| Summary: | 博士 === 淡江大學 === 財務金融學系博士班 === 101 === The past twenty years have witnessed increasing nonlinear econometric methods used in the empirical study of financial markets. Despite financial theory commonly does not provide enough reason to support the use of the nonlinear models yet, the nonlinear tools sometimes can capture the phenomenon which cannot be explained by linear models. Therefore researchers should attach importance to these tools. This dissertation describes three types of recently-developed nonlinear econometric methods and display how these methods are applied to the studies of the financial markets. In particular, they can help researchers explore the dynamic causal relationship between financial variables.
This dissertation consists of three essays. The first essay is "Linear and Nonlinear Dynamics between U.S. and Japanese gold futures market". Using Diks and Panchenko (2006) non-parametric causality test to explore the linkage between the world''s top two gold futures markets, the study results show evidence of two-way causality. After controlling for market volatility with FIAGRCH models, only the nonlinear causality from U.S. to Japanese gold futures market is found. This suggests that the nonlinear causal from Japanese to U.S. gold futures market is almost a result of market volatility, but the volatility effect can only explain a part of nonlinear reverse causality.
The second essay is "Asymmetric and Nonlinear Dynamic Relationship between Taiwan Spot and Stock Index Futures ". This article uses the Mackey-Glass (MG) time-series model to construct the relationship between spot and futures, and then performs the based MG causality tests. The study results show existence of the bidirectional nonlinear causality, and that the nonlinear causal almost is driven by market volatility. Further results demonstrate asymmetric nonlinear causality under good and bad news. Under good news, the spot returns nonlinearly causality index futures returns, but under the bad news the case is reversal.
The topic of the third essay is "Regime Dependence between Stock Prices and Trading Volume in Taiwan". This article uses the Markov Switching Vector Autoregression (MS-VAR) model to construct the price-volume relation in the Taiwan stock market. Results show that the two-regime MS-VAR model is suitable to describe the price-volume relationship. Moreover, the regime-dependent Granger causality tests and regime-independent impulse response functions are performed to analyze price-volume relationship under different regimes. Results show that the causality from volume changes to stock returns only exists in the high volatility regime, while stock returns cause volume changes irrespective of regimes.
|
|---|
