Testing for financial contagion based on a nonparametric measure of the cross‐market correlation
DOI | http://doi.org/10.1016/j.rfe.2014.05.001 |
Published date | 01 September 2014 |
Date | 01 September 2014 |
Author | Fuchun Li,Hui Zhu |
Testing for financial contagion based on a nonparametric measure of the
cross-market correlation☆
Fuchun Li
a,
⁎, Hui Zhu
b
a
Bank of Canada,Financial StabilityDepartment, Ottawa, ON K1A 0G9,Canada
b
Universityof Ontario Instituteof Technology, Faculty of Businessand Information Technology,Oshawa, ON L1H 7K4, Canada
abstractarticle info
Availableonline 9 May 2014
Keywords:
Financialcontagion
Financialcrisis
Nonparametricmeasure of the cross-market
correlation
Monte Carlosimulation
When contagion is defined as a significantincrease in market comovement after a shock to one country, we propose
atestforfinancialcontagion based on a nonparametric measure of the cross-market correlation. Monte Carlo sim-
ulation studies show that our test has reasonable size and good power to detect financial contagion, and that Forbes
and Rigobon's test (2002) is relativelyconservative, indicatin gtha t their test tends not to find evidence of contagion
when it does exist. Applying our test to investigatecontagion from the 1 997 East Asian crisisa nd the 2007 Subprime
crisis, we find that there existed international financial contagion from the two financial crises.
© 2014 Publishedby Elsevier Inc.
1. Introduction
Since 1987, international financialmarkets have experienced a se-
ries of financial crises such as the U.S. stock market crashin 1987, the
Mexican peso crisis in 1994, the East Asian crisis in 1997, the Russian
crisis in 1998, and the 2007 subprime crisis. A common characteristic
of these financial crises is that dram atic movements in the financial
market of a crisis country, such as large drops in asset prices a nd in-
creases in market volatility, can quickly spread to othermarkets with
differentsizes and structures acrossthe world. This leads many econo-
mists to raise the question of wheth er the high cross market
comovementsprovide empirical evidenceof contagion.
To answer this question, we need define contagion first. In this paper,
we adopt the definition of contagion introduced by Forbes and Rigobon
(2002), who define contagion as a significant increase in the cross-
market linkages after a shock to one country or group of countries.
1
Ac-
cording to this definition, contagion occurs only if cross-market linkage
increases significantly after the shock. Given the definition of contagion
above, the most common strategy of testing for contagion is to use
cross-market Pearson correlation coefficientas the measure of cross-
market linkage.
2
If there exists a significant increasein the correlation
coefficient after a shock, this suggests that the transmission mechanism
between the two markets increases after the shock and contagion occurs.
However, using a linear framewor k, Forbes and Rigobon (2002)
show that an increase in cross-market correlation coefficientsaround
crises maynot necessarily indicatecontagion due to econometric prob-
lems associated with heteroskedasticity,which can cause cross-market
correlationsto increase after a crisis, even if there is no increasein the
underlyingcorrelations. Consequently, Forbesand Rigobon (2002) sug-
gest one method of correctingfor this heteroskedasticity by adjusting
cross-market correlation coefficients. When the adj usted correlation
coefficient is used to test for contagion, they find no contagionduring
the 1997 East Asiancrisis. Instead, a high levelof cross-market correla-
tion coefficientafter a crisis onlyreflects a continuation of strongcross-
market linkages. Their conclus ion is that there is no contagion, only
interdependence.
Obviously, this adjustment is based on the assumptions that there
are no omitted variables and endogeneity,and the analysis of correla-
tion is limited to the caseof bivariate normal distributionbetween the
two markets. However, an increase in as set price correlations could
occurdue to changes in omitted variables,such as economic fundamen-
tals, risk perceptions, andpreference, even if contagion is not present.
Even though the correlation coefficientcan indicate the strengthof a lin-
ear relationship betweentwo variables, it may not be sufficientto eval-
uate this relationship, especially in the case where the assumption of
Reviewof Financial Economics 23 (2014)141–147
☆The authors are grateful to th e participants of the Bank of Ca nada's seminar,
International Risk Management Conference (Ve nice, June, 2009), and Far Eastern and
South Asian Econometric S ocieties (Tokyo, August, 2009). The auth ors thank the co-
editor Tarun K. Mukherjee, two anonymous refereesfor helpful comments and sugges-
tions. The views expressed in this paper are those of the authors. No responsibility for
them shouldbe attributed to the Bank of Canada.
⁎Correspondingauthor.
E-mailaddress: fuchunli@bankofcanada.ca (F.Li).
1
Itis important to mentionthat there are differentmethodsto identify contagionin the
literature.For example,contagionmay be viewed as the opening of newchannels of trans-
mission duringcrisis (Dungey & Martin, 2007). Other authors (e.g.,Bae, Karolyi, & Stulz,
2005;Billio & Caporion, 2005) use thresholdmodels to separate stableand crisis periods.
Morerecently, Markwat,Kole, and vanDijk (2009) view conationas a domino effectof cri-
ses, and Dungey, Milunovich,and Thorp (2010) propose an identified structuralGARCH
model to disentangle the h ypersensitivity of a domestic market and the conta gion
importedto a tranquil domestic marketfrom foreign crisis.
2
Pearsoncorrelation coefficientbetween two random variablesxand ywith expected
valuesμ
x
and μ
y
and standard deviations σ
x
and σ
y
is definedas ρx;y¼Ex−μx
ðÞy−μy
ðÞ½
σxσy:
http://dx.doi.org/10.1016/j.rfe.2014.05.001
1058-3300/©2014 Published by ElsevierInc.
Contents listsavailable at ScienceDirect
Review of Financial Economics
journal homepage: www.elsevier.com/locate/rfe
To continue reading
Request your trial