Journal of Economics and Management Volume 16, No. 2 September, 2020

Identifying the Dynamic Relationships among Four Pacific Rim Stock Markets by Bayesian Variable Selection

Feng-Chi Liu

Department of Statistics, Feng Chia University, Taiwan

Cheng-Lin Tsai

Department of Statistics, Feng Chia University, Taiwan

Abstract

With the increasing frequency of international trade and cross-border financial flows, the relationships among financial markets have become increasingly close. This research thus employs a vector autoregressive (VAR) approach to assess the dynamic correlations of four Pacific Rim stock index returns and considers the identification problem of VAR models. We treat the identification problem of the VAR by considering restrictions on the VAR coefficients and adopt a Bayesian variable selection method to simultaneously estimate the model parameters and identify the possible subsets of variables. For the purpose of finding possible subsets of variables, we propose a coding method and a visualized approach. For illustration purposes, we consider the dynamic relationships among the returns of four stock market indices:  S&P 500 of the U.S., Hang Seng Index of Hong Kong, Nikkei 225 of Japan, and Taiwan Capitalization Weighted Stock Index of Taiwan. We further employ three time periods of datasets to investigate the dynamic changes of the relationships among the four indices and find that there exists more complicated relationships among four indices during the global financial crisis of 2007-2009 and the period from 2016 to 2019 versus the relationships during the period from 2011 to 2015.

Keywords：Vector autoregressive model, Stochastic search variable selection, Bayesian inference, Markov chain Monte Carlo sampling.

JEL Classifications：C11, C58.

BACK