Journal of Financial Stability 3 (2007) 175–193

Effects of exchange rate depreciation on commercial

bank failures in Indonesia
Sahminan Sahminan ∗
Bank Indonesia, Directorate of Economic Research and Monetary Policy,
Jl. M.H. Thamrin No. 2, Jakarta 10110, Indonesia
Received 4 July 2006; received in revised form 9 April 2007; accepted 20 April 2007
Available online 4 May 2007

Abstract
This paper examines the effects of exchange rate depreciation on commercial bank failures in Indonesia
during the period from January 1995 to December 1999. This included the period of the Asian crisis during
which the Indonesian currency depreciated by about 75% in nominal terms or 25% in real terms. The
estimation results show that due to a higher amount of foreign currency assets relative to the amount of
foreign currency liabilities, exchange rate depreciation led to a lower probability of bank failure. Through
reduced profit on lending in foreign currency, exchange rate depreciation led to a higher probability of bank
failure.
© 2007 Elsevier B.V. All rights reserved.

JEL classification: F31; G21; E44

Keywords: Exchange rate depreciation; Bank failure; Indonesia

1. Introduction

Liberalization of financial markets in many developing countries has given more latitude for
banks in those countries to obtain funds and extend credits denominated in foreign currency. As
a result, holding assets and liabilities denominated in foreign currency is prevalent in the banking
sector of many developing countries. The evidence on the intensity of foreign currency assets and
liabilities in the banking sectors of developing countries, for example, is given by Arteta (2002).

∗ Tel.: +62 21 3817733; fax: +62 21 3502030.

E-mail address: [email protected].

1572-3089/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jfs.2007.04.002
176 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Using data of 37 developing and transition countries ranging from late 1970 to 2001, he finds that
the average ratio of dollar deposits to total deposits and dollar credits to total credits were 21.64%
and 27.23%, respectively.
Regardless of the reasons for choosing currency denomination of assets and liabilities, banks
are exposed to exchange rate risks when they hold assets and incur liabilities denominated in
foreign currency. In developing countries, generally, banks do not incur the cost of hedging
instruments in protecting risk due to the exchange rate changes. This could be due to the lack of
such instruments in developing countries, or for other reasons such as the high cost of hedging.
As pointed out by Sharma (2003), for example, short-term foreign borrowing was a relatively
inexpensive source of funds for banks in Indonesia because those banks do not incur the cost of
hedging against depreciation in the rupiah.
The purpose of this paper is to test the hypothesis that exchange rate depreciation led to a
higher probability of commercial bank failures in Indonesia during the period from 1995 to 1999.
Although commercial banks are engaged in various foreign exchange-related activities such as
derivative and non-asset based services, this paper focuses on the balance sheets; the components of
banks’ balance sheet are sufficient to capture bank’s net-worth, which is an indicator of bank insol-
vency. For the purpose of the study, we use data of individual commercial banks in Indonesia during
the period from January 1995 to December 1999. During the 1990s the amount of foreign currency-
denominated assets and liabilities in Indonesian banking sector increased significantly. The share
of credits in foreign currency increased from 15% in December 1991 to 37% in December 1997.
Similarly, the share of deposits in foreign currency also increased significantly, from 26% in
December 1991 to 46% in December 1997. Even using the rupiah exchange rate in 1991, the share
of credits and deposits in foreign currency in 1997 was substantially higher than the share of cred-
its and deposits in foreign currency in 1991. Using the exchange rate of the rupiah in 1991, in 1997
the share of credits in foreign currency was 23% while the share of deposits in foreign currency
was 30%.
Understanding the role of exchange rate depreciation on individual banks is important to
anticipate banking sector stability or potential fragility due to the exchange rate depreciation. This
paper provides empirical regularities between the exchange rate depreciation and the probability
of commercial bank failures in Indonesia that in turn can be used by banking regulators and
supervisors as an early-warning signal for banking sector fragility. In addition, the results of this
paper are also important for exchange rate policy makers in understanding the implications of an
exchange rate policy on the banking sector. Failure to understand the implication of the exchange
rate depreciation on the fragility of the banking sector may result in an exchange rate policy that
may lead to a large number of bank failures.
The remainder of the paper is organized as follows. In Section 2, we present a review of
the related literature. In Section 3, we present an overview of banking sector development in
Indonesia. In Section 4, we present the econometrics methodology. In Section 5, we present
descriptive statistics of the data. In Section 6, we discuss the estimation results. Finally, we
conclude this paper with Section 7.

2. Literature review

Many studies have investigated the relationship between exchange rate changes and banks’
behavior. While there is an extensive literature on the determinants of bank failure based on bank-
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 177

specific factors,1 only a few studies that test the role of macroeconomic conditions. Using data
of 31 commercial banks in Mexico, Gonzáles-Hermosillo et al. (1997) test the hypothesis that
bank fragility is determined by bank-specific factors, macroeconomic conditions, and potential
contagion effects. As an indicator for the fragility condition of an individual bank, they use a
condition on whether or not the bank receives government financial assistance. Thus, the dependent
variable in their model has a value of 1 when the bank receives financial assistance and a value
of 0 when the bank does not receive financial assistance. They find that bank-specific factors and
contagion effects have significant effects on the probability of bank failures. On the other hand,
macroeconomic factors – including exchange rate depreciation – do not have significant effects
on the probability of bank failures.
The studies by Demirgüç-Kunt and Detragiache (1998) and Hardy and Pazarbaşioğlu (1999)
provide cross-country evidence on the determinants of banking crisis. Demirgüç-Kunt and
Detragiache (1998) examine factors contributing to the probability of banking crisis in devel-
oping and developed countries. Their study covers the period from 1980 to 1994 with annual
frequency data. They define an indicator of banking crisis based on a number of factors including
the quality of banks’ assets, cost of rescue, the incidence of a large scale of bank nationalization
and extensive bank runs. The explanatory variables in their model consist of macroeconomic
variables and financial sector variables. The results of their estimation show that exchange rate
depreciation per se does not have a significant effect on the probability of banking sector fragility.
Hardy and Pazarbaşioğlu (1999) study 50 countries of which 38 countries experienced banking
system crisis or significant problems in the banking system. As the dependent variable, they use
a dummy variable with value of 2 in the period of banking sector difficulties, value of 1 in the
preceding period, and value of 0 otherwise. Explanatory variables in their study include real-sector
related variables, banking-sector variables, and shocks that may directly or indirectly affect the
health of the banking sector. Real exchange rate change is one of the variables they classify as a
shock to the health of banking sector. They find that a sharp depreciation of real effective exchange
rate significantly contributes to banking distress.
This paper contributes to the growing body of the literature on the effects of the exchange
rate depreciation on bank failure. While there have been a number of cross-country studies on
the determinants of banking failure that include exchange rate, relatively little attention has been
given to the effects of the exchange rate depreciation on individual banks. The main drawback of
the cross-country studies is that the aggregated data used in such studies can hide problems expe-
rienced by individual banks. This paper provides empirical evidence on the effects of exchange
rate depreciation on bank failure using data of individual banks.

3. Banking sector in Indonesia until 1999

An enormous expansion of the banking sector in Indonesia started in October 1988, that is,
when the Indonesian government introduced deregulation in the banking sector. The deregula-
tion of banking sector in Indonesia in 1988 lowered barriers for establishing new domestic and
joint-venture banks, and for opening private banks’ branches. The number of banks in Indonesia
increased from 111 in October 1988 to a peak of 240 in 1994. In line with the capital account
openness, in 1989 the Indonesian government eliminated quantitative limits on banks to borrow

1 Gonzáles-Hermosillo (1999) provides the list of selected empirical studies of US bank failures based on bank specific

variables.
178 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

from abroad. Furthermore, starting in 1990 Indonesian commercial banks have been allowed to
buy foreign exchange directly from foreign investors instead of through the central bank (Sharma,
2003).
It has been argued in the literature that banking deregulation can bring benefit to the economy
through a more efficient financial intermediation and credit allocation. However, the deregulation
in banking sector can also pose problems. This pattern seems to have taken place in the Indonesian
banking industry. As pointed out by Sharma (2003), for example, the rapid growth of the banking
sector in Indonesia has been achieved at the cost of banking sector soundness. The fragility of the
banking sector in Indonesia became more obvious following the currency crisis in 1997.
In restructuring banking industry, the Indonesian government has adopted various approaches
including recapitalization, mergers, and closures. After many years of being reluctant to close
problem banks, the Indonesian government closed 16 banks in November 1997. And although the
closure of those banks was intended to improve the soundness of the banking sector, that policy was
followed by the worsening situation of the banking sector. By early 1998, bank runs were prevalent
in the country due to the rumors of the preparation for the closure of other banks. As documented
by Enoch et al. (2001), 154 banks had experienced erosion in their deposits by mid-December
1997. As a part of the restructuring program in the banking sector, the Indonesian government
also introduced the new central bank law, bank liquidation law, new prudential regulations, and
bankruptcy law. Moreover, in addition to the closure of the 16 banks in November 1997, the
Indonesian government closed 38 more banks in March 1999. The restructuring program of the
Indonesian banking sector had been mostly achieved by the end of 1999 (Enoch et al., 2001).

4. Empirical framework

4.1. Model specification

To estimate the effects of the exchange rate depreciation on the probability of bank failures,
we formulate an empirical model as follows. Let Zit be a variable of bank i at time t with value of
1 if the bank is insolvent and value of 0 if the bank is solvent, then the specification of the model
is given by

Pr(Zit∗ = 1) = F (α0 + α1 et + α2 et (Lfi(t−1) − Dfi(t−1) ) + Xit ␤) (1)

where et is a change in exchange rate between period (t − 1) and period t, Xit a vector of
control variables, ␤ a vector of control variables’ coefficients, and F(·) is a continuous cumulative
distribution function.
The effect of the exchange rate depreciation on the probability of a bank failure through
mismatch in foreign currency assets–liabilities is captured by α2 (Lfi(t−1) – Dfi(t−1) ), and therefore
α1 can be used to capture the effects of the exchange rate depreciation through bank’s profit
from foreign currency lending. Thus, the total effects of the exchange rate depreciation on the
probability of bank failure is captured by α1 + α2 (Lfi(t−1) − Dfi(t−1) ).
The expected effect of the exchange rate depreciation on the banks’ profit from foreign currency
lending is negative, and therefore we expect ␣1 to be positive. The coefficient of the interaction
between exchange rate and the mismatch in foreign currency assets–liabilities (α2 ) is expected
to be negative. Thus, if the amount of foreign currency assets in foreign currency is larger than
the amount of foreign currency liabilities for a single bank then the effect of the exchange rate
depreciation on the probability of bank failure through currency mismatch is negative. On the
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 179

other hand, if the amount of foreign currency assets is less than the amount of foreign currency
liabilities, then the effect through currency mismatch is positive.

4.2. Estimation method

The probability of bank failure is estimated using a panel logit model. This type of model
has been widely used in the literature on estimating the probability of bank failure (for example,
Demirgüç-Kunt and Detragiache, 1998; Gonzáles-Hermosillo, 1996, 1999; Cole and Gunther,
1995).
With the logit model, F(·) is a logistic cumulative distribution function; and by defining Zit =
α0 + α1 et + α2 et (Lfi(t−1) − Dfi(t−1) ) + Xit ␤, then the model can be written as
exp(Zit )
Pr(Zit∗ = 1) = F (Zit ) = (2)
1 + exp(Zit )
and the log-likelihood function of the model is given by
 
ln L = {Zit∗ ln{F (Zit )} + [1 − Zit∗ ] ln{1 − F (Zit )}} (3)
t=1...T i=1...n

While the signs of the coefficients can be used to predict the effects of the explanatory vari-
ables on the probability of bank failures, the coefficients themselves cannot be interpreted as the
magnitude of the effects of the changes in explanatory variables on the probability of failure.
Instead, the coefficients are interpreted as the magnitude of the response of the log of odds failure
ratio, ln[Pr(Zit∗ = 1)/(1 − Pr(Zit∗ = 1))], to a change in an explanatory variable. The probability
of failure is estimated by

exp(Ẑit )
P̂it = (4)
1 + exp(Ẑit )
Hence, based on Eq. (4) the probability that bank i fails at time t depends on the magnitude of
the explanatory variables.

5. Data and descriptive statistics

5.1. Sample

In this study, the balance sheet data of individual commercial bank are obtained from Bank
Indonesia—the regulator and supervisor of the banking sector in Indonesia. Other data such
as exchange rate and other macroeconomic variables are obtained from International Financial
Statistics (IFS) of the International Monetary Fund. All balance sheet data are denominated in
rupiah. The period of the study ranges from January 1995 to December 1999 with monthly
frequency. Therefore, the study covers the period before and after the currency crisis that hit
Indonesia in 1997.
Basically, except that the data for a few banks are missing for certain periods, this study covers
all commercial banks in Indonesia. Excluding missing data, for every month the sample covers
more than 90% of the total number of commercial banks in Indonesia, and includes all 16 banks
closed in November 1997 and all 38 banks closed in March 1999. Thus, the sample captures not
180 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Table 1
Number of banks by category of ownership
Category December 1995 December 1996 December 1997 December 1998 December 1909

Government owned 34 34 34 34 32
Private forex 75 77 77 71 47
Private non-forex 90 87 67 59 45
Foreign 10 10 10 10 10
Joint 31 31 34 34 30
All banks 240 239 222 208 164

Source: Bank Indonesia.

only the information on the survived banks, but also the information on the closed banks before
they were liquidated.
With respect to the ownership of the banks, the sample covers all categories of banks: govern-
ment owned banks, domestic private banks, foreign banks, and joint banks. Government banks
include the banks owned by the central government as well as the banks owned by regional
governments. All foreign banks and joint banks are foreign exchange banks, that is, the banks
permitted to hold assets and liabilities in foreign currency. On the other hand, not all domestic
banks are foreign exchange banks. The number of banks by category of ownership is presented
in Table 1.

5.2. Descriptive statistics

5.2.1. Endogenous variable

The definition of bank failure used in this paper is economic failure as opposed to official failure.
A bank is defined as economically failed or market insolvent when its net-worth is negative, that
is, when its assets fall short of its liabilities. On the other hand, a bank officially fails when the
bank authority closes that bank or provides assistance to make the bank survives. As pointed
out by Gonzáles-Hermosillo (1996), economic failure or market insolvency does not necessarily
coincide with the official failure. A bank can be insolvent but remains in business if the bank
regulator does not liquidate that bank, and instead the regulator provides assistance such as an
injection of additional capital and other restructuring programs.
Let W be net-worth, A be total assets and L be total liabilities, then the net-worth can be defined
as
W =A−L (5)
Suppose A consists of domestic currency assets (Ad ) and foreign currency assets (Af ), and L
consists of domestic currency liabilities (Ld ) and foreign currency liabilities (Lf ), then
W = Ad + eAf − Ld − eLf (6)
W = (Ad − Ld ) + e(Af − Lf ) (7)
W = (Ad − Ld ) + e (Af − Lf ) + (Af − Lf ) e (8)
where e represents nominal exchange rate.
Based on the definition of the net-worth, we calculate the net-worth of individual banks for
each month during the period of sample. Summary statistics of the net-worth, assets and liabilities
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 181

Table 2
Summary statistics of net-worth, assets and liabilities (millions of rupiah)
Item 1995 1996 1997 1998 1999

Net-worth
Mean 117,283 138,512 182,749 89,909 −1,066,530
S.D. 278,814 334,075 458,419 1,261,076 5,916,176
Assets in rupiah
Mean 1,122,752 1,331,388 1,658,743 2,243,147 2,757,980
S.D. 3,366,569 3,916,969 4,843,196 6,724,640 9,539,204
Assets in forex
Mean 472,335 523,277 803,301 2,440,641 1,798,227
S.D. 1,662,866 1,827,374 2,775,609 8,562,696 5,983,079
Liabilities in rupiah
Mean 1,018,112 1,215,115 1,501,565 2,644,059 3,912,995
S.D. 3,164,524 3,685,524 4,546,367 8,639,118 13,271,781
Liabilities in forex
Mean 459,692 501,037 777,730 1,949,820 1,709,741
S.D. 1,602,135 1,735,976 2,682,053 6,956,957 5,985,408

in domestic and foreign currencies are presented in Table 2. The dynamics of the rupiah exchange
rate is presented in Fig. 1, and the dynamics of the number of banks with negative net-worth
during the period of the study is shown in Fig. 2A.
As can be seen from the solid line in Fig. 2B, the number of banks with negative net-worth
increased significantly in mid-1998, and reached its peak in February 1999. Before June 1998,
less than 5% of the banks had negative net-worth, and this jumped to about 30% in December
1998. The number of banks with negative net-worth had declined to about 15% by the end of
1999.
The question is: what were the driving factors that led to the increase in the number of banks
with negative net-worth? Before we estimate the econometric models as formulated in Section
4, we provide descriptive statistics of the variables related to the net-worth. Based on Eq. (8),
arithmetically, three factors may contribute to the change in banks’ net-worth: (i) a change in the

Fig. 1. Rupiah exchange rate.

182 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Fig. 2. Dynamics of the banks’ variables: (A) number of banks by net forex assets–liabilities; (B) number of banks with
negative net-worth; (C) number of banks by net domestic currency assets–liabilities; (D) average of assets and liabilities
in US dollar; (E) share of foreign currency assets and liabilities; (F) average total assets; (G) ratio of inter-bank deposits
to total assets; (H) ratio of liquid assets to total assets.

amount of the domestic currency assets relative to the domestic currency liabilities; (ii) a change
in the foreign currency assets relative to the foreign currency liabilities; and (iii) a change in the
nominal exchange rate.
Thus, to examine what factors contributed to a sharp increase in the number of banks with
negative net-worth during the period of mid 1998–end of 1998 we need to address the following
questions. First, do we observe an increase in the number of banks with domestic currency assets
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 183

less than domestic currency liabilities during mid 1998–end of 1998? Second, given that there
was a sharp depreciation in the rupiah during the period of mid 1998–end of 1998, do we observe
an increase in the number of banks with negative assets–liabilities in foreign currency during that
period?
A number of regularities we observe from the data are as follows. First, as shown in Fig. 2
A, the percentage of banks with negative net foreign exchange assets–liabilities was relatively
lower during the period from October 1997 to January 1999. This evidence indicates that currency
mismatch does not seem to be the cause of the increase in the number of banks with negative
net-worth in 1998. The increase in the number of banks with negative net-worth was not in line
with the increase in the percentage of banks with negative net foreign currency assets–liabilities.
Second, if we look at the percentage of banks with negative net-worth by category of banks in
terms of foreign exchange activity, as shown in Fig. 2B, the patterns of the percentage of banks
with negative net-worth in both foreign exchange banks and non-foreign exchange banks were
quite similar. That is, in both foreign exchange and non-foreign exchange banks there was a sharp
increase in the number of banks with negative net-worth that reached peaks at the beginning of
1999. However, since 1998 the percentage of foreign exchange banks with negative net-worth was
substantially higher than the percentage of non-foreign exchange banks with negative net-worth.
This indicates that a sharp depreciation in exchange rate during the crisis negatively affected the
solvency of both foreign exchange and non-foreign exchange banks, but the effects were much
more pronounced in the foreign exchange banks.
Third, Fig. 2C shows that the percentage of banks with negative net domestic currency
assets–liabilities increased from December 1997 to February 1998. Thus, domestic currency
assets–liabilities mismatch seems to be one of the factors that caused an increase in the number
of insolvent banks from 1997 to 1998.
Fourth, as shown in Fig. 2D, the average of assets and liabilities denominated in the US dollar
declined during the period of July 1997 to March 1998, but the average of liabilities in the US
dollar fell more than the average of assets. Thus, when the rupiah depreciated, the amount of
foreign currency assets and liabilities in the US dollar decreased. Since the average of liabilities
in the US dollar fell more than the average of assets in US dollar, then the banking system does
not seem to have the problem of the assets–liabilities mismatch in US dollar during the period
when there was an increase in the number of insolvent banks.
Finally, as shown in Fig. 2E, shares of foreign currency asset and liability movements follow
closely the movements of the rupiah per US dollar. When the rupiah depreciated, nominal value of
the foreign currency assets and liabilities in terms of the rupiah increased relative to the nominal
value of domestic currency assets and liabilities. Starting in July 1997, the share of foreign
currency assets and liabilities in terms of domestic currency had increased enormously. Then at
the beginning of 1999, the share of assets and liabilities in foreign currency started to decline,
and by the end of 1999 it reached the level that was lower than the share before the crisis.

5.2.2. Control variables

Control variables we use in this paper have also been used in a number of empirical studies
of bank failure (for example in Cole and Gunther, 1995; Gonzáles-Hermosillo, 1996, 1999). As
summarized in Table 3, the exogenous variables consist of three categories of variables: bank
specific variables, macroeconomic variables, and a dummy variable. Summary statistics of the
control variables during the period of January 1995–December 1999 are reported in Table 4.
The logarithm of total assets is included as an exogenous variable to control for the economies
of scale. In the literature it has been argued that a larger bank has a better capability in diversifying
184 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Table 3
Definitions of the control variables
Variable Definitions Expected sign Measure of

Bank specific
SIZE Logarithm of total assets − Economies of scale
IBD Ratio of deposits from other banks to total assets + Deposit runs
LIQ Ratio of cash and investment securities to total assets − Safety net for deposit
withdrawal
Macroeconomics
INTR Interest rate on loan + Interest rate shock
PROD Changes in industrial production index − Economic activity
INFL Inflation rate − Price shock
Dummy
DUMM 1 in the presence of deposit guarantee; 0 otherwise + The effect of deposit
guarantee

risks (Martin and Mauer, 2003). Thus, we expect that the larger the total assets of a bank the lower
is its probability to fail. Fig. 2F shows the average of total assets of commercial banks in Indonesia
during the period of the study.
When banks face liquidity difficulties, they will seek out additional funds from inter-bank
money markets, most likely at higher interest rates. Thus, the higher the ratio of deposits from
other banks, the higher is the probability of the borrowing banks to fail. As shown in Fig. 2G, there
was a sharp decline in the average ratio of inter-bank deposits to total assets from July 1997 to
July 1998. While Fig. 3G does not show the relationship between variation of inter-bank deposits
and bank insolvency, such a relationship is tested in the empirical models.
A bank with a relatively large amount of cash – or investment in risk-free securities – has a
higher ability to meet unexpected deposit withdrawals. Therefore, the higher the amount of liquid
assets held by a bank, the lower is its probability to fail. Fig. 2H shows that there was a sharp
increase in the average ratio of liquid assets from July 1997 to July 1998.
As the control for macroeconomic shocks we use industrial production index, interest rate, and
inflation rate. Change in industrial production index is used as an indicator for economic activity.
We expect that a change in industrial production index has a negative effect on the probability
of bank failure. As can be seen in Fig. 3A, the change in industrial production index during the
period of the crisis does not seem to have a different pattern compared to the period before the
crisis.

Table 4
Summary statistics of the control variables
Variables Mean S.D. Minimum Maximum

SIZE 13.39 1.59 9.07 19.02

NPL 19.34 22.76 0.00 99.81
IBD 8.36 11.65 0.00 84.49
LIQ 9.39 13.70 0.02 94.30
INTR 25.17 6.25 18.56 35.72
PROD 0.22 11.10 −27.33 31.90
INFL 1.70 2.76 −1.05 12.76
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 185

Fig. 3. Dynamics of macroeconomic variables: (A) changes in industrial production index; (B) interest rates on deposits
and loans; Indonesia’s inflation rate.

The effects of interest rates on the probability of bank failures are expected to be positive. As the
interest rates rise, the more difficult it is for the borrowers to meet their liabilities for the banks as
the lender, and the banks are more difficult to extend credits. As shown in Fig. 3B, interest rates on
both rupiah-denominated deposits and rupiah-denominated loans increased from September 1997
to January 1999, and then declined until the end of 1999. If we look at the period from May 1998 to
May 1999, the interest rates on deposits were substantially higher than the interest rates on loans.
186 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Finally, inflation rate is expected to have positive effects on banks’ performance, and therefore
negative effects on the probability of bank failure. As argued by Lindgren et al. (1996), a policy
that accommodates inflation may improve bank profitability. Banks can take advantage of the
high inflation by indexing lending rates and shifting their portfolio into assets whose price lead
inflation. As warned by Lindgren et al., however, there is also potential risk to the bank when the
monetary policy is reversed toward a tighter one. A tight monetary policy can cause banks facing
liquidity crisis that in turn can lead to insolvency. As shown in Fig. 3C, there was a sharp increase
in the rate of inflation during 1998 and reached its peak in March 1998.
In addition to the bank-specific and macroeconomic variables we include a dummy variable
for the presence of deposit guarantee. In January 1998, the Indonesian authority announced a
blanket guarantee for deposits both denominated in the rupiah and foreign currency. To capture
this effect, we define a dummy variable with the value of 1 for the data since February 1998, and
the value of 0 for the data before February 1998. We expect to obtain a negative sign if such a
deposit guarantee is effective to reduce the probability of bank failure in Indonesia during the
period from February 1998 to December 1999.

6. Results

In this section, we first discuss the estimation results of the empirical models. Then, based on
the coefficient estimates of the models we provide policy experiments related to exchange rate
depreciation and net foreign currency assets–liabilities.

6.1. Estimation results

The empirical models are estimated for two different samples. In the first estimation we use
a sample of all commercial banks, and in the second estimation we use a sample of only foreign
exchange banks. The estimation results with the full sample are reported in Table 5, and the
estimation results with the sample of only foreign exchange banks are reported in Table 6.
For each sample we estimate six different model specifications. The differences in the model
specifications are based on the types of interest rates used as explanatory variables, and whether
or not the size of the banks is included as an explanatory variable. Three different interest rates
we use in the model are interest rates on loan, interest rates on deposits, and the spread between
the interest rates on loans and the interest rates on deposits. While the interest rates on loans are
modestly correlated with the interest rates on deposits, it turns out that both of them are highly
correlated with the interest rate spread. Thus, instead of including the interest rates and the interest
rate spread in the same model we estimate them in different models.
In each of the model specifications 1–3 and 4–6 we use the interest rates on loans, the interest
rates on deposits, and the interest rate spread, respectively. With model specifications 1–3, the
models are estimated by including the size of the banks. On the other hand, with model specifi-
cations 4–6 we estimate the model without the size of the banks. This strategy is used to assess
the sensitivity of the model to the size of the banks.
To give some idea on the quality of the model, we report the χ2 statistics of each specification.
The χ2 statistics tests the joint significance of the explanatory variables in the model. For each
specification of the models, at 1% significance level we reject the hypothesis that there is no joint
effect of the explanatory variables on the probability of bank failures. This rejection is true for the
models using the sample of all banks as well as for the models using the sample of only foreign
exchange banks.
Table 5

S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Estimation results using sample of all banks
Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

e 1.2937* (2.84) 1.6202* (3.49) 1.7324* (3.67) 1.1259* (2.57) 1.3442* (3.02) 1.3984* (3.10)
e(Lf − Df ) −0.0028 (−1.88) −0.0028 (−1.83) −0.0027 (−1.82) −0.0030* (−2.02) −0.0029* (−1.98) −0.0029* (−1.98)
SIZE −2.0095* (−9.56) −2.0252* (−9.67) −2.0202* (−9.68) – – –
IBD 0.0114 (1.39) 0.0110 (1.35) 0.0114 (1.40) 0.0134** (1.78) 0.0132** (1.77) 0.0135** (1.82)
LIQ −0.0002 (−0.02) 0.0001 (0.01) −0.0012 (−0.10) 0.0043 (0.40) 0.0038 (0.36) 0.0025 (0.23)
INTR Loan 0.0884* (5.69) – – 0.0723* (4.96) – –
INTR Dep – 0.0314* (5.35) – – 0.0239* (4.29) –
NET INTR – – −0.0440* (−4.90) – – −0.0315* (−3.69)
PROD −0.0154* (−2.81) −0.0118* (−2.15) −0.0106** (−1.91) −0.0129* (−2.41) −0.0103 ** (−1.91) −0.0096** (−1.78)
INFL −0.4194* (−13.12) −0.4040* (−12.90) −0.3963* (−12.80) −0.4720* (−15.01) −0.4578* (−14.91) −0.4503* (−14.84)
DUMM 5.5101* (16.44) 5.9112* (18.66) 6.1641* (19.90) 4.6235* (15.89) 4.9531* (18.38) 5.1499* (19.89)
Log-likelihood −680.29 −682.35 −684.84 −731.69 −734.88 −737.36
LR ␹2 1557.19 1553.07 1548.08 1454.39 1448.01 1443.06

Note: Numbers in parentheses are z-statistics.

* Significant at 5%.
** Significant at 10%.

187
 188
Table 6

S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Estimation results using sample of foreign exchange banks
Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

e 1.4099* (2.82) 1.7333* (3.40) 1.8497* (3.57) 1.0934* (2.28) 1.2958* (2.66) 1.3521* (2.74)
e(Lf − Df ) −0.0030* (−1.97) −0.0029 ** (−1.91) −0.0029** (−1.88) −0.0030* (−2.01) −0.0030* (−1.97) −0.0030* (−1.96)
SIZE −2.2057* (−9.04) −2.2217* (−9.11) −2.2152* (−9.09) – – –
IBD 0.0073 (0.76) 0.0071 (0.73) 0.0076 (0.79) 0.0109 (1.19) 0.0110 (1.20) 0.0115 (1.26)
LIQ −0.0072 (−0.54) −0.0065 (−0.49) −0.0077 (−0.57) −0.0056 (−0.45) −0.0057 (−0.46) −0.0069 (−0.55)
INTR Loan 0.0883* (5.20) – – 0.0672* (4.25) – –
INTR Dep – 0.0319* (4.96) – – 0.0226* (3.74) –
NET INTR – – −0.0452* (−4.59) – – −0.0303* (−3.26)
PROD −0.0134* (−2.22) −0.0098 (−1.61) −0.0084 (−1.38) −0.0114** (−1.94) −0.0089 (−1.50) −0.0082 (−1.38)
INFL −0.4144* (−11.99) −0.3998* (−11.77) −0.3923* (−11.66) −0.4782* (−14.04) −0.4658* (−13.96) −0.4590* (−13.90)
DUMM 6.2380 * (15.21) 6.6383* (16.82) 6.8927* (17.70) 5.1472* (14.94) 5.4470* (16.83) 5.6286* (17.95)
Log-likelihood −556.05 −557.42 −559.32 −602.60 −604.71 −606.44
LR χ2 1427.37 1424.65 1420.84 1334.29 1330.06 1330.061

Note: Numbers in parentheses are z-statistics.

* Significant at 5%.
** Significant at 10%.
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 189

As shown in Table 5, et has positive significant coefficients, while et (Lfi(t−1) − Dfi(t−1) )
has negative significant coefficients. This means that exchange rate depreciation has a positive
effect on the bank failure through bank profits on lending in foreign currency, and negative effects
through currency mismatch. With different interest rates, the qualitative results of the effects of the
exchange rate depreciation do not change: both et and et (Lfi(t−1) − Dfi(t−1) ) remain significant
with correct signs.
The effects of other explanatory variables on the bank failure are as follows. In all model
specifications, ratio of liquid assets does not have a significant effect on the bank failure. Similarly,
inter-bank borrowing also does not have a significant effect on the bank failure in all model
specifications.
All macroeconomic variables used as control variables are significant in all model specifica-
tions. Interest rates – either interest rates on loans or interest rates on deposits – have significant
and positive effects on bank failure. Thus, the data support the hypothesis that the higher the
interest rates, the higher the probability of a bank to fail. The interest rate spread is also signifi-
cant with negative sign, meaning that the higher the difference between the interest rates on loans
relative to the interest rates on deposits, the lower is the probability of a bank to fail. This support
the hypothesis that a negative spread of the interest rates on loans relative to the interest rates on
deposits can lead to a higher probability of bank failure.
As measured by the change in industrial production index, in all model specifications, economic
activity has a negative and significant effect on bank failure. That is, the data support the hypothesis
that the higher the economic activity, the lower is the probability of a bank to fail. The negative
effect of the inflation rate on bank failure is also supported by the data. In all model specifications,
inflation rate has negative and significant coefficients. This indicates that commercial banks in
Indonesia were able to take advantage of the high inflation rate that took place during the period
of the study.
In all model specifications, as can be seen in Table 5, the dummy variable has positive sig-
nificant effects on the probability of bank failure. This result indicates that the introduction of
blanket guarantee at the beginning of 1998 had not been successful in mitigating the insolvency
of the banks in Indonesia until 1999. The positive sign of the dummy variable shows that after
January 1998 the probability of bank failure was higher than the probability of bank failure from
1995 to January 1998. Of course, given the severity of the problems facing banking system in
Indonesia until 1999, we cannot attribute such a higher probability to the introduction of the
blanket guarantee.
Whether or not the size of the bank is included in the models does not affect the estima-
tion results qualitatively. The inclusion of the size of the bank does not change the signs and
the significance of most of the explanatory variables. As shown in column (4) through (6)
of Table 5, the exchange rate depreciation still has significant and positive effects on a bank
failure through profits from a lending in foreign currency, and significant and negative effects
through currency mismatch. This evidence indicates that the estimation results are quite robust
to the size of the banks. The size of the bank itself – which is measured by the logarithm
of total assets – has a significant and negative effect on bank failure. Thus, the data support
the economies of scale hypothesis, that is, the larger the bank the lower is its probability to
fail.
As mentioned at the outset of this section, we also estimate the model using the data of only
foreign exchange banks. As shown in Table 6, the signs and the significance of the explana-
tory variables are very similar to those of the explanatory variables using the sample of all
banks.
190 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Table 7
Share of deposits by maturity and credits by types
Maturity/types 1997 1998 1999

Deposits
1-month 54.69 64.76 62.36
3 and 6-month 20.72 14.30 17.45
12 and 24-month 10.20 4.70 3.84
Credits
Investment 26.64 29.02 25.63
Working capital 63.67 64.46 63.68
Consumer 9.69 6.51 10.70

Source: Calculated from Bank Indonesia publications.

Another factor that may contribute to the increase in the number of insolvent banks in Indonesia
during the period of currency crisis was maturity mismatch. Due to the limitation of the series of
the data, we do not include the maturity mismatch into the model. The evidence on the maturity
mismatch experienced by the Indonesian banking sector is as follows. As shown in Table 7, during
the period of 1997–1999, we observe that commercial banks in Indonesia experienced significant
shifts in the maturity of the deposits. In 1997 the share of deposits with maturity of 1 month was
55% and increased to 65% in 1998. On the other hand, the share of credit for investments, which
presumably has the longest maturity among the type of credits, was relatively stable. In 1997,
the amount of credits for investment was accounted for 27% of the total credit and increased to
29% in 1998 before it returned to 26% in 1999. This fact tells us that the commercial banks in
Indonesia during the period of 1997–1999 also faced maturity mismatch that could lead to the
insolvency.

6.2. Policy experiments

As mentioned previously, the coefficients of the variables do not measure the marginal effects
of the explanatory variables on the probability of bank failure. Instead, the magnitudes of the
coefficients measure the marginal effects of the explanatory variables on Zt . The marginal effects
of the explanatory variables on the probability of bank failure are calculated from the predicted
probability of bank failures given by Eq. (4). To asses the magnitudes of the effects of exchange
rate depreciation and net foreign assets–liabilities on the probability of bank failure, we take the
following approach. First, using the coefficient estimates and the average values of the explana-
tory variables, we calculate the estimated values of Zt . Second, using the estimated values of
Zt we calculate the probability of bank failure under different levels of exchange rate depre-
ciation and net foreign currency assets–liabilities. Finally, under different levels of exchange
rate depreciation, we plot the probability of bank failures as a function of net foreign currency
assets–liabilities.
Based on the estimated values of model 1, under different rates of the depreciation we cal-
culate the probability of bank failures as a function of net foreign currency assets–liabilities;
the plots of the calculation results are shown in Fig. 4. A number of facts we can see in
Fig. 4 are as follows. First, as shown in Fig. 4A, if the exchange rate does not change then
the probability of bank failure is independent of net foreign assets–liabilities, other things being
equal.
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 191

Fig. 4. Plots of the probability of bank failures as a function of net foreign currency assets–liabilities: (A) e = 0; (B)
e = 10; (C) e = 20; (D) e = 25; (E) e = 100; (F)e = 1000.

Second, when the exchange rate depreciates by 10 units (Fig. 4B), as the banks’ net foreign cur-
rency liabilities short of their foreign currency assets by US$ 600-million or more, the probability
of the banks to fail approaches one. And if the exchange rate depreciates by 20 units (Fig. 4C),
the probability function shift to the right and the banks with foreign currency liabilities short of
foreign currency assets by US$ 100-million or more will fail with probability of approximately
one.
Finally, from Fig. 4D–F, we can see that when the exchange rate depreciates by 25-unit or more,
even the banks with positive net foreign currency assets–liabilities have high probability of failure.
The higher the depreciation rate, the higher is the value of net foreign currency assets–liabilities
required to avoid failure due to the exchange rate depreciation. If we look at Fig. 4F, when the
192 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193

Fig. 5. Distribution of banks based on the average values of net foreign currency assets–liabilities.

exchange rate depreciates by 1000 units or more – as happened in Indonesia during the currency
crisis – a bank needs more than US$ 450-million of net foreign currency assets–liabilities in order
to avoid adverse effects of the depreciation.
In Fig. 5, we show the distribution of commercial banks in Indonesia based on the average
values of their net foreign currency assets–liabilities during the period of January 1995–December
1999. As shown in Fig. 5, the majority of Indonesian commercial banks held net foreign currency
assets–liabilities between US$ 0 and 100-million, and there was no bank with negative net foreign
currency assets–liabilities less than negative US$ 100 million. On the right tail of the distribution,
we can see that less than 1% of total banks held net foreign currency assets–liabilities larger than
US$ 450-million. Thus, other things being equal, if the rupiah depreciated by only 10 units no
bank would fail. But if the rupiah depreciated by 25 units, about 50% of the banks would fail. In
the extreme case, if the exchange rate depreciated by 1000 units, almost all banks would fail due
to the exchange rate depreciation.

7. Conclusions

During the currency crisis that hit many East Asian countries in 1997, Indonesian economy
was characterized by a large exchange rate depreciation and a large number of banks failures.
In this paper, using monthly data of individual commercial banks in Indonesia during the period
from January 1995 to December 1999, we test the hypothesis that exchange rate depreciation led
to a higher probability of bank failure in Indonesia.
The estimation results show that, through mismatch in foreign currency assets–liabilities the
exchange rate depreciation results in a lower probability of bank failure, and through profit on
lending in foreign currency the exchange rate depreciation results in a higher probability of bank
failures. Other variables such as the size of the bank, interest rates, industrial production index,
and inflation rates also have significant effects on the probability of bank failure.
Our findings in this paper show that while commercial banks in Indonesia during the period
from 1995 to 1999 could avoid currency mismatch by extending loans in foreign currency, their
balance sheets suffered from exchange rate depreciation through the return on foreign currency
loans. The implication for the banking regulator and supervisor is that the adverse effects of the
exchange rate depreciation on commercial banks cannot be mitigated merely through monitoring
net foreign currency exposure of the banks; instead, the portfolios of the foreign currency loans
of the banks call for more attention.
 S. Sahminan / Journal of Financial Stability 3 (2007) 175–193 193

References

Arteta, C.O., 2002. Exchange rate regimes and financial dollarization: does flexibility reduce bank currency missmatches?,
International Finance Discussion Papers No. 738, Board of Governors of the Federal Reserve System.
Cole, R.A., Gunther, J.W., 1995. Separating the likelihood and timing of bank failure. J. Banking Finance 19, 1073–1089.
Demirgüç-Kunt, A., Detragiache, E., 1998. Determinants of banking crisis in developing and developed countries. IMF
Staff Papers, vol. 45, no. 1.
Enoch, C., et al., 2001. Indonesia: anatomy of a banking crisis, two years of living dangerously, 1997–1999. IMF Working
Paper, WP/01/02.
Gonzáles-Hermosillo, B., 1996. Banking sector fragility and systemic sources of fragility. IMF Working Papers, WP/96/12.
Gonzáles-Hermosillo, B., 1999. Determinants of ex-ante banking system distress: a macro–micro empirical exploration
of some recent episodes. IMF Working Papers, WP/99/33.
Gonzáles-Hermosillo, B., Pazarbaşioğlu, C., Billings, R., 1997. Determinants of banking system fragility: a case study of
Mexico. IMF Staff Papers, vol. 44, no. 3.
Hardy, D.C., Pazarbaşioğlu, C., 1999. Determinants and leading indicators of banking crises: further evidence. IMF Staff
Papers, vol. 46, no. 3.
Lindgren, C., Garcia, G., Saal, M.I., 1996. Bank Soundness and Macroeconomic Policy. International Monetary Fund,
Washington, DC.
Martin, A.D., Mauer, L.J., 2003. Exchange rate exposures of US banks: a cash flow-based methodology. J. Banking
Finance 27, 851–865.
Sharma, S.D., 2003. The Asian Financial Crisis: Crisis, Reform and Recovery. Manchester University Press, New York.