Determinants of Financial Stress in Emerging Market Economies

Journal of Banking & Finance xxx (2013) xxxxxx
Contents lists available at ScienceDirect
Journal of Banking & Finance

journal homepage: www.elsevier.com/locate/jbf
Determinants of nancial stress in emerging market economies

Cyn-Young Park , Rogelio V. Mercado Jr.
Economics and Research Department, Asian Development Bank, 6 ADB Avenue, Mandaluyong City 1550, Philippines
a r t i c l e
i n f o
Article history:
Available online xxxx
JEL classication:
F30
G01
G15
Keywords:
Financial stress index
Financial contagion
Emerging market economies
a b s t r a c t
The global nancial crisis of 20082009 illustrates how nancial turmoil in advanced economies could
trigger severe nancial stress in emerging markets. Previous studies dealing with nancial crises and contagion show the linkages through which nancial stress are transmitted from advanced to emerging markets. This paper extends the existing literature on the use of nancial stress index (FSI) in understanding
the channels of nancial transmission in emerging market economies. Using FSI of 25 emerging markets,
our panel regression estimates show that not only advanced economies FSI, but also regional and nonregional emerging market FSIs signicantly increase domestic nancial stress. Our ndings also suggest
that there is a common regional factor signicantly affecting domestic FSI in emerging Asia and emerging
Europe. Furthermore, the results from a structural vector autoregression model with contemporaneous
restrictions indicate that although a domestic nancial shock still accounts for most of the variation in
domestic FSI, regional shocks play an important role in emerging Asia.
2013 Elsevier B.V. All rights reserved.
1. Introduction
The recent global nancial crisis demonstrates the adverse effects of nancial globalization. Although nancial integration
brings direct and indirect benets to economies, it may increase
the countries vulnerability to nancial crises originating elsewhere (Stulz, 2005; Kose et al., 2006; Moshirian, 2008). For instance, the freezing of the credit markets in advanced economies,
particularly the United States (US), in late 2008 caused signicant
turmoil in emerging market nancial systems. As emerging markets are rapidly integrated into global and regional markets, the
origin of nancial stress is also becoming ubiquitous and the impact felt borderless. For example, the nancial crisis which started
in Thailand in 1997 quickly spread to the rest of East Asia, and then
to the Russia Federation and Brazil.
The transmission of nancial crises has been the subject of a
substantial body of economic literature. Cross-border transmission
of nancial crises is often manifested in co-movements of asset
prices and capital ows during times of crises. Earlier studies classied the causes of nancial contagion into two broad categories
(Calvo and Reinhart, 1996; Dornbusch et al., 2000; Kaminsky and
Reinhart, 1999, 2000; Moser, 2003). First, nancial asset prices
and capital ows can move similarly when economies share similar fundamentals and have strong macroeconomic interdependence through trade and nancial linkages. Similar fundamentals
Corresponding author. Tel.: +63 2 632 4444.

E-mail address: cypark@adb.org (C.-Y. Park).
may induce similar response to a shock, which leads to strong

co-movements in asset prices and capital ows. Second, the comovements may also result from herding behaviors and/or certain
decisions of investors, which affect different countries simultaneously. For example, a crisis in one country may prompt investors
to withdraw from all emerging market countries.
While the impact of nancial crisis is often devastating especially in emerging market economies, it has not been easy to monitor the buildup of a full-blown nancial crisis and to trace its
spread across borders. A number of studies have used a nancial
stress index (FSI) as a continuum and contemporaneous measure
of the severity of nancial crises. Using FSI shows that nancial
stress intensies due to greater fragility in the nancial systems
and exogenous shocks. Since the pioneering work of Illing and
Liu (2006) who dened nancial stress as episodes where economic agents are subjected to extreme uncertainty and varying
expectations of loss in nancial markets other authors developed
their own versions of FSI, including Hakkio and Keeton (2009) for
the Federal Reserve Bank of Kansas City; Hollo et al. (2012) for
European markets; Misina and Tkacz (2009) for selected advanced
economies; and Yiu et al. (2010) for Hong Kong Monetary
Authority.
The use of FSI has far-reaching benets for monetary authorities
and nancial regulatory and supervisory agencies. First, FSI combines various nancial market indicators into an aggregate index
to measure nancial market stress, hence eliminating the dependency on one or few indicators in the measurement of nancial
stress. Second, FSI allows for a measure of nancial stress to capture the degree and severity of nancial stress on a continuous
0378-4266/$ - see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jbankn.2013.09.018
Please cite this article in press as: Park, C.-Y., Mercado Jr., R.V. Determinants of nancial stress in emerging market economies. J. Bank Finance (2013),
C.-Y. Park, R.V. Mercado Jr. / Journal of Banking & Finance xxx (2013) xxxxxx
basis such that spikes in the stress index correspond to the periods
of severe nancial stress. It also allows fast, reliable identication
of crises or stressful periods. Third, FSI offers an aggregate measure
of nancial stability that is much simpler than other measures of
systemic nancial risks, without the complications of micro-level
assumptions of other measures.1 Nevertheless, FSI is found very
useful in alerting systemic risk conditions by its signaling properties
for identifying systemic stress episodes, grade thresholds, and corresponding probability of systemic stress (Oet et al., 2011). Of course,
FSI has its own shortcomings; particularly pertaining to its construction including choice variables, aggregation, and frequency choice.
Existing studies dealing with nancial stress, however, offer little insight on various channels for transmission of nancial stress
emanating from advanced economies and emerging market economies (either from the same region or from different regions) to
domestic nancial markets. Balakrishnan et al. (2009, 2011) explores the issue of nancial transmission from advanced and other
emerging economies to individual emerging market economies,
but paid little attention to geographical proximity in nancial
transmission and did not distinguish between regional and nonregional markets. Fernandez (2007) studied the impact of instability
in the Middle East to regional and nonregional emerging stock
markets; however, she focused on political stability as the source
of nancial market turmoil. This paper addresses the gap in the literature. Specically, it aims to examine the determinants of nancial stress in emerging market economies and to assess the
transmission of nancial shocks emanating from advanced and
other regional and nonregional emerging market economies to
individual emerging market economies. This paper adds to the previous literature in the following aspects.
First, it covers a longer period by extending the observations
from 1992 to 2012 to include a number of episodes of emerging
market crises in the early to mid-1990s as well as the latest crisis
episode that affected the global nancial markets in 20082009.
Extending the sample period with more crisis episodes will allow
for the analysis to provide more reliable results regarding the
transmission of a nancial shock.
Second, this paper employs two methodologies for constructing
domestic FSI for each emerging market in the sample one is the
variance-equal weights and the other is the principal component
analysis. This will allow for robustness checks on the overall patterns of individual FSIs.
Third, this study assesses the impact of external nancial shocks
on domestic FSI by differentiating their economic and geographic
origins, such as advanced versus emerging market economies as
well as regional versus nonregional emerging market economies.
The analysis specically focuses on whether or not a shock originating from emerging market economies would exert inuence
on the FSI of an individual emerging market economy in addition
to a shock from advanced economies. It will also assess the effect
of a common regional factor in domestic FSI for emerging markets.
The signicance of a common regional factor would help explain
the vulnerability of emerging market countries to regional nancial contagion.
Fourth, following the panel regression analysis on the magnitude and signicance of advanced and other regional and nonregional emerging market FSI on domestic FSI, we employ impulse
response functions and variance decompositions to assess the impact of a nancial shock coming from advanced and other emerging market economies on individual emerging market economies
FSI. A nancial shock generated from other emerging market economies are decomposed into regional and nonregional emerging
1
See Arnold et al. (2012) for discussion on the challenges in monitoring banking
systemic risks; and Allen et al. (2012) for discussion on the micro-level systemic risk
measures.
market nancial shocks. This will help assess whether or not the
impact of a nancial shock on domestic FSI would differ by the origins of the shock such as different economic (advanced versus
emerging market economies) and geographic (regional versus nonregional) groupings.
To carry out the empirical analysis of this study, aggregate
domestic FSI are constructed, drawing on the methodology used
by Cardarelli et al. (2011) and Balakrishnan et al. (2011), for a sample of emerging market economies using variance-equal weights
and principal component analysis as aggregation technique. To
verify the importance of global, country-specic, other countries
nancial stress, and regional factors in explaining domestic FSI, a
panel regression model involving 25 emerging markets from various regions including emerging Asia, emerging Americas, emerging
Europe, and other emerging countries is employed using quarterly
data from Q1 1992 to Q4 2012.2 Specically, it aims to determine
which factors including common regional factors contribute to
the increase of nancial stress in developing economies. Knowing
the signicance of advanced and other emerging market FSI in determining domestic FSI, we examine the impact of a nancial shock
emanating from advanced and other emerging economies on individual domestic FSI using a structural vector regression approach
with contemporaneous restrictions. This will allow us to determine
the magnitude and persistence of the effects of advanced and other
emerging market nancial shocks on individual domestic FSI.
This paper proceeds as follows. Section 2 discusses and constructs nancial stress index. Section 3 presents the determinants
of FSI and provides empirical specication for the panel regression
which will determine the signicance of advanced and other
emerging market nancial stress on domestic FSI, as well as the
importance of a common-regional factor. Section 4 provides the
structural vector autoregression specication and presents the impulse response functions and variance decompositions of FSI
shocks from advanced and other emerging market economies on
domestic emerging market FSI. Summary and some policy implication follow in Section 5.
2. Financial stress index

2.1. Literature review
There is abundant literature investigating the occurrence and
determinants of currency, banking, and sovereign debt crises in advanced and emerging economies; however, this literature failed to
account for the proper dating and intensity of said crises. For instance, Laeven and Valencia (2008) developed a database on the
timing and frequency of banking, currency, and sovereign debt crises for both advanced and emerging markets. Eichengreen et al.
(2004) looked into currency crises by developing an index of foreign exchange market pressure which incorporates foreign exchange depreciation and changes in international reserves.
Reinhart and Rogoff (2008) studied sovereign debt defaults and
found that crises usually emanate from nancial centers; and were
often accompanied by other crises like currency and banking crises. However, these studies devote little attention in dealing with
the contemporaneous severity of nancial crises. This comes from
the fact that most studies measure the occurrence of crises as a
simple binary variable, i.e., no crisis takes the value of zero (0)
and presence of crisis takes a value of one (1). As pointed out by
2
Emerging Asia includes the Peoples Republic of China; Hong Kong, China; India;
Indonesia; the Republic of Korea; Malaysia; the Philippines; Singapore; Taipei,China;
and Thailand. Emerging Americas includes Argentina, Brazil, Chile, Colombia, Mexico,
and Peru. Emerging Europe includes Czech Republic, Hungary, Poland, Russian
Federation, and Romania. Other emerging countries include Egypt, Israel, South
Africa, and Turkey.
16
16
16
16
12
12
12
-4
-4
-4
12
-8
-8
-8
-12
-12
-12
92
94
96
98
00
02
ARG_PC
04
06
08
10
12
92
ARG_SUM
94
96
98
00
02
ATR_PC
04
06
08
10
0
-4
-8
92
12
94
96
98
00
02
AUS_PC
ATR_SUM
04
06
08
10
-12
12
92
AUS_SUM
16
16
16
12
12
12
12
-4
-4
-4
-4
-8
92
94
96
98
00
02
CAN_PC
04
06
08
10
12
-8
-8
-8
-12
-12
-12
92
94
96
98
CAN_SUM
00
02
CHI_PC
16
04
06
08
10
12
92
94
96
98
CHI_SUM
00
02
COL_PC
04
06
08
10
92
12
16
16
12
12
12
-4
-4
-4
-4
-8
-12
92
94
96
98
00
02
DEN_PC
04
06
08
10
12
92
94
96
98
00
02
EGY_PC
04
06
08
10
12
-8
-8
-12
-12
EGY_SUM
92
94
96
98
00
02
FRA_PC
DEN_SUM
04
06
08
10
12
16
16
16
12
12
12
-4
-4
-4
-4
-8
-8
-8
-8
-12
-12
-12
-12
96
98
00
02
HKG_PC
04
06
08
10
92
12
94
96
98
00
02
HUN_PC
HKG_SUM
04
06
08
10
92
12
94
96
98
00
02
IND_PC
HUN_SUM
04
06
08
10
16
16
12
12
12
-4
-4
-4
-4
-8
-8
-8
-8
-12
-12
-12
-12
98
00
02
ISR_PC
04
06
08
10
92
12
94
96
98
00
02
ITA_PC
ISR_SUM
04
06
08
10
92
12
94
96
98
00
02
JPN_PC
ITA_SUM
04
06
08
10
92
12
16
16
12
12
-4
-4
-4
-4
-8
-8
-12
-12
92
94
96
98
00
MAL_PC
02
04
06
MAL_SUM
08
10
12
92
94
96
98
00
MEX_PC
02
04
06
MEX_SUM
08
10
12
-8
-8
-12
-12
92
94
96
98
00
96
98
00
94
96
98
00
NET_PC
94
96
02
04
06
NET_SUM
98
02
08
10
12
92
94
96
98
06
08
10
12
04
06
08
10
12
08
10
12
CZE_SUM
02
00
00
04
BRA_SUM
04
06
GER_SUM
02
04
06
08
10
12
08
10
12
INO_SUM
02
KOR_PC
12
12
94
JPN_SUM
16
16
02
INO_PC
16
96
92
92
12
12
94
98
IND_SUM
16
92
00
GER_PC
12
94
96
FRA_SUM
16
92
98
CZE_PC
16
-8
94
COL_SUM
12
-12
96
BRA_PC
16
-12
94
00
NOR_PC
04
06
KOR_SUM
02
04
06
08
10
12
NOR_SUM
Fig. 1. Individual country nancial stress index. PC = principal components; SUM = variance-equal weights. Note: The computed FSI corresponds to the aggregated values
shown in Eq. (6). Monthly average data on banking sector price index and the benchmark stock price index were taken from DataStream. The data were converted to year-onyear returns by taking the difference between current period and last years price index both in natural logarithm form. Monthly data for both foreign exchange and foreign
reserves are taken from the International Financial Statistics of the International Monetary Fund. Sovereign debt data refers to yield differentials between long-term domestic
government and US Treasury bonds in basis points. Monthly average data on sovereign debt spreads were taken from national sources accessed through CEIC Database.
However, for some countries where data started in the late 1990s, government treasury yield spreads of comparable tenure was used. In cases where data is unavailable for
certain months of the year, available data was extended to cover the whole year. For some countries with unavailable treasury yield data in the early 1990s, policy rates were
used. Source: Authors estimate.
16
16
16
12
12
12
-4
-4
16
12
8
4
0
-4
-4
-8
-8
-8
-8
-12
-12
-12
-12
92
94
96
98
00
02
PER_PC
04
06
08
10
12
94
96
98
00
02
PHI_PC
16
12
8
4
0
04
06
08
10
12
94
96
98
00
02
ROM_PC
04
06
08
10
96
98
00
02
04
06
08
10
92
12
16
12
12
98
00
02
RUS_PC
ROM_SUM
04
06
08
10
12
94
96
98
RUS_SUM
00
02
SAF_PC
04
06
08
10
12
-12
16
16
12
12
12
-4
-4
-8
-8
-8
-12
-12
-12
00
02
04
06
08
10
12
92
94
96
SPA_SUM
98
00
02
SWE_PC
04
06
08
10
92
12
94
96
98
00
02
SW I_PC
SW E_SUM
04
06
08
10
92
12
16
16
12
16
12
12
-4
-4
-4
-4
-8
-12
92
94
96
98
00
THA_PC
02
04
06
THA_SUM
08
10
12
-8
92
94
96
98
00
TUR_PC
02
04
06
08
10
12
96
98
00
02
12
04
06
08
10
12
08
10
12
SNG_SUM
94
96
98
00
02
TAP_PC
12
-8
94
SW I_SUM
16
-12
10
-8
SPA_PC
08
-4
-4
-12
98
06
PRC_SUM
SNG_PC
16
96
92
SAF_SUM
12
94
04
-8
92
16
92
02
-8
96
00
-4
-12
94
98
16
-4
92
96
PRC_PC
12
12
94
POL_SUM
16
-12
92
94
POL_PC
-8
-8
92
PHI_SUM
-4
-4
-12
92
PER_SUM
04
06
TAP_SUM
-8
-12
92
94
TUR_SUM
96
98
00
UKD_PC
02
04
06
UKD_SUM
08
10
12
-12
92
94
96
98
00
USA_PC
02
04
06
08
10
12
USA_SUM
Fig. 1 (continued)
Balakrishnan et al. (2011) and Illing and Liu (2006), the use of binary variable for crisis occurrence and dating does not provide a
measure of intensity of crisis and near-miss events.3 However,
some studies used a sector-specic index to measure sector-specic
intensity of crisis. For instance, Hanschel and Monnin (2005) derived
a banking stress index for Switzerland, although stress from other
nancial sectors is not considered. Furthermore, most of these studies do not include crises that emanate from the equity markets.
Against this backdrop, several authors proposed an FSI to address the weaknesses of previous literature in dating and measuring the severity of nancial crises. Several research works
developed FSI by capturing key features of nancial stress to identify a buildup in nancial stress and measure the intensity of nancial crisis. Illing and Liu (2006) created an index of nancial stress
for the Canadian nancial system, employing a continuous variable
with a spectrum of values where extreme values correspond to
periods of nancial crises. Their method was adapted and rened
by Cardarelli et al. (2008, 2011) which was used by the International Monetary Fund in their World Economic Outlook 2008; while
Balakrishnan et al. (2009, 2011) developed a similar index for 18
Some periods of heightened nancial market stress do not evolve into full-blown
nancial crisis. For example, the emerging market equity sell-off in June 2006 had
little macroeconomic impact, although it raised asset price volatility in some
countries.
emerging economies and used at the subsequent issue of IMFs

World Economic Outlook 2009.4
The use of FSI as a means of dating the duration and assessing
the severity of nancial crises has also gained popularity among
monetary authorities and nancial regulatory institutions. For
example, the Federal Reserve Bank of Cleveland, Kansas, and St.
Louis post their respective FSI on their website. Other researchers
covering advanced and emerging economies employ the said index, albeit with different component nancial variables to suit
domestic nancial market characteristics. For example, unlike the
use of foreign exchange market pressure index, Yiu et al. (2010)
used at-the-market implied volatility of Hong Kong dollar per US
dollar for the exchange rate component of the FSI. Van Roye
(2011) used several indicators to measure nancial stress in Germanys banking sector, including Treasury bill and Eurodollar future contract (TED) spread, money market spread, and banking
beta.
Another branch of literature on FSI examines the link between
nancial stress and economic activity. For example, Davig and Hakkio (2010) found the US economy uctuates between episodes of
4
The main difference between the advanced and emerging economies FSI used by
the IMF is the inclusion of indicators such as corporate bond spreads, inverted term
spread, and TED spread for advanced economies, which is inapplicable to emerging
economies given the low issuance of corporate bonds and data unavailability in
emerging economies.
Fig. 2a. Advanced and emerging markets nancial stress index and stress episodes (by region, using variance-equal weights).
Fig. 2b. Advanced and emerging markets nancial stress index and stress episodes (by region, using principal component analysis). Note: Aggregate and regional FSIs are
unweighted averages. Aggregate emerging market nancial stress periods are indentied based on the difference between the unweighted regional FSI and its unweighted
regional trend, where the regional trend is computed as the average of individual emerging market trend using Hodrick-Prescott lter method. Source: Authors calculation.
low nancial stress and high economic activity; and high nancial
stress and low economic activity. Other papers in this eld deals
with the contribution of nancial stress index to improving forecasts on economic activity. Ng (2011) showed FSI improves forecasting performance at horizons of 24 quarters for the US
economy. Afonso et al. (2011), Dovern and van Roye (2013), and
van Roye (2011) found that nancial stress lowers economic output and worsens scal positions. Cardarelli et al. (2011) identied
episodes of nancial turmoil in advanced economies using FSI and
assessed the impact of nancial stress on the real economy. They
found that nancial turmoil characterized by banking distress is
highly associated with severe and protracted downturns than
stress originating from securities or currency markets. In addition,
they also argued that economies with more arms-length nancial
systems appear to be particularly vulnerable to sharp contractions.
This paper follows the eld of literature set by Balakrishnan
et al. (2011) which employs FSI to examine cross-border transmis-
sion of nancial stress. Balakrishnan et al. (2011) created FSI for

emerging economies using the same methodology as Cardarelli
et al. (2011). They argued that domestic nancial stress index in
an emerging economy is inuenced by nancial stress in advanced
economies as well as common factors like global gross domestic
product (GDP) growth and interest rates; and country-specic factors like degree of nancial and trade linkages and other domestic
macroeconomic vulnerabilities. Their ndings suggest that nancial crises in advanced economies pass-through strongly to emerging economies; and that the depth of nancial linkages between
the two determines the extent of pass-through.
While the previous literature focused on the impact of nancial
transmission from advanced to emerging markets, this paper argues that a shock emanating from emerging markets is also important as emerging markets gain increasing presence in the global
nancial system. This paper also examines evidence of regional
nancial contagion by focusing on the transmission of a regional
Fig. 3a. Emerging markets nancial stress index and stress episodes (by country, using variance-equal weights).
Fig. 3b. Emerging markets nancial stress index and stress episodes (by country, using principal component analysis). Note: PC = principal components; SUM = varianceequal weights. Source: Authors calculation.
shock. We look into the signicance of a common regional factor as

another key determinant of domestic nancial stress index. If
domestic FSI is signicantly affected by a common regional factor,
we may conclude that there is evidence of regional nancial contagion. In this regard, the paper also assesses the effect of regional
and nonregional nancial shocks on domestic FSI.
2.2. Constructing emerging market nancial stress index
While we try to construct an index to capture the aforementioned conditions of nancial stress, there are three key issues.
First is to identify components of the index to cover key nancial
sectors. Second is to choose right variables to represent each component. And, third, what weighting scheme to use to aggregate
each component to a single FSI. Each is discussed accordingly, with
an explanation on how we construct an FSI for each of the emerging markets in this paper.
This paper follows the denition of nancial stress for emerging

markets as suggested by Balakrishnan et al. (2011). They dene
nancial stress as episodes when the nancial system is under
strain and its ability to intermediate is impaired. It is usually associated with the following: (1) large shifts in asset prices; (2) abrupt
increase in risk or uncertainty; (3) illiquidity of the nancial system; and (4) concerns about the health of the banking system.
2.2.1. Components and variable choices

We construct an FSI for each of the 25 emerging economies and
15 advanced economies following Balakrishnan et al. (2009, 2011),
Cardarelli et al. (2008, 2011), and Yiu et al. (2010). As in the previous studies, the composite FSI for each economy covers the four
major nancial sectors of the economy, which include:
(d) Emerging Americas
(a) Emerging Economies
(b) Advanced Economies
(e) Emerging Europe
(c) Emerging Asia
(f) Other Emerging Markets
Fig. 4. Components of nancial stress index. Note: Aggregate components of FSIs are unweighted averages of individual country. components computed using variance-equal
weights. Source: Authors calculation.
Banking sector: The lack of suitable data and institutional differences across countries make it hard to have a clear denition of
what constitutes a banking crisis. Some studies use ad hoc country-specic events to dene banking crisis. Others rely more on a
combination of qualitative and quantitative approach. For instance,
Kunt and Detragiache (1996) dene banking crisis as a situation
where any of the following conditions holds: (i) non-performing
loans is greater than 10%; (ii) the cost of bank rescue is at least
2% of GDP; (iii) banking problems result to large scale nationalization of banks; and (iv) extensive bank runs lead to emergency measures. Furthermore, some studies rely on quantitative methods
using aggregate balance sheet data of banks.
In constructing FSI, we include a measure of banking stress
called banking sector b as in Balakrishnan et al. (2011). This mea-
sure involves the ratio of bank share prices to total share prices.
It provides a stationary measure of relative equity-return volatility
and isolates banking sector-specic shocks. The banking sector b is
given by:
covr; m
varm
where r and m are the returns to the banking sector stock price index and the overall stock price index, respectively. If b is larger than
1, then the banking sector is relatively risky as the volatility of returns on bank shares is greater than the volatility of returns for
the overall market. The higher the banking sector b, the greater
the banking sectors stress. It must be noted that banking sector
beta is not a measure of co-movement between the two variables.

What it provides is a measure of how the banking sector returns
are more volatile than the overall stock price returns.
Some studies, including Cardarelli et al. (2008, 2011), van Roye
(2011), and Yiu et al. (2010), use risk spreads such as TED spread
and inverted term spread as a proxy for banking sector stress.
However, including other banking sector variables may pose as a
constraint for constructing FSI for each emerging market economies in the sample as most economies have relatively short time
series data for said indicators. For this reason, this paper will only
use banking sector b as a measure of banking stress in emerging
markets.
Monthly average data on the banking sector price index and the
benchmark stock price index were taken from DataStream. The
data were converted to year-on-year returns by taking the difference between current period and last years price index both in
natural logarithm form. Twelve-month rolling covariance and variance of returns were used to compute for the banking sector b. To
better capture banking sector stress, the derived series takes only
positive values exceeding a threshold of one and zero otherwise.
Foreign exchange market: Currency crises are dened as periods
of signicant devaluations, losses in foreign exchange reserves,
and/or defensive interest rate hikes. This study utilizes exchange
market pressure index (EMPI) as proposed by Eichengreen et al.
(2004) and used in Balakrishnan et al. (2011). The EMPI captures
the depreciation of the local currency with respect of US dollar
and the reduction in foreign exchange reserves. It is dened as:
Table 1
Principal component analysis: eigenvalue components.
Eigenvalue components
PC1PC3
Argentina
Brazil
Chile
Colombia
Czech Republic
Egypt
Hong Kong, China
Hungary
India
Indonesia
Israel
Rep. of Korea
Malaysia
Mexico
Peru
Philippines
Poland
Peoples Rep. of China
Romania
Russian Federation
Singapore
South Africa
Taipei, China
Thailand
Turkey
0.383
0.412
0.363
0.339
0.345
0.317
0.371
0.342
0.366
0.432
0.284
0.333
0.318
0.372
0.409
0.367
0.394
0.336
0.323
0.397
0.302
0.332
0.301
0.347
0.448
0.234
0.215
0.233
0.231
0.238
0.258
0.232
0.207
0.207
0.217
0.223
0.256
0.232
0.215
0.202
0.241
0.244
0.229
0.258
0.257
0.219
0.206
0.245
0.208
0.213
0.163
0.181
0.183
0.182
0.180
0.181
0.177
0.200
0.181
0.180
0.183
0.166
0.206
0.195
0.170
0.165
0.182
0.199
0.191
0.177
0.210
0.199
0.209
0.182
0.174
0.142
0.113
0.124
0.155
0.146
0.143
0.141
0.182
0.142
0.126
0.159
0.134
0.135
0.143
0.129
0.114
0.145
0.150
0.133
0.094
0.148
0.144
0.155
0.156
0.108
0.077
0.079
0.097
0.093
0.091
0.101
0.079
0.069
0.105
0.045
0.150
0.112
0.109
0.075
0.091
0.113
0.035
0.086
0.096
0.074
0.121
0.119
0.091
0.107
0.058
0.781
0.808
0.779
0.752
0.763
0.756
0.780
0.749
0.754
0.829
0.690
0.754
0.756
0.782
0.781
0.773
0.819
0.765
0.771
0.832
0.731
0.737
0.755
0.737
0.835
Source: Authors calculations.
market stress.5 Monthly average data on benchmark stock price index were taken from Datastream. The data were converted to
month-on-month returns by taking the difference between current
and previous months stock price index both in natural logarithm
form.
Aside from the stock market volatility measure, we also include
stock market returns as a component for equity market stress. The
stock market returns are computed from monthly average data on
benchmark stock price index availed from DataStream. The data
were converted to month-on-month returns by taking the difference between current and previous months stock price index both
in natural logarithm form.
Debt markets: Illing and Liu (2006) denes debt crisis as the
inability of sovereign nations or the private sector to service its foreign debts. Earlier literature on debt crises deals with a group of
emerging economies that were exposed to severe external indebtedness in the early-1980s. However, the occurrence of emerging
economies debt crises was identied mainly based on qualitative
information. The most common indicator of debt crises has been
the spread between risky and risk-free bond yields as a function
of expected losses. Spreads will widen if expectations of future
losses increase, or if greater uncertainty leads to lower condence,
implying a higher probable loss. Both factors are indicative of
stress.
This paper uses sovereign debt spreads to measure sovereign
debt stress. Data refers to yield differentials between long-term
(10-year) local government bonds and US Treasuries in basis
points. Monthly average data on sovereign debt spreads were taken from national sources accessed through CEIC Database. However, for some countries where data started in the late 1990s,
government treasury yield spreads of comparable tenure was used.
In cases where data is unavailable for certain months of the year,
available data was extended to cover the whole year. Furthermore,
where yt is the current periods equity return and yt1 is the previous periods equity returns. Balakrishnan et al. (2009, 2011), Cardarelli et al. (2008, 2011), and Yiu et al. (2010) used the same approach
in constructing equity market crisis index. This study will also use
time-varying volatility of stock returns as a measure of equity
5
The estimated variance equation for the equity returns following GARCH(1, 1)
shows that the estimated lagged squared residual and lagged variance are mostly
signicant for all countries in the sample. The estimated coefcients offer strong
support on the persistence of volatility of stock returns as the lagged variance is
signicant and greater than the lagged squared residual.
EMPIi;t
Dei;t li;De
ri;De
DRESi;t li;DRES
ri;DRES
where De and DRES denote month-on-month percent changes in

the foreign exchange rate of local currency per US dollar and foreign
exchange reserves; while r and l are standard deviation and mean,
respectively. Monthly data for both foreign exchange and foreign
reserves are taken from the International Financial Statistics of the
International Monetary Fund. Other methods of measuring foreign
exchange stress include hybrid volatility-loss approach such as
the CMAX calculation as used by Illing and Liu (2006) and foreign
exchange volatility following GARCH (1, 1) as in Bollerslev et al.
(1992). Eicher et al. (2009) developed an indicator of currency crisis
risk using price spreads between American Depositary Receipts
(ADR) and their underlying. They found ADR investors perceive
higher currency crisis risk when export commodity prices decline,
sovereign yield spreads increase, trading partners currencies depreciate, and interest rate spreads widen. This paper utilizes only the
EMPI as a measure of foreign exchange market stress.
Equity market: Previous studies dene equity crises as a sharp
decline in the overall stock price index. The drop suggests greater
expected loss, higher risk, or increased uncertainty about rms future prots. The simplest measure of equity crisis is the use of a
GARCH (1, 1) process to take into account time-varying
characteristics of movements in equity returns, following
Bollerslev et al. (1992). The volatility following a GARCH (1, 1) process is given by:
r2t x /1 e2t1 /2 r2t1
where r refers to the variance and e the error term in the regression given by:
yt ai;t byt1 ei;t
Table 2
Full sample panel estimates.
Dependen variable:
emerging country FSI
Expected Variance-equal weights

sign
(1)
(2)
(3)
Advanced economies
FSI
(+)
Emerging economies
FSI (excl country)
(+)
Emerging economies
FSI (excl region)
(+)
0.434*
0.277*
0.438*
0.301*
Regional FSI (excl

country)
(+)
0.12
0.420*
0.06
0.407*
0.12
0.413*
0.07
0.417*
LIBOR (3-month)
(+)
Global GDP growth
()
Global commodity
price change
Principal components
Variance-equal weights
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
0.566*
0.329*
0.346*
0.545*
0.449*
0.447*
0.574*
0.336*
0.353*
0.544*
0.445*
0.444*
0.05
0.05
0.856*
0.05
0.05
0.05
0.727*
0.05
0.04
0.05
0.853*
0.05
0.05
0.05
0.764*
0.05
0.10
0.07
0.09
0.07
()
0.279*
0.07
0.235*
0.05
0.033*
0.262*
0.06
0.115**
0.05
0.028*
0.12
0.265*
0.06
0.098***
0.05
0.029*
0.106**
0.04
0.119
0.07
0.020*
0.123**
0.04
0.013
0.07
0.020
0.07
0.121**
0.05
0.013
0.06
0.020*
0.284*
0.06
0.236*
0.05
0.034*
0.252*
0.06
0.118**
0.05
0.028*
0.11
0.253*
0.06
0.101**
0.05
0.029*
0.119**
0.04
0.112
0.07
0.022*
0.125*
0.04
0.001
0.06
0.020*
0.07
0.123**
0.04
0.002
0.06
0.020*
Financial openness
(t 1)
(+)
0.00
0.002*
0.00
0.002*
0.00
0.002*
0.00
0.002*
0.00
0.002*
0.00
0.002*
0.00
0.001
0.00
0.001
0.00
0.001
0.00
0.001**
0.00
0.001**
0.00
0.001**
Trade openness (t 1)
()
Current account
(t 1)
()
0.00
0.011
0.01
0.035
0.00
0.004
0.01
0.019
0.00
0.003
0.01
0.015
0.00
0.018*
0.01
0.052**
0.00
0.013**
0.01
0.030
0.00
0.013**
0.00
0.032
0.00
0.002
0.00
0.007
0.00
0.002
0.00
0.005
0.00
0.002
0.00
0.008
0.00
0.004**
0.00
0.022
0.00
0.005*
0.00
0.006
0.00
0.005*
0.00
0.005
Fiscal balance (t 1)
()
DForeign exchange
reserves (t 1)
()
0.03
0.065**
0.02
0.012
0.02
0.047*
0.02
0.009
0.02
0.047**
0.02
0.009
0.02
0.039**
0.02
0.027
0.02
0.036**
0.02
0.015
0.02
0.037**
0.02
0.016
0.01
0.047**
0.02
0.015
0.01
0.035**
0.02
0.007
0.01
0.035**
0.02
0.007
0.01
0.025***
0.01
0.029***
0.01
0.026***
0.01
0.015
0.01
0.025***
0.01
0.015
0.02
0.02
0.02
0.02
0.01
0.01
Dummy emerging Asia
(+)
Dummy emerging
Americas
(+)
0.02
0.014
0.18
0.156
0.02
0.099
0.16
0.067
0.02
0.078
0.16
0.090
0.02
0.250**
0.11
0.078
0.01
0.164
0.10
0.074
0.01
0.181**
0.09
0.066
Dummy emerging
Europe
(+)
0.21
0.237
0.18
0.224
0.18
0.109
0.10
0.221
0.09
0.335**
0.09
0.254***
Constant
0.938
0.73
0.506
0.62
0.587
0.63
0.452
0.44
0.325
0.42
0.309
0.42
0.22
0.20
1.406* 0.286
0.31
0.27
0.22
0.352
0.28
0.14
0.13
0.570** 0.208
0.24
0.23
0.15
0.197
0.24
R-squared (overall)
Observations
Country
Fixed effects
0.257
1845
25
Yes
0.418
1845
25
Yes
0.419
1837
25
Yes
0.118
1840
25
Yes
0.248
1840
25
Yes
0.256
1832
25
Yes
0.293
1845
25
No
0.439
1837
25
No
0.215
1840
25
No
0.307
1832
25
No
0.436
1845
25
No
0.300
1840
25
No
Note: Specications (1) to (6) do not include regional dummy variables and are estimated using xed-effects. Specications (7) to (12) include regional dummy variables and
are estimated using random-effects. Specications (1) to (3) and (7) to (9) use FSI aggregated using variance-equal weights; while (4) to (6) and (10) to (12) use principal
components. Specications (1), (4), (7), and (10) only include advanced economies FSI; (2), (5), (8), and (11) include other emerging market FSI excluding the specic country
FSI; (3), (6), (9), and (12) include emerging market FSI excluding the region and regional FSI excluding the country. Robust standard errors are used and are reported in italics.
Source: Authors estimate.
*
Signicant at 0.01.
**
Signicant at 0.05.
***
Signicant at 0.10.
for some countries with unavailable treasury yield data in the early
1990s, policy rates were used.
fore it is divided by its standard deviation, hence, the term variance-equal weights. Each component variable is computed as:
yt
2.2.2. Weighting scheme
The choice of a weighting scheme or how to combine the various components of nancial stress into one index is perhaps the
key to constructing an FSI. The difculty arises from the lack of a
reference series upon which meaningful weights can be derived
and tested. Hence, various weighting techniques are considered.
The most common method is the use of variance-equal weights.
With this approach, a nancial stress index is generated by giving
equal importance to each component variables. The variables are
assumed to be normally distributed and the series is demeaned
and standardized. The mean is subtracted from each variable be-
xt x
where yt is the demeaned and standardized series, x the mean of the

series, and r is the standard deviation of the series. The demeaned
and standardized components are then rebased from 0 to 100 (with
100 having the historically highest value) and averaged as done by
Cardarelli et al. (2008, 2011), and Yiu et al. (2010) or simply added
(without rebasing) as in Balakrishnan et al. (2009, 2011). The
advantage of this approach is that it is easily implemented and
applicable for cross-country comparisons. On the other hand, the
disadvantage is that it assumes that the demeaned and standardized series follows a normal distribution.
10
Table 3
Emerging Asia, xed effects panel estimates.
Dependent variable: emerging Asia country FSI
Expected sign
Advanced economies FSI
(+)
Emerging economies FSI (excl country)
(+)
Emerging economies FSI (excl region)
(+)
Regional FSI (excl country)
(+)
LIBOR (3-month)
(+)
Global GDP growth
()
Global commodity price change
()
Financial openness (t 1)
(+)
()
Current account (t 1)
()
()
DForeign exchange reserves (t 1)
()
Constant
R-squared (overall)
Observations
Country
Fixed effects
(1)
(2)
(3)
(4)
(5)
(6)
0.482*
0.05
0.254*
0.07
1.024*
0.16
0.297*
0.05
0.476*
0.07
0.405*
0.06
0.716*
0.09
0.415*
0.07
0.279**
0.10
0.279**
0.10
0.042**
0.01
0.002***
0.00
0.004
0.01
0.035
0.04
0.040
0.03
0.007
0.02
1.325
1.30
0.256**
0.09
0.076
0.08
0.036*
0.01
0.002***
0.00
0.000
0.01
0.020
0.04
0.031
0.02
0.022
0.01
0.952
1.00
0.334***
0.15
0.636**
0.22
0.263**
0.10
0.083
0.08
0.034
0.01
0.002**
0.00
0.001
0.01
0.020
0.04
0.034
0.02
0.019
0.01
0.814
0.99
0.231
804
10
Yes
0.376
804
10
Yes
0.392
804
10
Yes
0.067
0.06
0.079
0.14
0.023
0.01
0.002*
0.00
0.011
0.01
0.060
0.03
0.022
0.01
0.022
0.02
0.654
0.77
0.079
0.06
0.002
0.14
0.024*
0.01
0.002*
0.00
0.008
0.01
0.035
0.03
0.022
0.01
0.013
0.02
0.367
0.76
0.137
0.08
0.568*
0.10
0.105
0.07
0.006
0.14
0.022**
0.01
0.002*
0.00
0.010
0.01
0.033
0.03
0.027
0.02
0.016
0.02
0.441
0.79
0.157
804
10
Yes
0.261
804
10
Yes
0.266
804
10
Yes
Note: Specications (1) to (3) use FSI aggregated using variance-equal weights; while (4) to (6) use principal components. Specications (1) and (4) only include advanced
economies FSI; (2) and (5) include other emerging market FSI excluding the specic country FSI; (3) and (6) include emerging market FSI excluding the region and regional FSI
excluding the country. All specications were estimated using xed-effects to account for country heterogeneity. Robust standard errors are used and are reported in italics.
*
Signicant at 0.01.
**
Signicant at 0.05.
***
Signicant at 0.10.
Another popular approach is the use of principal component

analysis. The main idea behind using the principal component
analysis is to represent each component of the nancial stress index into a single variable by forming linear combinations of each
component. Through this approach, the resulting stress index captures the most common information from all components. The
resulting index is derived from the rst-three principal components, which refers to the coefcients of the linear combination
that maximizes the variance of the resulting composite nancial
stress index.6 Other weighting techniques used in the literature include credit aggregate-based weights, and transformations of the
variables using their sample cumulative distribution functions.
In this paper, the ve components of the nancial stress index,
given by:
6
The rst three principal components are used to derive the individual country FSI
as it captures around 7080% of information available from each component. 7080%
Eigenvalue proportion is widely used in empirical literature utilizing principal
component analysis. Table 1 presents the proportion of eigenvalue components. In
the earlier version of this paper, we used rst two principal components, as FSI using
the rst two principal components could identify more crisis periods. However, the
rst two principal components captured only about 5060% of available information.
As we use the rst three principal components, the number of identied crisis periods
declined somewhat, but the ndings from panel regression and structural vector
autoregression remain the same as those using rst two principal components. The
FSI for advance economies is constructed following the same method as that for
emerging market FSI to have comparable components and results. Advanced
economies include Australia, Austria, Canada, Denmark, France, Germany, Italy,
Japan, the Netherlands, Norway, Spain, Sweden, Switzerland, the United Kingdom,
and the United States.
EMFSI b Stockreturns Stockv olatility Debtspreads EMPI

6
are aggregated to a composite nancial stress index using the variance-equal weights and principal component analysis. Following
Balakrishnan et al. (2009, 2011), all components are demeaned
and standardized before adding for the variance-equal weights.
For the principal component analysis, the rst three components
are added and used as the emerging market FSI. In case of series
breaks, monthly FSIs were computed using the average of preceding
and succeeding monthly values. Fig. 1 presents the computed FSI for
each country covered in this study using variance-equal weights
and principal component analysis; including those for selected advanced economies. The gures illustrate that both methods of
aggregate FSI lead to comparable pattern of stressful and calm episodes. However, it can be noticed that the variance-equal weights
method lead to more erratic or volatile pattern than the one using
principal component analysis as aggregating technique.
2.2.3. Identifying episodes of nancial stress
Several approaches have been used in identifying episodes of
nancial stress based on the composite FSI. The simplest is through
a graphical inspection of the composite index done by Yiu et al.
(2010). Periods when the composite FSI peaks are considered
highly stressful episodes, while troughs are relatively calm periods.
Balakrishnan et al. (2009, 2011) and Cardarelli et al. (2008, 2011),
used a more rigorous approach. They identied episodes of nancial stress when the composite FSI reaches 1.01.5 standard deviation above trend. Illing and Liu (2006) used an event study
11
Table 4
Emerging Americas xed effects panel estimates.
Dependent variable: emerging Americas country FSI
Expected sign
(+)
Emerging Economies FSI (excl country)
(+)
(+)
(+)
LIBOR (3-month)
(+)
Global GDP growth
()
()
(+)
()
()
()
()
Constant
R-squared (overall)
Observations
Country
Fixed effects
Principal Components
(1)
(2)
(3)
(4)
(5)
(6)
0.651*
0.11
0.433**
0.15
0.908**
0.19
0.461**
0.16
0.494**
0.14
0.414***
0.16
0.901*
0.14
0.423***
0.17
0.437**
0.16
0.128
0.10
0.024**
0.01
0.014*
0.00
0.037***
0.01
0.078
0.07
0.124***
0.06
0.115
0.08
1.854
1.18
0.441**
0.14
0.222***
0.10
0.022**
0.01
0.011**
0.00
0.027
0.02
0.055
0.06
0.047
0.05
0.028
0.08
0.762
0.78
0.629***
0.30
0.256***
0.11
0.423**
0.13
0.202
0.12
0.022**
0.00
0.011**
0.00
0.035
0.03
0.050
0.07
0.027
0.05
0.030
0.07
0.484
0.83
0.352
440
6
Yes
0.482
440
6
Yes
0.473
432
6
Yes
0.188**
0.07
0.013
0.09
0.014**
0.00
0.009**
0.00
0.030**
0.01
0.091
0.06
0.072
0.04
0.042
0.03
0.529
0.93
0.234**
0.06
0.105
0.09
0.017**
0.00
0.008**
0.00
0.018
0.01
0.001
0.05
0.052
0.04
0.024
0.04
0.592
0.87
0.576**
0.21
0.278**
0.07
0.210**
0.05
0.107
0.08
0.015**
0.00
0.009***
0.00
0.032
0.02
0.008
0.05
0.021
0.04
0.019
0.03
0.059
0.90
0.266
435
6
Yes
0.383
435
6
Yes
0.362
427
6
Yes
*
Signicant at 0.01.
**
Signicant at 0.05.
***
Signicant at 0.10.
approach for Canada, where stressful events were drawn from annual and monetary policy reports of the Bank of Canada.
This study identies periods of nancial stress when the nancial stress indices exceed its long-run trend by one point for the
variance-equal weights and 0.7 point for the principal component
analysis.7 Given monthly structure of the dataset, stressful episodes
also include periods (months) in-between identied stressful periods. For instance, FebruaryApril 1998 are also counted as stressful
periods since they are in-between months of high stress levels
(December 1997January 1998; and June 1998March 1999). Aggregate advanced and emerging (regional and nonregional) markets,
and regional FSIs refer to unweighted average of individual country
FSI.8 To identify aggregate emerging market nancial stress periods,
the difference between the unweighted regional FSI and its unweighted regional trend are used. The identied stressful periods
are generally consistent with those of Balakrishnan et al. (2009) specically for 1990s; and they capture the recent global nancial crisis.
7
The trend was derived using the HodrickPrescott method where the smoothing
parameter k is set to 1600. Since the rst three principal component account for more
than 70% of variation, 0.7 deviation from the long-trend was used as criteria to
identify stressful episodes for the FSIs computed using principal component analysis.
Using one point rule as in the variance-equal weights method will fail to capture
signicant episodes of emerging market nancial stress.
8
Unweighted average is used so that individual country weight will not affect the
aggregate nancial stress index. For example, if the average FSI for emerging Asia is
weighted, then the impact of the Asian nancial crisis will be muted because of the
huge weight of Peoples Republic of China.
However, several points are noted on the identied stressful periods.

First, since the FSI is an ex post measure of nancial instability, it
would be inappropriate to use it in an ex ante context such that it
would be used to assess whether nancial system is fragile or not
today.9 Second, the identied periods are based on unweighted average values and, hence, may not capture stressful periods experienced
by one or relatively few countries in the sample. Third, the identied
periods are presented to illustrate the performance of the constructed FSI. They are not used in estimation in this paper.
2.3. Patterns of advanced and emerging markets FSI
Based on the computed composite domestic FSI for both advanced and emerging economies, several observations are noted
on the general patterns of nancial stress. First, episodes of nancial stress in emerging markets closely track those in advanced
economies (Figs. 2a and 2b). This pattern is clearly seen in the late
1998 and 20082009 nancial crises, where the crisis in advanced
countries instigated or aggravated emerging market nancial
stress. This observation is consistent with those from Balakrishnan
et al. (2011). For the episodes in the early 1990s, nancial stress
has been more pronounced in advanced than emerging economies
as northern Europe and Japan confronted banking sector woes.
Nonetheless, emerging market FSI showed a spike. For the late
9
See Borio and Drehmann (2009) on the discussion on the issues pertaining to
measures of nancial stability.
12
Table 5
Emerging Europe xed effects panel estimates.
Dependent variable: emerging Europe country FSI
Expected sign
(+)
(+)
(+)
(+)
LIBOR (3-month)
(+)
Global GDP growth
()
()
(+)
()
()
()
()
Constant
R-squared (overall)
Observations
Country
Fixed effects
(1)
(2)
(3)
(4)
(5)
(6)
0.643*
0.11
0.352**
0.06
0.666**
0.17
0.355*
0.07
0.696*
0.09
0.491**
0.10
0.861*
0.11
0.459**
0.11
0.185**
0.05
0.219
0.16
0.019***
0.01
0.005*
0.00
0.052**
0.01
0.042
0.02
0.026
0.03
0.002
0.03
2.245*
0.98
0.224**
0.05
0.127
0.13
0.019**
0.01
0.007*
0.00
0.040**
0.01
0.011
0.02
0.028
0.02
0.010
0.03
1.677***
0.78
0.480**
0.14
0.233**
0.08
0.231**
0.05
0.125
0.12
0.020**
0.01
0.007*
0.00
0.036**
0.01
0.009
0.03
0.029
0.02
0.016
0.03
1.435
0.75
0.170
306
5
Yes
0.402
306
5
Yes
0.433
306
5
Yes
0.097
0.08
0.270**
0.08
0.017***
0.01
0.003
0.00
0.053**
0.01
0.043
0.04
0.031
0.05
0.017
0.04
3.200**
1.18
0.170
0.09
0.091
0.07
0.020***
0.01
0.004***
0.00
0.044***
0.02
0.003
0.04
0.036
0.04
0.007
0.03
2.653
1.28
0.543**
0.16
0.351**
0.12
0.167
0.09
0.053
0.06
0.019***
0.01
0.003
0.00
0.033
0.02
0.003
0.05
0.036
0.04
0.008
0.03
1.885
1.61
0.131
306
5
Yes
0.273
306
5
Yes
0.273
306
5
Yes
*
Signicant at 0.01.
**
Signicant at 0.05.
***
Signicant at 0.10.
1998 episode, nancial crises in Latin America and the Russia

Federation compounded the nancial strain in the US due to the
collapse of Long Term Capital Management, causing both advanced
and emerging market FSIs to increase. Finally, during the 2008
2009 global nancial meltdown, advanced economies FSI reached
a new high and followed by emerging market FSIs. Emerging market FSI also tends to be larger than advanced countries FSI particularly in the 1990s. This may reect the fact that emerging
market economies experienced more nancial crises and therefore
greater nancial stress in the 1990s.
Second, although FSIs computed using variance-equal weights
and principal components exhibit a similar pattern during both
calm and stressful periods, there is a clear difference in their computed magnitudes. FSIs derived using variance-equal weights tend
to have greater/lesser magnitudes than those computed using
principal components, implying that the variance-equal weights
can better capture episodes of severe nancial stress and nearmiss events. In addition, there are also differences between the
stressful period identied using the variance-equal weights and
principal components. Variance-equal weight tends to capture
more stressful periods than principal components analysis.
Third, emerging market FSIs exhibit co-movement such that
individual country nancial stress index increases during periods
of great nancial market turmoil in emerging economies (Figs. 3a
and 3b). However, the peak of individual country FSIs can vary
across countries during episodes of nancial stress. For instance,
only few emerging market countries experienced severe nancial
strain in 1995 and 2002 compared to 19971998 and 20082009,
where almost all emerging market FSIs rose. In addition, there

are episodes of individual market stress that are specic to a country and not to emerging markets in general. For example, the spike
in Brazils FSI in late 2005 backs emerging market trend.
Fourth, the constructed FSIs seem to capture moments of stress
in emerging market nancial systems very well. Fig. 4af presents
the component breakdown of the unweighted average of advanced
and emerging economies FSIs, along with their regional groupings.
For the emerging economies as a whole (Fig. 4a), it can be observed
that banking, equity, currency, and debt markets were all under severe strain during the 1997/1998 crises; while equity and currency
markets were under strain during the 2008/2009 global nancial
crisis, reecting emerging markets healthy banking and scal positions. For the advanced economies (Fig. 4b), debt markets played a
huge role during the stress episode in the early 1990s, while almost
all components were under stress during the 2008/2009 crisis.
Interestingly, banking sector stress has been elevated since the recent global nancial crisis. Across emerging market regions
(Fig. 4cf), equity market stress dominates all nancial crisis
episodes. Nonetheless, there are differences in components per
episode across regions. Currency market stress spiked in late
19971998 for emerging Asia; while debt market stress was more
severe in emerging America during the same period.
3. Determinants of domestic emerging market FSI
Understanding the determinants of domestic FSI is important
for policymakers to ensure nancial stability. Given the increasing
13
Table 6
Fixed effects panel estimates for 1Q19924Q1999.
Expected sign
(+)
(+)
(+)
(+)
LIBOR (3-month)
(+)
Global GDP growth
()
()
(+)
()
()
()
()
Constant
R-squared (overall)
Observations
Country
Fixed effects
(1)
(2)
(3)
(4)
(5)
(6)
0.252**
0.10
0.272**
0.11
0.322
0.24
0.281**
0.11
0.234***
0.13
0.240***
0.13
0.024
0.15
0.263***
0.13
0.937**
0.35
0.745**
0.31
0.074*
0.02
0.006
0.01
0.017
0.03
0.050
0.04
0.077***
0.04
0.029
0.06
3.905***
2.00
0.661**
0.27
0.546
0.38
0.058*
0.02
0.005
0.01
0.018
0.03
0.038
0.03
0.069
0.04
0.021
0.06
3.130
1.91
0.034
0.19
0.337***
0.17
0.658**
0.27
0.556
0.39
0.059*
0.02
0.006
0.01
0.016
0.03
0.031
0.04
0.071
0.04
0.029
0.06
3.030
2.11
0.019
546
24
Yes
0.024
546
24
Yes
0.029
538
24
Yes
0.359
0.29
0.536**
0.22
0.033***
0.02
0.001
0.00
0.004
0.01
0.028
0.02
0.059***
0.03
0.005
0.05
0.914
1.17
0.352
0.28
0.526***
0.26
0.033**
0.01
0.001
0.00
0.004
0.01
0.028
0.03
0.059***
0.03
0.005
0.05
0.905
1.20
0.179
0.11
0.279**
0.11
0.369
0.30
0.477***
0.26
0.032***
0.02
0.001
0.00
0.007
0.01
0.026
0.03
0.060***
0.03
0.008
0.05
0.745
1.37
0.093
541
24
Yes
0.094
541
24
Yes
0.136
533
24
Yes
*
Signicant at 0.01.
**
Signicant at 0.05.
***
Signicant at 0.10.
degree of nancial integration, it is also crucial to know whether

there is nancial contagion and, if so, to what extent a nancial
shock originating elsewhere affects domestic nancial condition.
In this section, we employ a panel regression analysis to assess
the effects of nancial stress from various sources such as global,
country-specic, and other emerging market on domestic nancial
stress, as well as the role of global and domestic factors in explaining domestic FSI. We also include a dummy variable for region to
evaluate the effect of region-specic factors.
3.1. Data and methodology
The dataset includes quarterly data for 25 emerging market
economies. Data on individual emerging market FSI is taken from
the previous section, while the advanced, emerging excluding
country, emerging excluding region, and region excluding country
FSI are unweighted average of individual FSI in Section 2. Data for
the global GDP growth and scal balance (% of GDP) are taken from
the Oxford Economics. Data for London interbank offered rate (LIBOR), trade openness (exports plus imports as % of GDP), current
account (% of GDP), and foreign exchange reserves are sourced
from the International Financial Statistics and World Economic Outlook Database of the IMF and national sources accessed through
CEIC. Data for nancial openness is taken from the External Wealth
of Nations Database and extended using the International Investment Position Database of the IMF. Monthly data are converted
to quarterly series beginning Q1 1992Q4 2012 using the average

of 3 months of a quarter. In cases where quarterly data is unavailable, annual data is converted to quarterly series or the average annual value is used to ll the missing observations.
As mentioned in the previous section, FSIs computed using
principal component analysis have smaller values than those computed using variance-equal weights. Among the global indicators,
the change in global commodity price has the highest mean and
standard deviation compared to global GDP growth and LIBOR.
For openness indicators, de facto nancial integration has greater
cross-country variation compared to trade openness. For the
domestic indicators, foreign exchange reserves has the highest
standard deviation compared to current account and scal balance,
implying that cross-country differences in emerging market foreign exchange reserve holdings is relatively high.
Panel unit root test of the Augmented Dickey Fuller (ADF)
type was used to check for stationarity of all variables. The results reject the null hypothesis that all variables in the panel
contain unit root, in favor of the alternative hypothesis of no
unit root. However, for foreign exchange reserves, the null
hypothesis cannot be rejected, therefore the lagged value of
the rst-differenced foreign exchange reserves was used in
the estimation. For the global indicators, a similar test was
conducted using the same procedure and specication for time
series data. The results show that the indicators do not contain
unit root at 10% level of signicance.
14
Table 7
Expected sign
(+)
(+)
(+)
(+)
LIBOR (3-month)
(+)
Global GDP growth
()
()
(+)
()
()
()
()
Constant
R-squared (overall)
Observations
Country
Fixed effects
(1)
(2)
(3)
(4)
(5)
(6)
0.511*
0.07
0.311*
0.07
0.817*
0.11
0.323*
0.07
0.435*
0.06
0.389*
0.06
0.752*
0.12
0.386*
0.06
0.354*
0.06
0.625*
0.12
0.002
0.01
0.001
0.00
0.011
0.01
0.051***
0.03
0.072**
0.03
0.007
0.03
1.346***
0.64
0.257
0.05
0.249***
0.14
0.022*
0.01
0.001
0.00
0.001
0.01
0.038
0.02
0.043***
0.02
0.041
0.02
0.141
0.61
0.440*
0.14
0.381*
0.09
0.261*
0.05
0.257***
0.14
0.020*
0.01
0.001
0.00
0.002
0.01
0.034
0.02
0.044***
0.02
0.042***
0.02
0.233
0.61
0.127
775
25
Yes
0.379
775
25
Yes
0.380
775
25
Yes
0.188*
0.05
0.433*
0.13
0.003
0.00
0.000
0.00
0.011
0.01
0.023
0.02
0.058**
0.02
0.024
0.02
1.124***
0.53
0.136**
0.05
0.138
0.14
0.015*
0.00
0.000
0.00
0.005
0.01
0.012
0.02
0.044**
0.02
0.040***
0.02
0.252
0.61
0.605*
0.14
0.118
0.13
0.140**
0.05
0.151
0.15
0.012**
0.00
0.000
0.00
0.004
0.01
0.013
0.02
0.045**
0.02
0.041***
0.02
0.181
0.64
0.163
775
25
Yes
0.342
775
25
Yes
0.341
775
25
Yes
*
Signicant at 0.01.
**
Signicant at 0.05.
***
Signicant at 0.10.
3.2. Panel least squares regression

The model specication is as follows:
EMFSIi;t ai b1 AEFSIt Rj bj EMFSIX i;t Rj bj Globalt

Rj bj Domestici;t1 Rj bj Dumi ei;t
where EMFSIi,t is the individual country FSI computed using variance-equal weights and principal component analysis; AEFSIt is
the unweighted average of advanced economies nancial stress inP
dex;
jbjEMFSIXi,t refers to measures of other emerging market
nancial stress indexEMFSI excluding country, EMFSI excluding
region, and regional FSI excluding country, where each index is
computed as residual using unweighted average of other emerging
market FSIs as dependent variables; and advanced economies FSI
P
and global indicators as the regressors. jbjGlobalt includes indicators such as global interest rates, global output growth, and global
P
commodity price increase. jbjDomestici.t1 refers to lagged value
of country-specic factors including nancial and trade openness,
current account balance, scal balance, and change (rst difference)
P
in foreign exchange reserves. jbjDumt refers to regional dummy
variables for emerging Asia, emerging Americas, and emerging
Europe, where the value takes one (1) if the country belongs to
the region and zero (0) otherwise.
Eq. (7) is rst estimated without regional dummy variables
P
( jbjDumt) using xed-effects ordinary least squares estimation.
Based on the results of the Hausman test, country-specic effects
are adequately modeled by xed-effects regression. However, since
we also want to test the signicance of a common regional factor,
Eq. (7) is estimated including three regional dummy variables

using random-effects generalized least squares estimation. Using
xed-effects with regional dummy variables for Eq. (7) would be
inappropriate since cross-country heterogeneity is already captured by the regional dummy variables.10
To avoid possible endogeneity, lagged values of country-specic
factors were used in the estimation. Simple pairwise correlation of
residuals reveals weak correlation between the estimated residuals
and independent variables, implying that endogeneity is not a concern under the current specications. To address possible heteroskedasticity, robust standard errors are used. To conduct
robustness checks, Eq. (7) is estimated for each region by dropping
countries that are not member of the region. For example, Eq. (7) is
estimated using xed-effects only for emerging Asia countries and
the same is conducted for the other regions. To test the evolution of
nancial contagion or transmission from advanced and other
emerging market (regional and nonregional) economies to an individual emerging market FSI over time, the sample period was split
into three. The rst covers 1Q19924Q1999 or the 1990s; the second covers 1Q20004Q2007 or before the global nancial crisis
period (pre-GFC); and the third includes 1Q20084Q2012 or after
the global nancial crisis period (post-GFC).
10
The three regional dummy variables represent emerging Asia, emerging Americas,
and emerging Europe, respectively. The intercept for the random-effects generalized
least squares estimation for Eq. (7) corresponds to the common factor for other
emerging market economies including Egypt, Israel, South Africa, and Turkey.
15
3.3. Empirical results

Balakrishnan et al. (2011) argued that there are several factors
inuencing emerging market nancial stress. First, the nancial
turmoil in advanced economies tends to increase the nancial
stress in emerging economies. This reects the nancial contagion
from advanced economies to emerging economies. Second, domestic nancial stress can be also exacerbated by common global factors such as changes in GDP growth, commodity prices, and
interest rates. Third, country-specic factors such as the degree
of openness (nancial and trade) and macroeconomic vulnerabilities (current account, scal balance, and foreign exchange reserves)
seem to affect the individual emerging market FSIs.
Previous literature notes that nancial transmission can be
caused by some common factors, which would affect individual
emerging market FSIs simultaneously. Such common factors can
include global shocks and may manifest through investors herding
behavior, cross-country contagion, and common credit conditions.
Adding to the existing studies, this paper argues that there may be
shocks that are region-specic, in line with the observed regional
nancial integration. These shocks may explain the rapid transmission of a nancial shock in a particular region, such as regional
nancial contagion that was witnessed in emerging Asia during
the nancial crisis of 19971998 and emerging Americas in the
1990s and early 2000s.
Table 2 presents the panel estimates on the determinants of
emerging market FSI. The estimates show that specications
including other emerging market FSI as a determinant of domestic
FSI have better t than those that include advanced economies FSI
only, as shown by their higher overall R-squared. For robustness
checks, dataset for each region is estimated separately and period
coverage is divided into three. Tables 35 present results for
emerging Asia, emerging Americas, and emerging Europe, respectively. Similar to the estimates in Table 2, specications including
other emerging market FSI as a determinant of domestic FSI have
better t than those that include advanced economies FSI only.
Tables 68 show estimates for the subsample periods, namely
1Q19924Q1999 (1990s); 1Q20004Q2007 (pre-GFC period); and
1Q20084Q2012 (post-GFC period).
The estimates presented in Table 2 offer several key ndings.
First, nancial stress from both advanced and emerging market
economies (excluding a country), signicantly increases domestic
FSI for the (excluded) emerging market economy. This nding is
consistent with the earlier studies including Aizenman and Pasricha (2010) and Balakrishnan et al. (2011). The estimates in this paper also show that emerging market (excluding the particular
region) and regional FSIs (excluding the particular country) significantly increases domestic FSI as seen in specications (3), (6), (9),
and (12). These ndings support the view that nancial contagion
could originate from both advanced and other emerging
economies.
Second, both global and domestic factors signicantly inuence
domestic FSI. Higher global interest rates tend to increase domestic
nancial stress, suggesting tightening conditions in international
credit markets can have adverse effects on the domestic nancial
condition. Higher global GDP growth reduces domestic nancial
Table 8
Expected sign
(+)
(+)
(+)
(+)
LIBOR (3-month)
(+)
Global GDP growth
()
()
(+)
()
()
()
()
Constant
R-squared (overall)
Observations
Country
Fixed effects
(1)
(2)
(3)
(4)
(5)
(6)
0.533*
0.06
0.347*
0.06
0.745*
0.20
0.352*
0.06
0.527*
0.08
0.447*
0.07
0.629*
0.18
0.432*
0.07
0.163
0.13
0.190*
0.06
0.011**
0.00
0.002
0.00
0.004
0.01
0.086*
0.02
0.002
0.02
0.002
0.02
1.915*
0.61
0.273**
0.13
0.086
0.06
0.024*
0.01
0.004*
0.00
0.005
0.01
0.073*
0.02
0.008
0.02
0.014
0.02
1.018
0.67
0.602**
0.23
0.158
0.13
0.290**
0.13
0.080
0.05
0.024*
0.01
0.004*
0.00
0.005
0.01
0.072*
0.02
0.008
0.01
0.013
0.02
1.045
0.65
0.278
475
25
Yes
0.244
475
25
Yes
0.246
475
25
Yes
0.067
0.12
0.130**
0.06
0.008**
0.00
0.003
0.00
0.012**
0.01
0.046**
0.02
0.006
0.01
0.011
0.02
2.338*
0.66
0.117
0.12
0.013
0.05
0.016*
0.00
0.004***
0.00
0.006
0.01
0.037***
0.02
0.009
0.01
0.005
0.02
1.955**
0.70
0.675*
0.18
0.024
0.16
0.143
0.12
0.004
0.05
0.015*
0.00
0.004***
0.00
0.005
0.01
0.036***
0.02
0.011
0.01
0.001
0.01
1.855**
0.68
0.069
475
25
Yes
0.068
475
25
Yes
0.073
475
25
Yes
*
Signicant at 0.01.
**
Signicant at 0.05.
***
Signicant at 0.10.
16
(a) Argentina
Response of DOM_SUM to Shock2
-1
-1
-1
-2
-2
2
(b) Brazil
Response to Structural One S.D. Innovations 2 S.E.

2
10
12
-2
2
10
12
10
12
Response of DOM_PC to Shock2
-1
-1
-1
-2
10
12
-1
-1
-1
-2
-2
2
10
(c) Chile
12
-2
2
10
12
10
12
10
12
-2

2
-1
-1
-1
-2
10
12
-2
10
12
-2
-1
-1
-1
-1
-2
10
12
10
12
-2
-2
10
-2
2
12
10
12

2
-1
-1
-1
10
(e) Colombia
12
-2
10
12
-2

2
-1
-1
-1
10
10
12
-2
10
12

2
-1
-1
-1
10
(g) Egypt
12
-2
10
12
12
10
12
-2

2
-1
-1
-1
10
12
-2
10
12
-2
12
-1
-1
-1
-2
-2
2
10
12
10
12
10
12

2
-1
-1
-1
-2
10
12
-2
10
12
-2

2
-1
-1
-1
-2
10
12
-2
10
12
-2
10
12

2
-1
-1
-1
-2
10
12
-2
10
12
-2
-1
-1
-1
-1
-1
-1
10
12
-2
10
12
10
12
10
12
10
12
12
10
-2

2
12
(h) Hong Kong,China
10
12
10

-2
2
-2

10
-2
-2
(f) Czech Republic
-2

-2
12
-2
2
10
-2
-1
-2
-1
(d) Peoples Rep.ofChina
-2

-2
2
10
12
10
12
-2
10
12
-2
10
12
Fig. 5. Impulse responses of domestic FSI to advanced economies, other emerging markets, and domestic nancial shock (response to structural one standard deviation).
17
(i) Hungary
(j) India
-1
-1
-1
-1
-1
-1
-2
10
12
-2
10
-2
12
10
12
-2
10
12

-1
-1
-2
10
(k) Indonesia
12
-2
10
10
12
-2
-1
-1
-1
-1
10
12
-2
10
12
(l) Israel
-2
10
12
-2
-1
-1
-1
-1
-2
-2
10
12
10
12
-2
10
12
10
12

-2
10
12
-2

2
-1
-1
-1
-1
-1
-1
-2
-2
-2
-2
-2
-2
10
(m) Rep. of Korea
12
10
12
10
12
10
12
(n) Malaysia
10
12
-1
-1
-1
-1
-1
-1
10
12
-2
10
12
-2
10
-2
12
10
12

-2
10
12
-1
-1
-1
-1
-1
-1
10
(o) Mexico
12
-2
10
12
-2
10
-2
12
10
(p) Peru
12
-2
10
12
-2

2
-1
-1
-1
-1
-1
-1
-2
8
10
12
-2
2
10
12
10
12
-2
2
10

-1
-1
-1
-1
-1
-1
-2
4
10
12
-2
-2
2
10
12
10
12
10
12
10
12
-2
2
10
12
10
12
10
12
-2
-2
12
2
4
6
8
10
12
2
-2
12
10
-2
12
-2
10
-2
-2
12
10
-1
-2
-1
-2
12
-2
-2
2
10
12
10
12
Fig. 5 (continued)
18
(q) Philippines
(r) Poland
-1
-1
-1
-1
-1
-1
-2
-2
2
10
-2
12
10
12
10
-2
-2
12
10
12
10
12
-1
-1
-1
-1
-1
-1
-2
2
10
(s) Romania
-2
-2
12
10
12
10
-2
2
12
10
12
10
12
-1
-1
-1
-1
-1
-1
10
12
-2
-2
-2
10
12
10
-2
2
12
10
12

10
12
-1
-1
-1
-1
-1
-1
-2
6
10
(u) South Africa
10
12
10
12
10
(v) Singapore
-2
-2
-2
12
12
10
12
-1
-1
-1
-1
-1
-1
-2
-2
-2
-2
-2
10
12
10
12
10
12
10

12
10
12
-2
-1
-1
-1
-1
-1
-1
10
(w) Taipei, China
12
-2
10
12
-2
10
12
-2
10
(x) Thailand
12
-2
10
12
-2
-1
-1
-1
-1
-1
-1
10
12
-2
10
12
-2
10
12
-2
10
12
-2
10
12
-2
-1
-1
-1
-1
-1
-1
-2
2
10
12
10
12
-2
-2
-2
2
10
12
10
12
10
12
10
12
-2

12
-2
10
-2
12
-2
2
10
-2
2
-2
12
10
-2
2
-2
(t) Russian Federation
-2

-2
2
10
12
10
12
-2
2
10
12
10
12
Fig. 5 (continued)
19
(y) Turkey
2

2
2
-1
-1
-1
-2
10
12
-2
10
12
-2
10
12

2
2
2
1
-1
-1
-1
-2
10
12
-2
10
12
-2
10
12
Fig. 5 (continued)
stress, implying that as global demand conditions improve nancial stress declines. In contrast, sound domestic macroeconomic
conditions have mitigating effect on domestic nancial stress. Current account surplus, scal surplus, and higher foreign exchange
reserves lower domestic nancial stress. Among these domestic
indicators, scal surplus signicantly lowers domestic FSI across
specications. This means that scal space of the country or its
ability to increase domestic spending during episodes of nancial
market turmoil remains crucial in lowering domestic nancial
stress. These results are consistent with the ndings of Cardarelli
et al. (2011), Balakrishnan et al. (2011), and Afonso et al. (2011).
Third, trade and nancial openness are both signicant but have
opposing inuence on domestic FSI. Greater de facto nancial integration tends to signicantly increase domestic nancial stress;
while greater trade openness reduces it. These ndings are consistent with Aizenman and Pasricha (2010), Balakrishnan et al.
(2011), and Dovern and van Roye (2013) and the existing view that
nancial openness can increase emerging market countries vulnerability to nancial shocks through increased capital ows and
its volatilities. Greater trade openness tends to improve economic
performance as diversied trading partners could positively inuence economic stability, thereby reducing domestic FSI.
Fourth, the effects of dummy variables for different regions are
also found signicant for emerging Asia and emerging Europe, suggesting that a common regional factor also plays an important role
in driving domestic nancial stress in these emerging market
countries. While it is difcult to pin down what this common regional factor represents, it could be similar economic fundamentals, institutional set ups, and anything about the region that
causes common perception and herding behavior of foreign investors. This nding also suggests that both regions are vulnerable to
nancial contagion within the region, wherein a crisis in one member country will have signicant impact on the other countries in
the region. The experience of emerging Asia during the 1997
1998 nancial crisis concurs with this nding.11
Fifth, the results are robust when one looks into the xedeffects estimates for the emerging market regions. Tables 35
present the results for emerging Asia, emerging Americas, and
emerging Europe, respectively. As in the full sample results, advanced economies FSI and emerging market FSI (excluding the particular country) signicantly increases domestic emerging market
FSI of a particular country. Interestingly, unlike emerging Americas
and emerging Europe, trade openness is insignicant for emerging
11
To check for the robustness of the results, we also conducted a separate
regression by removing the Hong Kong, China and Singapore, both of which are
considered as the Asian nancial hub in the emerging Asia sample, to determine
whether the signicance of a common Asian regional factor is due to the inclusion of
these two nancial hubs. The panel regression results show that a common Asian
regional factor is still positive and signicant.
Asia; but unlike emerging Asia and emerging Europe, scal surplus
is signicant for emerging Americas.
Sixth, the subsample period results in Tables 68 show several
interesting ndings. First, the effects of advanced and emerging
market FSIs on domestic FSIs increased over time, in line with
the trend of nancial globalization. Especially, the effects of (nonregional) emerging market FSIs have become much more signicant in the 2000s, suggesting the greater inuence of emerging
markets in the global nancial system. Second, regional nancial
stress (excluding the country) plays a lesser role in the post-GFC
period, perhaps reecting the fact that the global nature of GFC
makes differentiating regional or nonregional shocks meaningless
in tracing the origin of nancial stress. Third, trade and nancial
openness indicators signicantly inuence domestic FSI in the
post-GFC period, but not in the 1990s and pre-GFC period. This
suggests more open economies are more exposed to the effects
of nancial globalization and economic integration, which appear
to be strengthening in recent years similar to the previous two
points.
4. Impact of advanced and other emerging market nancial
shocks on domestic emerging market FSI
Knowing that advanced and other emerging market economies
FSI signicantly increase domestic FSI, we can examine the magnitude and persistence of nancial shocks coming from advanced
and other emerging economies on domestic FSI using a structural
vector autoregression (SVAR) with contemporaneous or short-run
restrictions.12
4.1. Structural vector autoregression
The SVAR model used in this paper takes a structural equation
form, given by:
ALY t et
where A(L) is a matrix of polynomial in the lag operator L. Yt is a

3 1 data vector corresponding to advanced economies FSI, other
emerging market countries FSI (excluding the particular country),
and domestic FSI.13 et is a 3 1 structural disturbance. The reduce-form equation is given by:
Y t BLY t1 lt
where B(L) is a matrix polynomial in lag operator L and Var(ut) = X.

To recover the parameters and structural shocks in Eq. (8), contemporaneous short-run identifying restrictions are imposed, following
Blanchard and Watson (1986), Bernanke (1986); and Sims (1986).
The orthogonal factorization matrixes are given the form:
B
A @ NA
0
1
NA
C
B
0A B @ 0
NA NA 1
0
NA
0
C
0 A
NA
10
12
This paper uses structural vector autoregression (SVAR) instead of vector
autoregression (VAR) to impose structural restrictions where domestic FSI does not
inuence advanced and other emerging market FSIs. This restriction is largely in line
with the ndings that a nancial shock spreads from advanced to emerging market
economies and the assumption that each emerging market is not large enough to
inuence the global and regional markets. Nonetheless, a VAR model was also
estimated for robustness tests. The impulse responses and variance decompositions
from the VAR models are similar to the SVAR results discussed in this paper.
Specically, the VAR results conrm that domestic nancial shock dominates
variation in domestic FSI and that for most emerging Asia countries, the impact of
regional (excluding the country) shock is greater than nonregional shock.
13
A separate structural vector autoregression model is also estimated decomposing
other emerging market countries FSI into nonregional and regional emerging market
FSI excluding the country. The same procedure would apply for the four-variable
SVAR.
20
(a) Argentina
(b) Brazil
Response of DOM_SUM to Shock1 Response of DOM_SUM to Shock2 Response of DOM_SUM to Shock3 Response of DOM_SUM to Shock4
Response of DOM_SUM to Shock1 Response of DOM_SUM to Shock2
-1
-1
-1
-1
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9101112
-2
1 2 3 4 5 6 7 8 9101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(c) Chile
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
-1
-1
-1
-1
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
-2
-1
-1
-1
-1
-1
-1
-1
-1
1 2 3 4 5 6 7 8 9 101112
-2
(e) Colombia
1 2 3 4 5 6 7 8 9 101112
-2
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(f) Czech Republic
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
(d) Peoples Rep.ofChina
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(g) Egypt
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(h) Hong Kong,China

1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
-2
-2
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
1.6
1.2
1.2
1.2
1.2
0.8
0.8
0.8
0.8
0.4
0.4
0.4
0.4
0.0
0.0
0.0
0.0
-0.4
-0.4
-0.4
-0.4
-0.8
1 2 3 4 5 6 7 8 9 101112
-0.8
1 2 3 4 5 6 7 8 9 101112
-0.8
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
1.5
1.5
1.5
1.5
1.0
1.0
1.0
1.0
0.5
0.5
0.5
0.5
0.0
0.0
0.0
0.0
-0.5
-0.5
-0.5
-0.5
-1.0
-1.0
-2
1 2 3 4 5 6 7 8 9 101112
1.6
1 2 3 4 5 6 7 8 9 101112
1.6
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
1.6
-0.8
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
-1.0
1 2 3 4 5 6 7 8 9 101112
-1.0
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
Fig. 6. Impulse responses of domestic FSI to advanced economies, nonregional emerging markets, regional emerging markets, and domestic nancial shock (response to
structural one standard deviation).
21
(i) Hungary
(j) India
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 1011 12
-2
1 2 3 4 5 6 7 8 9 1011 12
1.2
1.2
1.2
1.2
0.8
0.8
0.8
0.8
0.4
0.4
0.4
0.4
0.0
0.0
0.0
0.0
-0.4
-0.4
-0.4
-0.4
-0.8
-0.8
1 2 3 4 5 6 7 8 9 101112
-0.8
-0.8
1 2 3 4 5 6 7 8 9 101112
(k) Indonesia
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(l) Israel
-2
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
-2
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
(m) Rep. of Korea

-2
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(n) Malaysia
-2
-2
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(o) Mexico
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(p) Peru
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
Fig. 6 (continued)
22
(q) Philippines
(r) Poland
-1
-1
-1
-1
-2
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112

1 2 3 4 5 6 7 8 9 101112
-2
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
-2
1 2 3 4 5 6 7 8 9 101112
(s) Romania
-2
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(t) Russian Federation
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112

-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(u) South Africa
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(v) Singapore
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-1
-1
-1
-1
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
-1
-1
-1
-1
-1
-1
-1
-1
-2
-2
-2
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
(w) Taipei, China
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
(x) Thailand
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
-1
-1
-1
-1
-1
-1
-1
-1
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112

-2
1 2 3 4 5 6 7 8 9 101112

1 2 3 4 5 6 7 8 9 101112
-2
-2
1 2 3 4 5 6 7 8 9 101112
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
Fig. 6 (continued)
(y) Turkey
Response to Structural One S.D. Innov ations 2 S.E.
Response of DOM_SUM to Shock1 Response of DOM_SUM to Shock2 Response of DOM_SUM to Shock3 Response of DOM_SUM to Shock4
-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
Response to Structural One S.D. Innov ations 2 S.E.

-1
-1
-1
-1
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
-2
1 2 3 4 5 6 7 8 9 101112
Fig. 6 (continued)
such that domestic FSI responds to advance and emerging market

countries nancial shocks; as well as its own nancial shocks (third
row of matrix A); while its covariance with advanced and other
emerging market nancial shocks is assumed zero (third row of
matrix B). The resulting impulse response functions and variance
decomposition will show the size, speed of adjustment, and persistence of impact of advanced and other emerging market countries
(nonregional and regional) nancial shocks on domestic emerging
market FSI.
To estimate Eqs. (8)(10), an augmented ADF unit root test was
conducted for nancial stress indexes for advanced economies,
emerging market (excluding the particular country), nonregional
emerging market, regional excluding the country, and domestic
FSI using both variance-equal weights and principal component
analysis. The results reject the null hypothesis of a unit root at
0.10 level of signicance for almost all countries. Since most FSI
indicators do not contain unit root, SVAR was estimated using
the level data for all FSI indexes. A granger causality test has also
been done, and the results do not reject the null hypothesis that
domestic FSI does not Granger cause both the advanced and other
emerging market FSIs.
To determine the appropriate lag order to use in the SVAR, lag
structure test was conducted using the nancial predictor error,
Akaike information criterion, Schwarz information criterion, and
HannanQuinn information criterion. The results from the lag order test show that up to two lags of each variable must be included
in the SVAR estimation.
4.2. Impulse responses and variance decompositions
Fig. 5ay presents individual emerging market impulse responses of domestic FSI to a structural one standard deviation
nancial shock coming from advanced economies (Shock 1), other
emerging markets (Shock 2), and its own domestic market (Shock
3) using both variance-equal weights and principal component
analysis. The responses are shown together with their +/ two
standard error values. Table 9 shows the variance decompositions
of domestic FSI on nancial shocks coming from advanced economies (Shock 1), other emerging markets (Shock 2), and its own
domestic market (Shock 3). Fig. 6ay presents individual emerging
market impulse responses of domestic FSI to a structural one standard deviation nancial shock coming from advanced economies
(Shock 1), nonregional emerging markets excluding the country
(Shock 2), regional emerging markets excluding the country (Shock
3), and its own domestic market (Shock 4) using both varianceequal weights and principal component analysis. Table 10 presents
the variance decompositions.
Based on the impulse responses and variance decomposition results, several key ndings are noted. First, the impulse responses
from the SVAR model on the impact of advanced economies, other
emerging markets, and domestic market nancial shocks on
23
domestic FSI (Fig. 5ay) are in line with the panel regression results. Specically, nancial shocks originating from advanced and
other emerging market countries increase individual emerging
market domestic FSI. Among the different sources of nancial
shocks, domestic FSI tends to respond strongly to its own nancial
shock. However, some emerging market domestic FSIs respond faster to a shock from advanced economies than from emerging market economies; while others (mostly emerging Asia) respond more
strongly to a shock from emerging market economies than to from
advanced economies. Hence, there are differences in speed and
magnitude of responses to external nancial shocks across individual emerging market countries.
Second, the variance decompositions in Table 9 show that
domestic nancial shock dominates variation in domestic FSI.
A nancial shock from advanced economies accounts for greater
variation in domestic FSI in the rst two quarters after the shock,
while a shock from emerging market economies accounts for
greater variation after the second quarter. This suggests that
advanced economies nancial shock has more instantaneous but
less persistent impact on domestic nancial stress; while emerging
market nancial shock has delayed but relatively lasting impact on
domestic nancial stress. For emerging Asia and emerging
Americas, a nancial shock originating from emerging markets
consistently accounts for greater variation in domestic FSI than
that coming from advanced countries.
Third, looking into the responses of domestic FSI to nancial
shocks from other emerging countries excluding the region (nonregional) and regional emerging countries excluding the country
(Fig. 6ay), the estimates reveal that a shock from nonregional
emerging markets increase domestic nancial FSI as much as a regional shock or more for many emerging market countries in the
sample. But for most emerging Asia, a shock from regional emerging markets has greater impact than that from nonregional emerging markets.14 The variance decompositions also illustrate that for
emerging Asia, regional (excluding country) nancial shocks explain
considerably larger variation in domestic FSI than nonregional
shocks. These observations are consistent with the panel regression
estimates showing the signicance of a common regional factor for
emerging Asia.
5. Summary and conclusion

This paper extends the literature analyzing the cross-border
transmission of nancial crisis using the FSI. From a policy viewpoint, FSI offers an important tool which can help measure a degree of nancial stress and identify the sources of the stress. This
paper utilizes FSI to understand global, regional, and domestic factors inuencing the nancial stress condition of emerging market
economies. The analysis also includes FSIs in various economic
and geographic groupings of the countries to understand the transmission of nancial shock from these different groups.
Earlier studies indicated that nancial shocks spread from advanced to emerging market economies. While verifying the earlier
ndings, this study highlights that nancial stress originated from
emerging economies also exerts signicant inuence on the nancial stress condition of other emerging market economies. Specically, the panel regression results show that nancial stress
originating from emerging market economies (excluding the country of particular interest) signicantly increases nancial stress in
14
A separate SVAR estimation was conducted by removing the Hong Kong, China
and Singapore (both Asian nancial hub) in the emerging Asia sample, to determine
whether a regional nancial shock will still dominate compared to a nonregional
shock. The impulse response and variance decomposition results show that a
regional nancial shock is still greater than nonregional shock despite removing the
regional nancial hubs.
24
Table 9
Variance decomposition of domestic FSI.
Table 9 (continued)
Period
Period
S.E.
Shock1
Shock2
Shock3
(a) Emerging markets

Financial stress index (equal-variance weights)
1
1.035
11.947
2
1.540
16.000
3
1.791
16.851
4
1.901
16.538
6
1.961
16.248
8
1.972
16.497
12
1.978
16.773
8.920
13.982
18.850
22.462
25.550
26.124
26.208
79.133
70.019
64.299
60.999
58.202
57.379
57.019
Financial stress index (principal

1
0.753
2
1.019
3
1.151
4
1.213
6
1.251
8
1.258
12
1.261
8.708
13.050
17.332
20.185
22.354
22.698
22.734
83.899
77.499
72.922
69.919
67.331
66.598
66.218
component)
7.393
9.451
9.746
9.897
10.315
10.705
11.048
S.E.
Shock1
Shock3
3.625
7.702
13.313
18.221
23.596
25.161
25.523
76.644
69.862
63.967
59.419
54.884
53.617
53.285
Financial stress index (principal component)

1
0.760
8.787
2
1.023
9.892
3
1.155
9.644
4
1.213
9.301
6
1.245
9.041
8
1.249
9.071
12
1.249
9.160
4.262
9.445
13.195
15.383
17.078
17.429
17.486
86.951
80.664
77.161
75.316
73.881
73.501
73.353
Note: Factoralization is structural. Values refer to unweighted average of individual

country variance decomposition.
Source: Authors calculation.
(b) Emerging Asia

1
1.025
8.665
2
1.527
10.860
3
1.778
10.545
4
1.882
10.024
6
1.928
10.495
8
1.932
11.401
12
1.935
12.139
14.426
19.442
23.491
25.930
27.169
26.932
26.639
76.909
69.698
65.964
64.046
62.336
61.667
61.222

1
0.748
6.811
2
1.004
8.141
3
1.132
8.061
4
1.187
8.193
6
1.217
9.041
8
1.221
9.713
12
1.222
10.175
8.897
11.672
14.679
16.539
17.628
17.748
17.805
84.291
80.187
77.259
75.268
73.331
72.539
72.020
(c) Emerging Americas

1
1.025
8.797
2
1.515
11.458
3
1.751
11.904
4
1.859
11.377
6
1.937
10.990
8
1.954
11.294
12
1.962
11.619
8.857
14.512
20.420
25.532
30.715
32.152
32.644
82.345
74.030
67.676
63.091
58.296
56.554
55.737

1
0.750
4.674
2
1.014
7.791
3
1.141
7.558
4
1.202
7.243
6
1.248
7.031
8
1.258
7.125
12
1.263
7.290
11.936
17.142
23.400
27.677
31.305
31.970
32.119
83.390
75.067
69.042
65.080
61.664
60.905
60.592
(d) Emerging Europe

1
1.064
16.064
2
1.591
26.582
3
1.862
30.703
4
1.987
31.104
6
2.065
29.848
8
2.083
29.152
12
2.095
28.692
2.218
7.447
12.114
15.235
17.678
18.043
18.169
81.719
65.971
57.182
53.661
52.474
52.805
53.139
Financial stress index (principal

1
0.761
2
1.050
3
1.200
4
1.278
6
1.330
8
1.340
12
1.346
8.012
13.779
18.664
22.327
25.286
25.686
25.530
81.282
72.513
65.514
60.708
56.890
56.023
55.658
component)
10.706
13.708
15.821
16.965
17.823
18.291
18.812
Shock2
(e) Other emerging markets

1
1.041
19.731
2
1.545
22.436
3
1.796
22.720
4
1.901
22.360
6
1.953
21.521
8
1.960
21.222
12
1.963
21.192
domestic nancial systems of the (excluded) emerging economy.

Whether this shock from emerging market economies is from the
same region or not did not seem to matter. The SVAR impulse
responses and variance decompositions reveal that the impact of
a nonregional shock is as strong as that of a regional shock or
stronger on domestic FSI for most emerging market economies in
the sample, except the ones in emerging Asia.
Another important nding of this study is the importance of a
common regional factor in affecting domestic nancial stress.
The panel regression results indicate that a common regional factor
signicantly increases domestic nancial stress in emerging Asia
and emerging Europe. The variance decomposition in Table 10
indicate that regional emerging market nancial shocks have lasting impact in emerging Asia, which corroborates the implications
of the panel regression results.
The ndings of this paper encourage future researchers in the
eld to consider the importance of nancial shocks originating
not only from advanced economies but also from emerging market
economies. Given the increasing importance of emerging market
economies in the global economic and nancial scene, it is natural
to expect increasing inuence of emerging market economies in
international transmission of shocks.
Together with the trend in nancial integration, the ndings of
nancial contagion present a cautionary tale. While nancial
activity is increasingly global, regulation remains extremely national. The growing cross-border nature of banking and nancial
services underscores the need for a coordinated oversight of international nancial institutions and markets. Without an integrated
global framework, regulatory arbitrage and inadequate cross-border information and data sharing would continue to expose national and international nancial systems to nancial risks and
instability (Moshirian, 2011). The global nancial crisis prompted
a wave of reforms in the global nancial architecture. While the
institution of global regulator is realistically difcult, there have
been efforts to establish a global framework that will mandate
minimum consistency across jurisdictions in regulatory principles
that would apply to similar markets, institutions, services, and
products. What is also important is that the regulatory oversight
should be more inclusive of emerging market economies given
the growing importance of emerging market nancial systems.
25
Table 10
Variance decomposition of domestic FSI.
Table 10 (continued)
Period
Period
S.E.
Shock1
Shock2
Shock3
Shock4
(a) Emerging markets

Financial Stress Index (Equal-Variance Weights)
1
1.036
13.374
5.170
2
1.551
17.235
8.967
3
1.803
17.654
12.506
4
1.914
16.964
15.491
6
1.981
16.339
18.270
8
1.995
16.599
18.823
12
2.006
16.967
18.873
7.245
9.322
11.202
12.619
14.100
14.681
15.046
74.211
64.475
58.638
54.926
51.291
49.897
49.113

1
0.756
7.933
5.680
2
1.029
9.836
8.693
3
1.166
10.113
11.582
4
1.230
10.078
13.792
6
1.271
10.222
15.605
8
1.279
10.465
15.870
12
1.282
10.665
15.827
7.243
9.259
11.191
12.527
13.731
14.178
14.505
79.144
72.211
67.115
63.603
60.442
59.487
59.003
Shock2
Shock3
Shock4
(e) Other emerging markets

1
1.047
22.285
2.237
2
1.563
25.733
6.466
3
1.822
25.962
11.011
4
1.927
24.942
14.736
6
1.979
23.567
18.675
8
1.990
23.163
19.874
12
1.995
23.146
20.202
S.E.
Shock1
0.652
0.946
3.411
5.579
7.386
8.405
9.297
74.826
66.855
59.616
54.744
50.372
48.558
47.355

1
0.756
9.110
4.382
2
1.021
10.313
10.151
3
1.165
9.938
13.182
4
1.233
9.415
15.218
6
1.269
9.131
16.603
8
1.272
9.198
16.841
12
1.273
9.277
16.929
0.941
0.868
2.486
3.846
5.263
5.620
5.731
85.566
78.668
74.394
71.521
69.002
68.341
68.063
Note: Factoralization is structural. Values refer to unweighted average of individual

country variance decomposition.
Source: Authors calculation.
(b) Emerging Asia

1
1.037
9.457
7.029
2
1.550
11.761
8.660
3
1.802
11.340
10.420
4
1.906
10.634
11.678
6
1.953
10.979
12.442
8
1.958
12.103
12.408
12
1.961
12.909
12.376
12.097
17.302
20.818
22.859
24.070
23.859
23.544
71.417
62.277
57.423
54.829
52.510
51.630
51.171

1
0.760
7.189
4.353
2
1.017
8.413
4.049
3
1.147
8.390
4.837
4
1.204
8.457
5.505
6
1.234
9.284
5.941
8
1.238
10.031
5.971
12
1.239
10.473
5.990
8.118
13.547
16.872
18.884
20.165
20.317
20.355
80.340
73.991
69.902
67.154
64.611
63.680
63.182
(c) Emerging Americas

1
1.010
10.153
7.187
2
1.509
13.285
12.101
3
1.750
13.532
17.885
4
1.866
12.718
23.360
6
1.960
11.801
29.384
8
1.981
11.814
31.197
12
1.992
12.023
31.785
6.144
5.638
5.551
5.297
4.917
4.982
5.170
76.515
68.976
63.032
58.625
53.899
52.007
51.022

1
0.744
5.616
9.132
2
1.025
8.703
12.992
3
1.151
8.375
18.481
4
1.215
7.910
22.726
6
1.268
7.431
26.912
8
1.280
7.413
27.725
12
1.285
7.550
27.730
8.458
7.888
7.735
7.449
7.238
7.585
8.075
76.795
70.418
65.408
61.915
58.419
57.277
56.645
(d) Emerging Europe

1
1.059
17.946
1.376
2
1.596
26.126
7.824
3
1.854
28.582
11.420
4
1.975
28.334
14.281
6
2.063
26.720
16.267
8
2.088
26.084
15.963
12
2.119
26.076
15.308
4.133
4.484
4.984
6.560
10.553
12.984
14.503
76.545
61.566
55.013
50.825
46.459
44.969
44.114

1
0.762
11.261
5.231
2
1.065
13.659
11.658
3
1.221
15.783
15.513
4
1.298
16.453
18.504
6
1.352
16.318
20.568
8
1.365
16.007
20.663
12
1.372
15.898
20.334
9.078
9.043
10.938
12.853
15.430
16.659
17.538
74.430
65.640
57.766
52.190
47.684
46.671
46.229
Deepening regional integration in many developing regions

imply there are additional benets in creating responsible regional
institutions and regulators for monitoring of regional market
conditions and maintaining nancial stability at the regional level
in coordination with national and global ones. The ndings also
suggest there may be a common regional factor in determining
the domestic nancial stress condition. A shock emerging from a
country may affect the countries in the same region more than
those of other regions, due to similar fundamentals, integrated regional trade, nancial and investment network, and/or similar responses by international investors. Especially in emerging Asia,
where regional integration has made substantial progress, a number of nancial issues and activities require regionally coordinated
responses to improve effectiveness of these actions as well as to
avoid adverse consequences of unilateral national responses.
References
Afonso, A., Baxa, J., Slavik, M., 2011. Fiscal Developments and Financial Stress: A
Threshold VAR Analysis. ECB Working Paper Series No. 1319. ECB, Frankfurt am
Main.
Aizenman, J., Pasricha, G.K., 2010. Determinants of Financial Stress and Recovery
during the Great Recession. NBER Working Paper Series No. 16605. NBER,
Massachusetts.
Allen, L., Bali, T., Tang, Y., 2012. Does systemic risk in nancial sector predict future
economic downturns? Review of Financial Studies 25 (10), 30003036.
Arnold, B., Borio, C., Ellis, L., Moshirian, F., 2012. Systemic risk, macroeconomic
policy frameworks, monitoring nancial systems, and the evolution of capital
adequacy. Journal of Banking and Finance 36, 31253132.
Balakrishnan, R., Danninger, S., Elekdag, S., Tytell, I., 2009. How Linkages Fuel the
Fire: The Transmission of Financial Stress from Advanced to Emerging
Economies. World Economic Outlook (April 2009). International Monetary
Fund, Washington, DC.
Balakrishnan, R., Danninger, S., Elekdag, S., Tytell, I., 2011. The transmission of
nancial stress from advanced and emerging economies. Emerging Markets
Finance and Trade 47, 4068.
Bernanke, B., 1986. Alternative explorations of the moneyincome correlation.
Carnegie-Rochester Series on Public Policy 25, 4999.
Blanchard, O., Watson, M., 1986. Are business cycles all alike? In: Gordon, Robert
(Ed.), The American Business Cycle: Continuity and Change. University of
Chicago Press, Chicago.
Bollerslev, T., Chou, R., Kroner, K., 1992. ARCH modeling in nance: a review of the
theory and empirical evidence. Journal of Econometrics 52, 559.
Borio, C., Drehmann, M., 2009. Towards an Operational Framework for Financial
Stability: Fuzzy Measurement and its Consequences. Bank for International
Settlements Working Paper No. 284.
Calvo, S., Reinhart, C., 1996. Capital ows to Latin America: is there evidence of
contagion effects? In: Calvo, Goldstein, Hochreiter (Eds.), Private Capital Flows
26
to Emerging Markets after the Mexican Crisis. Peterson Institute for

International Economics.
Cardarelli, R., Elekdag, S., Lall, S., 2008. Financial Stress and Economic Downturns.
In: World Economic Outlook (October 2008). Financial Stress, Downturns, and
Recoveries. International Monetary Fund, Washington, DC.
Cardarelli, R., Elekdag, S., Lall, S., 2011. Financial stress and economic contractions.
Journal of Financial Stability 7, 7897.
Davig, T., Hakkio, C., 2010. What is the effect of nancial stress on economic
activity? Kansas City Fed Economic Review Second Quarter, 3562.
Dornbusch, R., Park, Y.C., Claessens, S., 2000. Contagion: understanding how it
spreads. World Bank Research Observer 15, 177197.
Dovern, J., van Roye, B. 2013. International Transmission of Financial Stress:
Evidence from GVAR. Kiel Working Papers No. 1844. Kiel Institute for World
Economy, Kiel.
Eichengreen, B., Rose, A., Wyplosz, C., 2004. Contagious currency crises. In:
Eichengreen, B. (Ed.), Capital Flows and Crises. Massachusetts Institute of
Technology Press.
Eicher, S., Karmann, A., Maltritz, D., 2009. The ADR shadow exchange rate as an
early warning indicator for currency crises. Journal of Banking and Finance 33,
19831995.
Fernandez, V., 2007. Stock market turmoil: worldwide effects of the Middle East
conicts. Emerging Markets Finance and Trade 43, 58102.
Hakkio, C., Keeton, W., 2009. Financial Stress: What is it, How can it be Measured,
and Why Does it Matter? Federal Reserve Bank of Kansas City.
Hanschel, E., Monnin, P., 2005. Measuring and Forecasting Stress in the Banking
Sector: Evidence from Switzerland. BIS Working Paper No. 22.
Hollo, D., Kremer, M., Lo, M., Duca, M., 2012. CISS A Composite Indicator of
Systemic Stress in the Financial System. European Central Bank Working Paper
Series No. 1426. Frankfurt am Main.
Illing, M., Liu, Y., 2006. Measuring nancial stress in a developed country: an
application to Canada. Journal of Financial Stability 2, 243265.
Kaminsky, G., Reinhart, C., 1999. The twin crises: the causes of banking and balanceof-payments problems. American Economic Review 89, 473500.
Kaminsky, G., Reinhart, C., 2000. On crises, contagion, and confusion. Journal of
International Economics 51, 145168.
Kose, A.M., Prasaad, E., Rogoff, K., Wei, S., 2006. Financial Globalization: A
Reappraisal. IMF Working Paper.
Kunt, A., Detragiache, E., 1996. The determinants of banking crises in developed and
developing countries. International Monetary Fund Staff Papers 45, 81109.
Laeven, L., Valencia, F., 2008. Systemic Banking Crises: A New Database.
International Monetary Fund Working Paper WP/08/224.
Misina, M., Tkacz, G., 2009. Credit, asset prices, and nancial stress. International
Journal of Central Banking 5, 95122.
Moser, T., 2003. What is international nancial contagion? International Finance 6,
157178.
Moshirian, F., 2008. Globalization, growth and institutions. Journal of Banking and
Finance 32, 472479.
Moshirian, F., 2011. The global nancial crisis and the evolution of markets,
institutions and regulation. Journal of Banking and Finance 35, 502511.
Ng, T., 2011. The Predictive Content of Financial Cycle Measures for Output
Fluctuations. Bank for International Settlements Quarterly Review.
Oet, M., Bianco, T., Gramlich, D., Ong, S., 2011. Financial Stress Index: A Lens for
Supervising the Financial System. Federal Reserve Bank of Cleveland Working
Paper No. 12/37.
Reinhart, C., Rogoff, K., 2008. This Time is Different: A Panoramic View of Eight
Centuries of Financial Crises. NBER Working Paper Series No. 13882. National
Bureau of Economic Research.
Sims, C., 1986. Are forecasting models usable for policy analysis? Quarterly Review
of the Federal Reserve Bank of Minneapolis 10, 216.
Stulz, R.M., 2005. The limits of nancial globalization. Journal of Finance 60, 1595
1638.
Van Roye, B., 2011. Financial Stress and Economic Activity in Germany and Euro
Area. Kiel Institute Working Paper No. 1743. Kiel Institute for World Economy.
Yiu, M., Ho, W.-Y., Jin, L., 2010. A Measure of Financial Stress in Hong Kong Financial
MarketThe Financial Stress Index. HKMA Research Note 02/2010. Hong Kong
Monetary Authority.

Determinants of Financial Stress in Emerging Market Economies

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Determinants of Financial Stress in Emerging Market Economies

Uploaded by

Copyright:

Available Formats

Journal of Banking & Finance xxx (2013) xxxxxx

Contents lists available at ScienceDirect

Journal of Banking & Finance

Determinants of nancial stress in emerging market economies

Corresponding author. Tel.: +63 2 632 4444.

may induce similar response to a shock, which leads to strong

2. Financial stress index

emerging economies and used at the subsequent issue of IMFs

sion of nancial stress. Balakrishnan et al. (2011) created FSI for

shock. We look into the signicance of a common regional factor as

This paper follows the denition of nancial stress for emerging

2.2.1. Components and variable choices

(d) Emerging Americas

(a) Emerging Economies

(b) Advanced Economies

(e) Emerging Europe

(c) Emerging Asia

(f) Other Emerging Markets

beta is not a measure of co-movement between the two variables.

Source: Authors calculations.

where De and DRES denote month-on-month percent changes in

r2t x /1 e2t1 /2 r2t1

yt ai;t byt1 ei;t

Expected Variance-equal weights

Regional FSI (excl

Global GDP growth

Dummy emerging Asia

where yt is the demeaned and standardized series, x the mean of the

Advanced economies FSI

Emerging economies FSI (excl country)

Emerging economies FSI (excl region)

Regional FSI (excl country)

Global GDP growth

Global commodity price change

DForeign exchange reserves (t 1)

Another popular approach is the use of principal component

EMFSI b Stockreturns Stockv olatility Debtspreads EMPI

Advanced economies FSI

Emerging Economies FSI (excl country)

Emerging economies FSI (excl region)

Regional FSI (excl country)

Global GDP growth

Global commodity price change

DForeign exchange reserves (t 1)

However, several points are noted on the identied stressful periods.

Advanced economies FSI

Emerging economies FSI (excl country)

Emerging economies FSI (excl region)

Regional FSI (excl country)

Global GDP growth

Global commodity price change

DForeign exchange reserves (t 1)

1998 episode, nancial crises in Latin America and the Russia

where almost all emerging market FSIs rose. In addition, there

Advanced economies FSI

Emerging economies FSI (excl country)

Emerging economies FSI (excl region)

Regional FSI (excl country)

Global GDP growth

Global commodity price change

DForeign exchange reserves (t 1)

degree of nancial integration, it is also crucial to know whether

to quarterly series beginning Q1 1992Q4 2012 using the average

Advanced economies FSI

where yt is the demeaned and standardized series, x the mean of the