Measuring contagion risk among vietnam’s listed commercial banks

Policies and Sustainable Economic
Development | 631

Measuring Contagion Risk among
Vietnam’s Listed Commercial Banks
NGUYEN THI MINH HUE
National Economics University - /

NGUYEN THE TUNG
National Economics University

TRAN DANG MINH
National Economics University

NGUYEN HUU QUANG
National Economics University

LAM HONG PHONG
National Economics University

TRAN MINH KHOI
National Economics University

Abstract
Commercial banking system is a dominant sector in the Vietnamese economy.
The weakness and failure of one particular bank may have a big impact on the
banking system as well as the economy. Given the case study of Lehman Brother
Holding Inc. (2008) beyond the United State and the worldwide economies,
contagion risk among commercial banks has been an intriguing issue in the
literature. During the recession period of Vietnam, there is a need to restructure
the banking system and to discover potential risks among commercial banks.

Due to the data unavailability, we follow Gropp and Moerman (2004) to raise
awareness of the contagion risk across eight listed commercial banks of Vietnam.
Particularly, we try to measure how other banks are affected when one bank falls
into the extreme event and finds the signal of contagion risk. The result of this
paper may lead to implication of the overall picture of the banking system. The
names of the listed bank are coded randomly to protect the confidential
information of the banks. Three of them (D, E, H) are showing extraordinary
vulnerability, whereas two others (B, C) are more stable. While the strongest
resistance to contagion is found for the case of Bank F, showing the lowest
interconnectedness for both “co-exceedances” and “Granger” test, the other
banks in the sample reveal controversial results. To conclude, banks are
susceptible to others’ operating situation, and the research should be expanded
throughout the whole system with the legal enhancement of data transparency
and adequacy.

Keywords: contagion risk; banks; Vietnam; signal

1. Introduction


632 | Policies and Sustainable Economic Development

In recent year, contagion risk is becoming a hot trend globally and has
been debated intensely ever since. The world economic are getting more
and
more
integrated,
financial
institutions
increased

their
interconnectedness to each other. It is needless to say that this integration
provides us with huge advantages of labor, information and cost savings,
which ultimately leads to improvement of the liquidity and depth of the
market. However, with great advantages come great obstacles to
overcome. In this situation, it is the case of “contagion” in the system.
The collapse of the Lehman Empire which led to 2008 crisis is one
example of contagion effect. If this is the situation of countries in the first
world-group, it becomes direr when it comes to developing countries such
as Vietnam. Thus, in this paper, we attempts to make a contribution to
measure the contagion risk among bank of Vietnam. We will be discussing
a methodology by Gropp and Moerman (2004) to locate the direction of
contagion from one institution to another. We propose to use market data,
especially the distance to default, to examine bank contagion. The
approach based on the use of “co-exceedance” and extreme value theory
in examining contagion risk among institutions.
In this paper, for the first part, we will discuss about different literature
on the subject of contagion and methods to measure contagion risk. In the
second part, we will show the detail method that was applied in this paper.
The third part will be focused on analyzing result of the research, and for
the final part: conclusion and implication for stakeholders.
2. Literature review
The term financial contagion has created controversy throughout the
past years. Until now, there has not been an official announcement or
mutual agreement upon this matter. Thus, to what degree responsibility
should be assigned to contagion in financial crisis is still ambiguous. Some
analysts even believe that there is no contagion but only the worsening
economic conditions that cause depositors to take their money out of
weak banks and put it into healthier ones. Taking the situation of this
research into concern, we decide on the definition of contagion as “one

bank being hit by an idiosyncratic shock that is also transmitted to other
banks” as suggested in the work of Gropp in 2004 to be likely the most
feasible term to use for the purpose of investigating the regional
contagion. We will not specify channel of transmission, but one could
imagine money markets, payment systems, equity (ownership) links and
“pure” contagion. The definition could avoid problem of overstate or
understate the effects of contagion. Its advantages allow us to develop a
method to identify the direction of contagious influence among banks
rather than the degree of contagion effects on financial system. As such,
in this paper, we avoid talking about the exact quantities but only try to
measure the extent to which Vietnamese banking system has become
interconnected and how a problem of an individual bank could spread
across our market. The methods debated bellows will follow what we have
previously stated as definition.


Policies and Sustainable Economic
Development | 633

In some papers, contagion is time to time referred to the
interconnectedness as these two terms is closely related even when there
are certain differences between them. Therefore in this work we will not
specifically clarify the two of them but to use them interchangeably.
We employ the Model taken from the study of Gropp and Moerman
(2004) as the basis for this paper. The work is based on Longin and
Solnik’s (2001) and Bae et al.’s (2003) evidence that the observed coexceedances (i.e. the presence of two or more banks in the tail of the
distribution simultaneously) are consistent with a different simulated
distribution which suggests it could be a reliable method to measure
contagion risk. The co-exceedances mentioned in the paper is calculated
on the basis of Distance to Default (DD), which in Gropp et al. (2003), was

shown as a particularly suitable way to measure bank risk. The DD is
outstanding from other risk measurement as it could avoid problems such
as subordinated debt spreads. The distance to default combines
information on stock price returns with asset volatility and leverage and
represents the number of standard deviations away from the default point.
The default point is defined as the point at which the liabilities of the bank
are just equal to the assets.
For the case of Vietnam, we choose this method as we deemed it to be
the most suitable considering our situation. Not only is it proved to be a
relatively effective way of measuring contagion, there are high chances
that this method can be applied on the market of Vietnam. Our domestic
stock market has developed enough for data and information in full
disclosure to be found. Furthermore, most of the biggest banks
domestically have been listed on the stock market. It is possible for us to
collect sufficient data for calculation of the co-exceedance.
Through the studying from other studies, we were able to get some
insight and by conducting our own one, we can say that it is different from
the previous papers. First of all, the researches in Vietnam only focus on
systemic risk but the one focusing on contagion risk is nearly nowhere to
be found. In a developing country such as our, the new term of contagion
risk may be deemed as a relative new concept. Through gathering
information, in our knowledge, this paper may be the first research that
tries to measure the effect of contagion in the system of Vietnam banking.
Secondly, the method we would be using to measure bank risk is called
the distance to default. This combines the information on leverage, asset
volatility with data of full exposure which is superior to those
measurements of risk that have previously been done domestically.
3. Methodology and research model
We employ the model taken from the study of Gropp and Moerman
(2004) as the basis for this paper. The approach is related to the

suggestion of Bae et al. (2003) that the behavior of tail observations for
financial market data is quite different from the behavior of other
observations (extreme value theory).
We empirically examine contagion in sample of eight listed commercial
banks in Vietnam. These banks were coded ramdomly as A, B, C, D, E, F, G
and H. For each bank, we will calculate their


634 | Policies and Sustainable Economic Development

distances-to-default (DD) by days using market data. The distance to
defaults was chosen, as in Gropp et al. (2002) argued that specifically with
respect to banks, DD may be a particularly suitable and all-encompassing
measure of bank risk. As it combines both the information about stock
returns with leverage and volatility information, it could avoid problems of
other measures (like subordinated debt spreads or unadjusted stock
returns). The distance to default of each bank can be calculated following
the Black-Scholes model.
After calculating the distance to default of each bank, we then obtain
the change in DD by taking ln(ddt/ddt-1), using the weekly first different to
default which in the following will be denoted as ln(∆dd). Hence, ln(∆dd)
measures the percentage change in the number of standard deviations
away from the default point.
For the next part, we will test the banks contagion risk using a concept
called “co-exceedance” (Bae et al., 2003). We define a co-exceedance as
period (in this paper: a week) during which two or more banks first
difference in the distance to default was in “extreme value”. The more
times a bank in a co-exceedance, the more likely it can create contagion
risk on others. We will implement the test for the “extreme value” of 5%
bottom tail. As our database is quite small, we decided to extend the test

to 10th percentile to provide a more comprehensive view about the
contagion risk in banks of Vietnam. While the 5% bottom tail has
significant implication in predicting the effect of the risk in recession
period, the co-exceedances of the 5% top tail of the distribution could give
us an overview about the trend of contagion. Thus, we also perform the
calculation for 5 and 10% top tail (also known as 90th and 95th
percentile).
In order to test whether the result of co-exceedances can explain the
contagion risk, we then use the Granger-causality (GC) method suggested
in Chan-Lau et al. (2007) as the robust of the model. We implement a set
of pairwise tests on the 8 banks to determine the broad trends in the
ln(ΔDD) relationships between pairs of banks. “While this could be
captured by calculating the rolling correlations between the ln(ΔDD)s, GC
tests go beyond standard correlations; they show whether the past
ln(ΔDD)s of a particular bank help to explain the current ln(ΔDD)s of
another bank, after also taking into account the past ln(ΔDD)s of the
latter” (Chan-Lau et al., 2007). The Granger-causality test may not as
reliable as the Monte Carlo test, as correlations in bank soundness during
normal times provide little information on the likelihood of contagion.
However, it could give us an overview look about the Vietnam market.
4. Contagion risk results

4.1. Co-exceedances and identification of contagion risk
Through the use of co-exceedances, there are possibilities to apprehend
the direction of contagious influence among banks. For the research
purpose of clearly understood the situation, we will proceed to calculate
and analyze the result for co-exceedances of the eight selected banks in
the quantile of 5% and 10% respectively.

Policies and Sustainable Economic
Development | 635

Table 1
Overall descriptive statistics of the research

Distance to Default
Log-differenced DD (weekly 1st different)
Number of observations (weeks)
Number of tail observations (5%)
Number of tail observations (10%)

Table 2
Summary statistics of (co-) exceedances for weekly log-differenced
distance to defaults for 8 individual banks (5% bottom tail)

>6
A

1

B

1

C

1

D

1

E

1

F

1

G

1

H

1

Table 2 shows the number of times one bank co-exceedances with other
banks in the whole period. For example, looking at second column of Bank
A, it means that during the period of observation, the bank has two weeks
exceedance with five banks, including itself.
The table indicates the number of co-exceedances at the 5th quantile
among 8 banks. As we expected, the bank G, with the highest standard
deviation of ln(∆dd), is the bank with the most exceedances in the bottom
tails with 20 weeks in the tail distribution. This means that the bank G is
rather unstable relative to others, its changes in distance to default has
declined below the 5th quantile more than any other banks has. However,
when taken contagion into account, the bank G is not the most vulnerable

bank. Half of the time of its exceedances, there is only one bank that is
also in the tail event at the same time with the bank G, ultimately it leads
to only 10 weeks of co-exceedances between the bank G and the other
two and more banks
Besides, the bank D, with the record of 12 exceedances, there are 11
times the bank in the tail event when three or more banks are also in the
same situation (co-exceedances), which is relatively high compared to the
result of others. This suggested the high level of interconnectedness of the


bank D with other banks which means the need for its extra regulation in
term of contagion.


636 | Policies and Sustainable Economic Development

The result for the bank B and C are also match with what we predicted.
Among six banks with about 300 observations, those banks witness the
two low numbers of weeks of co-exceedances, only co-exceedances of
more than 2 banks for seven times. Notably, the bank C, although is one of
the three banks with the largest total assets; ranked last position among
the six banks. This could hint that the size of the bank isn’t the main
source of contagion.
From all 8 banks, bank F is the bank that most unlikely to be involved in
contagion, with the lowest number of both exceedances and coexceedances, 5 and 3 respectively. Some may argue that this low record is
the result from the fact that their number of observations is only 193
compare to approximately 300 of other banks. Nevertheless, taking note
that its’ counted number of co-exceedances is even lower than that of the
bank I, which is only observed for 76 weeks.
Only observed for such a short period, the observation time for the bank

A is quite low compare to other banks, thus the result may be not much of
significant. Despite that, the bank A still shows itself as an influential bank.
Considering the five times co-exceedances in only less than a quarter
period of observation of other banks, we believe this bank shows a
potential risk for contagion.
In conclusion, the tables illustrate that there is significant
interconnectedness among banks although the magnitude of it is rather
unsubstantial. In order to make up for this shortcoming one way or
another, we also conducted the test on the 10% tails. However, as it
represents smaller shock, the result for the risk of contagion could not be
varied and not as precise as the 5% threshold. Result for the 10% test is
available in Appendix
The following part will examine the specific bank-to-bank contagion
using the cross-table for number of pair co-exceedances. The result shown
in the cross table could give us a better view about the direction of
contagious influence among Vietnamese banks.
Table 3
Cross table for number of times 2 banks in 5% bottom tails of 8 individual
banks
Number of times 2 banks in 5% bottom tails
A
A

-

B

2

C

3

D

5

E

2

F

2

G

2

H

3

Total

We can see the details in Table 3. For example, bank B and A has two
weeks that they are both in the 5% bottom tail. Still, from the table, it
seems like bank D is showing potential in creating

19

Policies and Sustainable Economic
Development | 637

contagion on a large scale as the bank are co-exceeded with others for no
least than three times each. It also holds the record for the highest time
that the bank co-exceeded with others. For instance, seven weeks with the
bank H. As the average number of pair co-exceedances is four, it suggests
a very close relationship between the two banks. The only other pair that
has the same number is bank E-G.
Two pair of banks that have the lowest number of week in tails
distribution together is bank F-C and bank F-G. This one strengthens the
evidence for low contagion risk in the bank F.
There is no pair that has zero co-exceedances. Overall, the pair coexceedances are ranged from three to four, which in turn suggest a decent
correlation of banks in Vietnam financial system.
As we have mentioned before, we want to closely examine whether the
contagion risk among big banks is higher than the rest. However, the
result again seems to reject our hypothesis. The pair co-exceedances for
bank A, C and D do not seem to show much of a problem, with three and
four for bank C-A and bank C-D respectively. The result for pair of bank D-A
is somehow more significant, with five week of co-exceedances, which in
turn suggests a real threat of contagion between these two banks.
Regardless of the fact, this alone is still not enough to prove for any
relationship between banks size and contagion.
To sum things up, the average pair co-exceedances are ranged from
three to four. For any pair with more than four co-exceedances, we believe
these two should receive extra regulation in term of contagion risk.
However, as same as the problem we faced in the previous part; the
magnitude of co-exceedances is quite low that made the outcome shown

above rather ambiguous.
We also conducted the test on the 10% tails . However, as it represents
smaller shock, the result for the risk of contagion could not be varied and
not as precise as the 5% threshold. Result for the 10% test is available in
Appendix
4.2. Robustness test: Granger-causality
During the Granger test, it appears that the bank D is showing quite
substantial “causality” with significant of less than 1% with all other banks.
Once again, bank D is showing proof to be the bank that to be the most likely
to succumb to contagion risk among the others documented in the sample.

While the bank D seems to be in the red light, the bank E is also not in
the safe zone either. The bank has “causality” with six other banks,
following closely to the bank D and taking the second out of those banks
to be the most exposed to the danger. We could not deny the vulnerability
of the bank to contagion.
With the result almost identical to the previous one, the bank H is
causing “granger” to four banks and at the same time, being “caused” by
the rest of the banks taken into the test.


638 | Policies and Sustainable Economic Development

The pairs of bank C-D and bank H-D show result not out of our
expectation. These two pairs witness the two-way “causality”, which in
turn point out that these banks are having high chances of causing
contagion to each other.
The test also suggests the bank A is not so likely to expose to contagion
which is close to the result in 10% but opposite to our study in the 5% tail. On
the other hand, the case of bank G emerges to be the most interesting. The

finding in the Granger test gives out rather different comparing to what we
have discovered throughout the previous testing which gradually suggests a
low risk of contagion. In addition, the test also implies that the contagion risk
could only be transferred from the bank E to bank G but not in versa. This
case should be considered further in future studies.

Table 4
Pairwise Granger-Causality between 8 Commercial banks of Vietnam December
2009 – January 2016
Banks

A

A
B
C
D

X

E
F
G
H

Factors in the columns are Granger-caused by factors in the rows. The
Granger-causality test used 30 lags. The F-statistic of the joint significance
of the lags of bank 2 in explaining bank 1 is noted in the matrix, an “X ” is
noted if the F-statistic shows significant GC at 1 percent or less. In short,
“X” denotes that “the row” Granger-cause “the column.”

5. Implication and conclusion

5.1. Implication about the contagion risks among Vietnamese
commercial banks
Throughout our examination, the eight banks are showing phenomenon
signs. The financial system is showing trend of interconnectedness
domestically, though vividly. The evidence concluded from the result can
only be verified by testing with difference assumption of the distribution
(we suggest the use of Monte Carlo model). However, throughout the
research we can still see the hint of movements from contagion. Despite
that as the fact, some banks are showing extraordinary vulnerability. The
bank D is taking the lead as the bank to be the most easily to be affected
by contagion. Closely standing beside it is the bank E, showing intense
weakness to the risk. The bank H, while being not as in danger as the two
previous banks, is also not in the clear either. These three banks would
better be preparing for further measurements and attention.


Policies and Sustainable Economic
Development | 639

The bank G and bank A are proving to be the most interesting cases.
The bank G is showing quite dangerous result during the use of the model
but through the Granger test, the bank tends to be not as much being in
the danger area. We recommend further research being done on the
matter as for now, the situation is uncertain. The bank A is showing
significant result to be in the red as much as the rest of them. During the
5% quantile, contagion tends to be the most likely to be in concern.
However, during the 10% threshold and in “Granger causality”, it said to
be otherwise. There is hardly any sign of contagion for the bank. For this

reason, we conclude that the banks’ exposure is still unclear.
The two banks B and C are proving to be most stable. Healthy as they
are, their resistance is quite remarkable. The last but not least, the bank
that has the strongest resistance to contagion must be bank F, showing a
lowest contagion risk for both “co-exceedances” and “Granger” test.
Throughout the research, we could conclude that, the risk of contagion
in Vietnamese banking system is a certain possibility. According to the
tests result, the selected banks have shown quite a number of coexceedances. However, the experiences of measuring contagion risk have
not being clear in other countries, which leads to the lack of benchmark for
Vietnam. This ultimately put the situation of Vietnam regarding contagion
was still in the mist.
5.2. Implication to Commercial Banks in Vietnam
From our study, we figure out that among 8 banks, bank D is the bank
that vulnerable to contagion the most. It has the potential to create
contagion on most of the others. Thus, I believe the bank should receive
extra regulation for their activities. Bank D should be more cautious with it
banking business to prevent any cause of contagion.
The same thing should be applied for the bank H and bank E as these
two are also shown a high risk of contagion. These banks should be more
careful about its activities as well as keep a close eye to their
counterparties.
As the result for the bank G and bank A is rather inconsistence, we
cannot clearly say anything about these two. However, they should not let
their guard down or else, somehow in the future, they may find
themselves as the victims of the contagion.
The overall studies suggest a low risk of contagion for bank B and bank C,
so currently, these bank don’t have to worry much about the issue. Although
the bank C is in the safe zone regarding contagion, we found some other
aspect worth looking into. It is shown in the result that the distance to default
of the bank is decreasing. The risk of bankruptcy is coming closer and closer

as we are discussing and the bank would better focus on this problem for now.
The bank B is doing quite well compare to its recent past. The distance to
default of this bank used to be seen as the lowest and it had been struggling
to get out of this situation. However, nowadays it appears that the bank B is
improving its performance day by day. The case of bank F is also quite good.
Its standing among eight banks is safest from the risk of contagion. Though
the bank is not likely to face the problem


640 | Policies and Sustainable Economic Development

anytime soon, it brings no harm to come up with some prevention
methods to minimize the damages if needed.
5.3. Conclusion
In conclusion, we believed this paper has achieved our initial goal of
making a contribution to recognize the contagion risk among banks of
Vietnam. There are some syndromes for contagion that can only be
verified by testing with difference assumption of the distribution, which we
suggest the use of Monte Carlo model. Throughout the research, we can
see some sign of the risk direction. The result varies among banks, which
suggests bank D, H, and E with a high risk of contagion and bank B, C and
F with low one. For the two banks A and G, the result is inconsistent and
need to be examined in further studies. From what have been done, we
apprehended contagion risk in Vietnam as a real threat. However, its
magnitude and influence is uncertain. The cross-border contagion risk is
also ignored. The model need to be more developed, which need to have
the result more comprehensively. In the near future, we hope to improve
our work in order to provide a better view about the contagion risk in
Vietnam.
Appendix

Appendix 1: Co- exceedances for weekly log-differenced distance to
defaults for 8 individual banks (10% bottom tail)
Number of (co-) exceedances in the bottom tail (10%)
>6
A

2

B

4

C

5

D

5

E

4

F

3

G

4

H

5

Appendix 2: Cross-Table for pair co-exceedances (10% bottom tails)
Number of times 2 banks in 10% bottom tails
-

A

A

-

B

3

C

4

D

5

E

2

F

4

G

3


Policies and Sustainable Economic
Development | 641

H

3

Total

24

References
Bae, K., Karolyi, G., & Stulz, R. (2003). A new approach to measuring financial contagion.
Review of Financial Studies, 16, 717-763.
Blank, S., Buch, C., & Neugebauer, K. (2009). Shocks at large banks and banking sector
distress: The banking granular residual. Journal of Financial Stability, 5(4), 353-373.
Chang, E. C., Pinegar, J. M., & Ravichandran, R. (1993). International evidence on the
robustness of the day-of-the-week effect. Journal of Financial and Quantitative Analysis,
28 (December), 497-513.

Chan-Lau, J. A., Mitra, S., & Ong, L. L. (2007). Contagion risk in the international banking
system and implications for London as a global financial center. Working Paper
No.07/74. International Monetary Fund.
Cochrane, J. H., & Zingales. L. (2009). Lehman and the financial crisis. Retrieved from
/>De Bandt, O., & Hartmann, P. (2001). Systemic risk: A survey. In C. Goodhart & G. Illing (Eds.),
Financial crisis, contagion

and the lender of last resort: A book of readings (pp. 249-298). London: Oxford
University Press.
French, K. R. (1980). Stock returns and the weekend effect. Journal of Financial
Economics, 8, 55-69.
Gropp, R., & Moerman, G. (2004). Measurement of contagion in banks' equity prices.
Journal of International Money and Finance, 23(3).
Jaffe, J., & Westerfield, R. (1985). The weekend effect in common stock returns: The
international evidence. Journal of Finance, 40(2), 433-454.
Keim, D. B., & Stambaugh, R. F. (1984). A further investigation of the weekend effects in
stock returns. Journal of Finance, 39(3), 819-840.
Lakonishok, J., & Smidt, S. (1988). Are seasonal anomalies real? A ninety-year
perspective. Review of Financial Studies, 1(1), 403-425.
Longin, F., & Solnik, B. (2001). Extreme correlation of international equity markets. Journal
of Finance, 56(2), 649-676.
Rogalski, R. J. (1984). A further investigation of the weekend effect in stock returns:
Discussion. Journal of Finance, 39(3) 835-837.
Taylor, J., & Williams, J. C. (2011). Further results on a black swan in the money market.
Financial Crisis, 66 J.FIN.1109.
SIEPR Discussion Paper No. 07-46.