate the impact of the phenomena and provide the fact on contagion effect which occurred and affected economies and financial markets differently due to different contexts. The notions lying above have formed a motif for the development of this academic work. The important goal lies in providing alternative ways of estimation techniques for dynamic conditional correlation approach (DCC) with one-step estimation to replace the two-step estimation of the conventional DCC approach, and more flexible approaches to provide better understanding on the contagion effect in financial markets in the context of trade war and COVID-19. We estimate the DCC parameters with one-step estimation to avoid the problem of biased and inconsistent estimators. Moreover, we extend the conventional DCC with Markov-switching (MS) and Copula into our study to provide more flexible and accurate model. The MS-DCC-GARCH with Jumps helps understand the characteristic of data which could have different explained parameters due to the different state of regime, namely low and high volatility regimes. While the Copula- based DCC-GARCH could benefits our study on the tail-dependence which financial data usually persists of fat-tailed distribution, replacing the normal and student's t distributions in the conventional DCC model, and GARCH with Jumps model has ability to capture the extreme case values which the extreme cases usually occur in financial time-seiries. Moreover, we also develop the Copula-based MS-DCC-GARCH model which provided more flexible with by having both MS and Copula concepts. To the best of my knowledge, the estimation of Copula-based MS-DCC models is new to literature, and thus this academic study is the first ever employed with the approach. This thesis suggests applying the DCC approaches including, the Copula-based DCC- GARCH, MS-DCC-GARCH, and Copula-based MS-DCC approaches, to be alternative ways of dynamic conditional correlation due to the evidence that those models outperform the conventional DCC approach. Those approaches could be alternative approaches when researchers want to investigate the dynamic conditional correlation. For a good direction of research, the one-step estimation is suggested instead of using the biased and inconsistent two-step estimation.