Tail dependence networks of global stock markets

• Author: Fenghua Wen,Xin Yang,Wei‐Xing Zhou
• Published date: 01 January 2019
• Date: 01 January 2019
• DOI: http://doi.org/10.1002/ijfe.1679

Received: 27 May 2018 Revised: 17 July 2018 Accepted: 10 September 2018

DOI: 10.1002/ijfe.1679

RESEARCH ARTICLE

Tail dependence networks of global stock markets

Fenghua Wen1,2 Xin Yang3Wei-Xing Zhou4,5

1School of Business, Central South

University, Changsha, China

2Supply Chain Management and Logistics

Optimization Research Centre, Faculty of

Engineering, University of Windsor,

Windsor, ON,Canada

3School of Mathematics and Statistic,

Changsha University of Science and

Technology,Changsha, China

4Department of Finance, East China

University of Science and Technology,

Shanghai, China

5Department of Mathematics, East China

University of Science and Technology,

Shanghai, China

Correspondence

Wei-XingZhou, East China University of

Science and Technology,130 Meilong

Road, P.O.Box 114, Shanghai 200237,

China.

Email: wxzhou@ecust.edu.cn

Funding information

Fundamental Research Funds for the

Central Universities, Grant/Award

Number: 222201718006; National Natural

Science Foundation of China,

Grant/AwardNumber: 71873146,

71873147, 71532009 and 71431008

Abstract

The Pearson correlation coefficient is used by many researchers to construct

complex financial networks. However, it is difficult to capture the structural

characteristics of financial markets that have extremefluctuations. To solve this

problem, we resort to tail dependence networks. We first build the edge infor-

mation of the stock network by adopting Pearson's correlation coefficient and

the symmetrized Joe–Clayton copula model, respectively. By using the planar

maximally filtered graph method, we filter the edge information, obtain Pear-

son's correlation coefficient and tail dependence network, and compare their

efficiencies. The community structure of the constructed networks is investi-

gated. We find that the global efficiency of tail-dependent networks is higher

than that of the Pearson correlation networks. Further analysis of the nodes in

the upper- and lower-tail dependence networks reveals that the European mar-

kets are more influential than Asian and African markets during a booming

market and a recession market.In addition, different cliques are found in the two

tail dependence networks. The finding indicates that financial risks will impact

geographically adjacent markets.

KEYWORDS

community structure, complex network, Pearson's correlationcoefficient, SJC copula, stock market

1INTRODUCTION

With the rapid development of international trade and the

mass flow of capital and labour forces, the economic ties

among countries are becoming ever closer than before. The

global economy,which is a large complex system, sets each

country in its system, and any changes in the subsystem

(e.g., the stock market) will have an impact on the entire

system. Many scholars have demonstrated that the theory

of complex network is a powerful tool to cope with com-

plex systems. Therefore, we can adopt the complex net-

work theory to portray the stock market network system

(Boccaletti, Latora, Moreno, Chavez, & Hwang, 2006;

Buldyrev, Parshani, Paul, Stanley, & Havlin, 2010; Costa

et al., 2011; Miccichè, Bonanno, Lillo, & Mantegna, 2003;

Newman, 2003).

In previous studies, many scholars calculated the Pear-

son correlation coefficient to obtain the edge information

of the stock network, applying the complex network the-

ory. Mantegna (1999) was the pioneer to take the Pearson

correlation coefficients as the S&P 500 stock edge informa-

tion to build the network, and he had also analysed the

clustering level. Since then, numerous similar researches

applied this method. Kim, Lee, Kahng, and Kim (2002)

558 © 2018 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/ijfe IntJ Fin Econ. 2019;24:558–567.