Banking Industry Volatility and BANKING CRISIS

Available online at www.sciencedirect.
com
Int. Fin. Markets, Inst. and Money 19 (2009) 351–370
Banking industry volatility and

banking crises夽
Fariborz Moshirian a,∗ , Qiongbing Wu b
a School of Banking and Finance, The University of New South Wales,
Sydney, NSW 2052, Australia
b Newcastle Graduate School of Business, The University of Newcastle,
Newcastle, NSW 2300, Australia
Received 24 July 2006; accepted 5 February 2008
Available online 2 March 2008
Abstract
While studies using balance sheet information of banks and macroeconomic indicators to forecast banking
crises are prolific, empirical research using market information of banks is relatively sparse. We investigate
whether banking industry volatility, constructed with the disaggregated approach from Campbell et al.
[Campbell, J.Y., Lettau, M., Malkiel, B.G., Xu, Y., 2001. Have individual stocks become more volatile?
An empirical exploration of idiosyncratic risk? The Journal of Finance 56, 1–43] using exclusively publicly
available market information of banks, is a good predictor of systemic banking crises in the analyses including
data from 18 developed and 18 emerging markets. We find that banking industry volatility performs well in
predicting systemic banking crises for developed markets but very poor for emerging markets, which suggest
that the impact of market forces on the soundness of the banking system might be different for developed
and emerging markets. We also find that those macroeconomic and banking risk management indicators
have different impact on the probability of banking crises. Therefore, the traditional cross-country results of
the studies on banking crises need to be interpreted cautiously.
© 2008 Elsevier B.V. All rights reserved.
JEL classiﬁcation: G15; G21; G28
Keywords: Banking crises; Volatility; Market forces
夽 The paper is based on Chapter 3 of Qiongbing Wu’s PhD thesis from the University of New South Wales.
∗ Corresponding author. Tel.: +61 2 93855859.
E-mail addresses: f.moshirian@unsw.edu.au (F. Moshirian), Linda.wu@newcastle.edu.au (Q. Wu).
1042-4431/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.intfin.2008.02.002
352 F. Moshirian, Q. Wu / Int. Fin. Markets, Inst. and Money 19 (2009) 351–370
1. Introduction
Recent empirical research has strongly supported the theoretical view that a well-functioning
banking system is important in a country’s economic development. Banks have boosted growth
at the country level (King and Levine, 1993; Levine and Zervos, 1998; Beck et al., 2000; Beck
and Levine, 2004), at the industry level (Rajan and Zingales, 1998; Cetorelli and Gambera, 2001;
Beck and Levine, 2002), and at the firm level (Demirguc-Kunt and Maksimovic, 1998, 2002).
The positive effect of banks on economic development is robust to different econometric methods
(Levine, 2005).
Since banks have played such an important role in economic development, banking crises can
generate serious disruptions of a country’s economic activity (Hoggarth et al., 2002). Therefore,
to ensure the soundness of the banking system and prevent the occurrence of banking crises is
undoubtedly a main concern of policy makers and regulators. The role of market discipline in
ensuring financial stability is becoming so prominent that the New Basel Capital Accord developed
by the Basel Committee on Banking Supervision (2003) has included market discipline as one
of the three pillars1 to recognize its importance in promoting safety and soundness in banks and
financial systems. The strength of market discipline derives from the immense power of the price
system to aggregate information (Crockett, 2002). There is extensive literature on the forecasting
of banking crises using balance sheet information of banks as well as macroeconomic indicators.
However, studies on the probabilities of banking crises using market information of banks are
relatively sparse. This research contributes to the literature by using the publicly available market
information of banks to predict the probability of systemic banking crises. Publicly traded banks
are broadly representative of a country’s banking sector, so that the stock prices of banks listed
in the domestic exchanges will reflect the performance of a country’s banking sector. A certain
degree of price fluctuations is desirable since it reflects the information flows in an efficient market,
while excessive changes of stock prices might signify uncertainty of the future economic status.
Therefore banking industry volatility could indicate the stability of a country’s banking sector
performance. From this point of view, we investigate whether banking industry volatility, among
those leading macroeconomic variables, is a useful predictor of systemic banking crises.
To address this issue, we first construct the portfolios of banks listed in domestic stock
exchanges for 36 markets which consist of 18 developed markets and 18 emerging markets.2
Secondly, we construct the value-weighted banking industry volatility using the unique disaggre-
gated approach from Campbell et al. (2001). Thirdly, we construct the macroeconomic variables
that are traditionally thought to be the leading indicators of banking crises for each of the 36
markets, including real GDP growth rates, real interest rates, inflation rates, changes of exchange
rates, domestic credit growth rates, ratios of M2 against reserves, and the volatility of GDP growth
rates. Finally we use a Logit econometric model to test whether banking industry volatility is a
good predictor of banking crises with the controls of those macroeconomic indicators; we also
test whether those banking institutional characteristics that affect the risk management of banks,
including government ownership of banks, bank accounting disclosure standards, bank audit man-
agement, and the existence of deposit insurance scheme, would also affect the predictive power of
1 The other two pillars are minimum capital requirement (Pillar 1) and supervisory review power (Pillar 2).
2 Traditional studies generally examine larger sample sizes by utilizing low-frequency macroeconomic data (e.g.,
Demirguc-Kunt and Detragiache, 1998 include 45–65 economies in their study). Studies using market information of
banks are subject to the availability of high frequency data. To our knowledge, this is the first empirical research using
market information of banks to predict banking crises for a large number of countries.
F. Moshirian, Q. Wu / Int. Fin. Markets, Inst. and Money 19 (2009) 351–370 353
bank volatility. We run the tests for the full sample of all markets and the subsamples of developed
markets and emerging markets, respectively, which enable us to identify the differences between
developed and emerging markets.
We contribute to the literature in a few respects. First, we use the bank stock prices to assess
the probability of systemic banking crises for a large number of countries, with controls of the
macroeconomic and banking institutional determinants of systemic banking crises. Although the
research on the probability of systemic banking crises using macroeconomic and banking insti-
tutional indicators is prolific, empirical studies using the market information of banks to predict
systemic banking crises are relatively meager. We adopt the unique disaggregated approach from
Campbell et al. (2001) to measure bank volatility using exclusive market information of banks.
This approach is superior to the traditional volatility measure of moving standard deviation; it
extracts the individual industrial shock from the market and takes into account the market capital-
ization of the components and the variations of all individual bank stock prices within the period
rather than between the periods. Secondly, we improve the estimations by using lagged variables
on the right hand side. Previous research uses almost exclusively contemporaneous variables on
the right hand side (e.g., Demirguc-Kunt and Detragiache, 1998; Beck et al., 2006), and therefore
the direction of causality is ambiguous, which limits the usefulness of the findings. Thirdly, most
of the previous research applies a common methodology to all countries without differentiating
the level of market development. We acknowledge this difference by analyzing the full sample of
all countries as well as the subsamples of developed and emerging markets, respectively. We find
that the marginal contribution of bank volatility to the probability of the occurrence of systemic
banking crises is negligible in the sample for all markets, and this result is mainly driven by
the data from the emerging markets. When we test the subsamples of developed and emerging
markets, however, we find that bank volatility is a significant predictor of banking crises for the
subsample of developed markets, even after being controlled for those macroeconomic indicators.
This result indicates that the market forces are more powerful in promoting the safety and sound-
ness of the banking system in developed markets. We also find that those macroeconomic and
banking risk management indicators have different impacts on the probability of banking crises
for the emerging and developed markets, which suggests that the leading indicators of banking
crises could be more country-specific. Therefore, caution needs to be taken in interpreting the
traditional cross-country results.
The paper is structured as follows: Section 2 describes the data sets and the summary statis-
tics. Section 3 presents the methodology. Section 4 reports the empirical results, and Section 5
concludes.
2. Data and summary statistics
The data sets include the information of listed banks, macroeconomic environment, and bank
institutional characteristics from 18 developed and 18 emerging markets covering the period from
1980 to 2001. The selection of markets is based on data availability.3 In column 2 of Appendix
A is the number of listed banks that we use to construct the bank portfolio for each market, with
the maximum number of 1236 listed banks for the U.S. and the minimum of 5 for Finland. The
3 The selection of markets is based on the availability of data on bank equity prices, quarterly macroeconomic time
series and short-term interest rates. After this selection process, we have complete data sets on 18 developed markets and
18 emerging markets.
sample period for each market is shown in column 3 of Appendix A, with the longest period of
22 years and the shortest of 6 years.
We construct the dummy variable CRIS that takes a value of 1 when a country was undergoing
a systemic banking crisis. The crisis periods are derived from Caprio and Klingebiel (2003)4 and
presented in column 4 of Appendix A.Four out of 18 developed markets, and 8 out of 18 emerging
markets have experienced at least 1 year of systemic banking crises during their sample periods.
2.1. Banking industry volatility
Our primary interest variable is banking industry volatility. We use quarterly and weekly data
of individual banks and market price index as well as interest rates to construct quarterly bank
volatility for each market. The weekly and quarterly stock prices and market capitalizations
of individual banks and the market price index for each market are extracted from Datastream
International, adjusted for capital changes. The interest rates are derived complementally from
International Financial Statistics (International Monetary Fund) and Datastream International.
We use the disaggregated approach from Campbell et al. (2001) to measure banking industry
volatility. The construction of the banking industry volatility variable is achieved in a few steps.
First, we construct the portfolio of listed banks for each market (the number of banks for each
market is shown in column 2 of Appendix A), then we construct the value-weighted excess returns
on bank portfolios for each market. The excess return is computed as the continuous stock return
over risk-free rate Rf . For risk-free rate, we use the 3-month Treasury Bill rate, call money market
rate, 3-month deposit rate, in superior turns, which depends on the availability of data. Weights are
based on market capitalization (MC). The weight of bank j in period t is the market capitalization
of bank j over the total market capitalization of the banking sector at the end of period t − 1 and
keeps constant within period t.
In the second step, we construct the excess return on the market index for each market, and
then run the following regressions to obtain the beta for each market, assuming the beta is constant
over the sample period:
Rit = βim Rimt + ε̂it
where Rit is the quarterly value-weighted bank excess return in country i, Rimt the quarterly market
excess return of country i, and βim is the beta of the banking industry with respect to the market
in country i. ε̂it is the error term.
In the third step, we use the weekly data to construct quarterly bank volatility VOLit . The
details are as follows:
2
VOLit = Var(Rit ) = βim
2
Var(Rmit ) + σ it
where
2

Var(Rmit ) = (Rmiw − μmit )2 , σ it = (Riw − βim Rmiw )2
w∈t w∈t
where Riw is the value-weighted weekly excess return of the banking industry in country i, Rmiw
the weekly excess market return in country i, μmit the average weekly excess market return for
4 Caprio and Klingebiel (2003) identify 117 episodes of systemic banking crises (defined as much or all of bank capital
being exhausted) covering 93 countries since the late 1970s.

country i over period t (here t is quarter), and βim is the beta of the banking industry with respect to
the market in country i. To get weekly excess return, we subtract the weekly risk-free rate which
is derived by dividing the annual short-term interest rate by 52 weeks.
2.2. Other variables
Based on the findings of previous empirical studies (e.g., Demirguc-Kunt and Detragiache,
1998; Hardy and Pazarbasioglu, 1998, 1999; Kaminsky and Reinhart, 1999) that systemic banking
crises tend to erupt in a weak macroeconomic environment, we construct the variables of the
macroeconomic determinants of banking crises with the definitions and data sources presented in
panel A of Table 1 .
We also identify a few banking institutional indicators that are mainly related to banking risk
management, and investigate their impact on banking volatility in forecasting systemic banking
crises. Panel B of Table 1 provides the descriptions and data sources of these variables.
2.3. Summary statistics of the data
Table 2 presents the summary statistics of the quarterly time-series variables and their correla-
tion matrix for the full sample, while those for the subsamples of developed markets and emerging
markets are presented in panels A and B of Table 3, respectively. The information on government
ownership of banks and other banking risk management indicators for each market is presented
in Appendix A.
In panel A of Table 2 is the descriptive statistics of the dependent variable, our primary
independent variable—the lagged banking industry volatility, and the lagged macroeconomic
time series for all markets; in panel B is their correlation matrix while letters ‘a’, ‘b’, and ‘c’
indicates the significance level of the correlation at 1%, 5%, and 10% respectively. The average
variance of the quarterly returns on the banking industry is 0.72%, with the range from 0.02% to
343.31%. The average quarterly growth rates is 0.34% with the range from −13.11% to 19.44%.
The mean of the real interest rates is 3.19% which is close to their median of 3.35%, with the
maximum rate of 51.73% and the minimum of −90.12%. The inflation rates average 7.45% and
range from −5.88% to 123.7%. Some markets have experienced deflation during their sample
periods. The volatility of growth rates spans from 0.02% to 15.64%, with a mean of 1.42%. The
average of domestic credit growth rates, the ratios of M2 against reserves, and the change rates
of exchange rates is 1.34%, 8.58, and −0.95%, respectively.
From the simple correlation matrix in panel B of Table 2, the banking crisis is positively
correlated with the lagged banking industry volatility, lagged real interest rates, and the lagged
growth volatility, but negatively related to other variables. Banking industry volatility is positively
correlated with the contemporaneous inflation rate, growth volatility and domestic credit growth
rate, but negatively associated with the contemporaneous real GDP growth, real interest rate,
M2-to-reserves ratio, and the change of exchange rate. The inflation rate is highly correlated
(P-value = 0.00) with the contemporaneous growth volatility and the real interest rate, with the
simple correlation of 0.705 and −0.495, respectively.
The summary statistics of the variables and their correlation matrices for developed and emerg-
ing markets are reported in panels A and B of Table 3, respectively. Overall, the emerging markets
have higher mean values and larger ranges of variables. The mean value of real quarterly growth
rates for the sample of developed markets is 0.27%, compared to that of 0.46% for the sample
of emerging markets. The quarterly banking volatility is much higher for the sample of emerg-
Table 1
Descriptions and sources of the variables
Variables Descriptions and data sources
(A) Macroeconomic determinants of banking crises

GDP growth rates (GROWTH) GROWTH is measured as the log differences of the GDP
time series. The quarterly GDP time series are
complementally from International Financial Statistics
(IFS), Datastream International, OECD national account.
For Bangladesh and India, we use industrial production
series due to the unavailability of the quarterly GDP series.
The GDP time series are constant prices; except for Austria,
Hong Kong, Greece, Indonesia, Malaysia, Peru,
Philippines, Poland, Taiwan, Thailand and Turkey, all series
are seasonally adjusted
Inflation rates (INFLATION) INFLATION is measured as the rate of changes in
consumer price indices (CPI), the data are obtained from
IFS (line 64). The CPI for Taiwan is derived from
Datastream International
Interest rates (REALINT) REALINT is measured as the difference between short-term
interest rates and the contemporaneous inflation rates. The
short-term interest rates are the 3-month Treasury Bill rates,
or call money market rates, or 3-month deposit rates, in
superior turns, which depends on the availability of data.
The interest rates and inflation rates are acquired
complementally from IFS and Datastream International
Domestic credit growth rates (GCREDIT) GCREDIT is defined as the growth rates of domestic credit
(line 32 of IFS) minus the contemporaneous inflation rates.
The domestic credits and inflation rates are from IFS. We
have the data set for all markets except for Taiwan. Most of
the member countries of the European Monetary Union lose
the observations from 1999 after the circulation of Euro
currency
M2/reserves (M2RESERVE) M2RESERVE is the ratio of M2 to foreign exchange
reserves of the Central Bank. M2 is the money plus
quasi-money (line 34 plus line 35 from IFS). Reserves are
from IFS (line 14). The observations for most of the
member countries of the European Monetary Union are up
to the end of 1998
Depreciation/devaluation (DEPRE) DEPRE is the rate of change of the exchange rates. We use
the Special Drawing Right (SDR), a currency basket
provided by the IMF, as our base currency. We convert the
series of national currency per SDR in IFS to compute the
change rates of each national currency
Growth volatility (VGROWTH) VGROWTH is measured as the four-quarter moving
standard deviations of the real GDP growth rates
(B) Indicators of banking risk management
Deposit insurance scheme (DINS) DINS takes a value of 1 when the country has an explicit
deposit insurance scheme during the systemic banking
crisis. Data sources: Demirguc-Kunt and Sobaci (2001)
Table 1 (Continued )
Variables Descriptions and data sources
Government ownership of banks (GOV) La Porta et al. (2002) document the percentage of
government ownership of banks in 1970 (Gov70) and 1995
(Gov95), respectively, for 92 markets. We construct the
dummy variable GOV that takes a value of 1 when the
percentage of government ownership of banks exceeds the
median of the sample group (all markets, developed markets
and emerging markets) using the values of Gov70 for years
prior to 1990 (included) and the values of Gov95 for years
after 1990
The index of bank accounting disclosure standards We construct this index based on the following five
(ACCT) questions: (1) Does the income statement include accrued
or unpaid interest or principal on nonperforming loans? (2)
Are banks required to produce consolidated financial
statements, including nonbank financial affiliates or
subsidiaries? (3) Are off-balance sheet items disclosed to
the public? (4) Are banks’ directors legally liable for
misleading or erroneous information? (5) Have penalties for
compliance failure been enforced? We assign a value of 1
for each question when the answer is YES, resulting in a
maximum value of 5. Data source: Barth et al. (2001)
The index of bank audit management (AUDIT) We construct the index from Barth et al. (2001) based on the
following eight questions: (1) Is an external audit
compulsory? (2) Are there specific requirements for the
extent of the audit? (3) Are auditors licensed or certified?
(4) Is the auditor’s report given to a supervisory agency? (5)
Can supervisors meet auditors to discuss the report without
bank approval? (6) Are auditors legally required to report
misconduct by managers/directors to a supervisory agency?
(7) Can legal action against external auditors be taken by
the supervisor for negligence? (8) Has legal action been
taken against an auditor in the last 5 years? We assign a
value of 1 to each question if the answer is YES; therefore
the maximum index is eight if the answer is YES for all
questions
ing markets than for that of developed markets, with the mean value of 1.48% and the range
from 0.02% to 343.31% for emerging markets, compared to the mean value of 0.27% and the
range from 0.02% to 5.46% for the developed markets. The simple correlation between banking
crisis and lagged banking industry volatility is significantly positive for the developed markets
(P-value = 0.00), but weakly negative for the emerging markets, with the corresponding figure of
0.146 and −0.011, respectively. The banking crisis is more correlated with lagged M2-to-reserves
ratio for emerging markets, but is more associated with lagged banking industry volatility, lagged
inflation rate and lagged domestic credit growth rate for the developed markets. The inflation rate
is highly correlated with the contemporaneous real interest rate, growth volatility and domestic
credit growth rate for emerging markets, with the simple correlation of −0.558, 0.769 and 0.407,
respectively. The simple correlations show that the developed markets and emerging markets
could have different significant determinants of the occurrence of systemic banking crises.
Overall, the simple correlations indicate that systemic banking crises are more likely to occur
in a developed country with higher banking industry volatility, lower inflation and lower domestic
358
Table 2
Summary statistics of banking crises and macroeconomic indicators: all markets
F. Moshirian, Q. Wu / Int. Fin. Markets, Inst. and Money 19 (2009) 351–370

CRIS VOL (%) GROWTH (%) REALINT (%) INFLATION (%) VGROWTH (%) GCREDIT (%) M2RESERVE DEPRE (%)
(A)
Mean 0.11 0.72 0.34 3.19 7.45 1.42 1.34 8.58 −0.95
Median 0 0.25 0.32 3.35 3.77 0.39 0.94 8.06 −0.36
S.D. 0.32 7.98 2.49 6.46 13.72 2.47 2.55 4.95 5.50
Min 0 0.02 −13.11 −90.12 −5.88 0.02 −10.00 1.53 −45.39
Max 1 343.31 19.44 51.73 123.70 15.64 35.42 37.04 35.21
NOB 1917 1917 1917 1917 1917 1917 1723 1724 1825
No. of countries 36 36 36 36 36 36 35 35 36
(B) Correlation matrix
CRIS 1
VOL 0.010 1
GROWTH −0.021 −0.003 1
REALINT 0.057 b −0.019 −0.011 1
INFLATION −0.005 0.058 b 0.013 −0.495 a 1
VGROWTH 0.066 a 0.029 0.020 −0.180 a 0.705 a 1
GCREDIT −0.040 c 0.015 −0.057 b −0.036 0.388 a 0.353 a 1
M2RESERVE −0.020 −0.045 c −0.021 0.058 b −0.277 a −0.273 a −0.155 a 1
DEPRE −0.044 c −0.020 −0.046c −0.016 −0.333 a −0.304 a −0.359 a 0.117 a 1
CRIS is the dummy variable that takes a value of 1 when a country was experiencing a systemic banking crisis. VOL, GROWTH, REALINT, INFLATION, VGROWTH,
GCREDIT, M2RESERVE, and DEPRE are the bank volatility, real GDP growth rates, real interest rates, inflation rates, growth volatility, domestic credit growth rates, M2-to-
reserves ratios, and the changes of exchange rates, respectively. The data sources and the construction of variables can be found in Section 3. a, b, and c indicate the significant
level of 1%, 5%, and 10%, respectively.
Table 3
Summary statistics of banking crises and macroeconomic indicators: developed and emerging markets
CRIS VOL (%) GROWTH REALINT INFLATION VGROWTH GCREDIT M2RESERVE DEPRE (%)
(%) (%) (%) (%) (%)
(A) Developed markets

Mean 0.06 0.29 0.27 3.30 3.78 0.81 0.86 10.37 −0.22

Median 0 0.20 0.29 3.12 2.80 0.25 0.74 10.17 0.00
S.D. 0.24 0.38 1.44 2.81 3.29 1.44 1.95 4.92 3.78
Min 0 0.02 −6.37 −6.43 −5.88 0.02 −5.53 1.97 −19.52
Max 1 5.46 9.40 17.39 21.75 6.59 35.42 37.04 13.95
NOB 1220 1220 1220 1220 1220 1220 1084 1085 1136
No. of countries 18 18 18 18 18 18 18 18 18
Correlation matrix
CRIS 1
VOL 0.146 a 1
GROWTH −0.032 −0.006 1
REALINT 0.038 0.118 a −0.027 1
INFLATION −0.137 a 0.102 a −0.032 −0.130 a 1
VGROWTH −0.027 0.146 a 0.006 −0.060 b 0.111 a 1
GCREDIT −0.117 a −0.028 −0.020 0.007 0.147 a −0.034 1
M2RESERVE 0.026 −0.028 −0.006 −0.027 −0.181 a −0.086 a −0.024 1
DEPRE −0.005 −0.128 a 0.032 −0.036 −0.034 −0.014 −0.067 b 0.030 1
(B) Emerging markets

Mean 0.20 1.48 0.46 3.00 13.89 2.49 2.15 5.53 −2.17
Median 0 0.46 0.48 3.99 7.05 1.34 1.71 4.57 −1.17
S.D. 0.40 13.19 3.65 10.05 20.84 3.36 3.18 3.21 7.36
Min 0 0.02 −13.11 −90.12 −5.81 0.04 −10.00 1.53 −45.39
Max 1 343.31 19.44 51.73 123.70 15.64 33.87 15.71 35.21
NOB 697 697 697 697 697 697 639 639 689
No. of countries 18 18 18 18 18 18 17 17 18
Correlation matrix
CRIS 1
VOL −0.011 1
GROWTH
359
360
Table 3 (Continued )
CRIS VOL (%) GROWTH REALINT INFLATION VGROWTH GCREDIT M2RESERVE DEPRE (%)
(%) (%) (%) (%) (%)
−0.028 −0.006
REALINT 0.075 b −0.020 −0.007 1
INFLATION −0.090 b 0.035 0.004 −0.558 a 1
VGROWTH 0.007 0.003 0.009 −0.211 a 0.769 a 1
GCREDIT −0.077 b −0.003 −0.089 b −0.032 0.407 a 0.440 a 1
M2RESERVE 0.210 a −0.026 −0.006 0.100 a −0.206 a −0.234 a −0.096 b 1
DEPRE −0.012 −0.007 −0.066 c −0.019 −0.357 a −0.361 a −0.470 a 0.057 1
CRIS is the dummy variable that takes a value of 1 when a country was experiencing a systemic banking crisis. VOL, GROWTH, REALINT, INFLATION, VGROWTH,
GCREDIT, M2RESERVE, and DEPRE are the bank volatility, real GDP growth rates, real interest rates, inflation rates, growth volatility, domestic credit growth rates, M2-to-
reserves ratios, and the changes of exchange rates, respectively. The data sources and the construction of variables can be found in Section 3. a, b, c indicate the significant level
of 1%, 5%, and 10%, respectively.
credit growth. But the emerging markets have a different pattern, the banking crises are more
likely to happen in an emerging country with a higher real interest rate and especially with a
higher M2-to-reserves ratio.
Appendix A presents the summary information on indicators related to bank risk management
for each market. In column 4 is the period when a country experienced a systemic banking
crisis. In columns 5 and 6 are the percentages government ownership of banks in 1970 (Gov70)
and 1995 (Gov95), respectively. The median percentage government ownership of banks for all
markets falls to 26.845 in 1995 from 52.04 in 1970. The developed markets have a much lower
government ownership of banks, with the median percentage values of 22.87 for Gov70 and
12.84 for Gov95, while the corresponding figures for emerging markets are 73.41 and 55.19,
respectively. Columns 7–9 are the index of bank accounting disclosure standards (ACCT), the
index of bank auditing management (AUDIT), respectively. These two variables are constructed
from Barth et al. (2001), we do not have the information of Hong Kong (China), Norway and
Colombia. A higher value of ACCT indicates greater disclosure, and a higher value of AUDIT
indicates stricter audit management for banks. Bangladesh, India and Taiwan have the lowest
ACCT value of 2, while Austria, Germany, Argentina and Bangladesh have the greatest AUDIT
value of 8. The last column of Appendix A presents the year when an explicit deposit insurance
scheme was enacted for each market. Most of the markets have an explicit deposit insurance
scheme, except for two developed markets5 and six emerging markets.
3. Methodology
We use a general Logit model to estimate the probability of banking crises. This approach
has been widely used in the studies on banking crises, and is the preferred method in our study
for the following reasons: first, the estimation is easy to implement by using the conventional
maximum likelihood estimation; while the presence of individual effects, either fixed effects or
random effects, complicates the estimation when it comes to discrete dependent variable models.6
Secondly, a general Logit model also allows us to undertake the analyses for a large number of
countries by including those countries that did not experience banking crises.7 However, since
our study does not account for the cross-sectional individual effects that are independent of the
explanatory variables included in the regressions, our results need to be explained with caveat.
Our dependent variable Y(i,t) is the crisis dummy, it takes a value of 1 when a country is
experiencing a systemic banking crisis, and 0 otherwise. Let P(i,t) denotes the probability that a
banking crisis will occur in country i at period t, we assume that the P(i,t) follows the logistic
5 We construct a dummy variable that takes a value of one when a country has an explicit deposit insurance scheme
during the systemic banking crisis. Sweden established an explicit deposit insurance scheme (1996) after the occurrence
of the systemic banking crisis (1991), therefore we assign a value of zero for Sweden.
6 For a random effect model, the joint likelihood of the occurrence of an event can no longer be written as the product of
the marginal likelihoods, which therefore complicates the derivation of maximum likelihood and the estimation becomes
problematic (Baltagi, 2001, p. 206). For a fixed-effect Probit model, there is no feasible way to remove the heterogeneity,
and with large numbers of cross-sectional units, estimation of the fixed effects is intractable (Greene, 2000, p. 837).
7 For a fixed-effect Logit model, the parameters can be estimated using the conventional maximum likelihood Logit
program by sweeping away the fixed effects, however, it requires those observations for individuals that do not switch
status to be omitted from the panel data (Greene, 2000). In another words, with the fixed-effect Logit model, only those
countries that have experienced systemic banking crises can be included in the estimation. This will eliminate a large
amount of useful information from the countries that did not experience banking crises, and lead to infeasibility of the
estimation with a large number of countries.
distribution:
1
P(i,t) = E(Yi,t = 1|X) = = F (β X)
1 + e−β X
where X is the vector of explanatory variables including the constant term, β the vector of unknown
coefficients, and F(β X) is the logistic distribution function.
The estimation of binary choice model is usually based on the method of maximum likelihood.
The joint probability or the likelihood function of the model is
n
Y
[F (β Xi )] i [1 − F (β Xi )] i
1−Y
L=
i=1
where n is the number of observations, which is the sum of time periods for all countries, Yi the
dummy that takes a value of 1 when a banking crisis occurs. The vector of coefficients could be
estimated by maximizing the following log-likelihood function:
n

Ln L = {Yi Ln F (β Xi ) + (1 − Yi ) Ln[1 − F (1 − β Xi )]}
i=1
In the linear probability model (LPM), the slope coefficient directly measures the change in the
probability of an event occurring as the result of a unit change in the value of the corresponding
explanatory variable. However, the slope coefficient in the Logit model, instead, measures the
change of the log-odds, Ln(P(i,t) /(1 − P(i,t) )), in favor of the occurrence of an event given a unit
change in the value of the corresponding independent variable. Therefore, in interpreting our
regression results, we need to bear in mind that the coefficient reflects the effect of a change of
a corresponding variable on the log-odds in favor of the occurrence of systemic banking crises.
The rate of change in the probability with respect to an explanatory variable depends not only on
its coefficient but also on the initial values of all the explanatory variables and their coefficients.
While the sign of the coefficient does indicate the direction of the change in the probability, the
magnitude of the change in the probability relies on the level of the probability from which the
change is measured. In other words, the effect of the value change of an explanatory variable on
the probability of the occurrence of a banking crisis depends on the country’s initial probability
of a banking crisis. If the initial crisis probability of a country is close to 0 or 1, a unit change
in the explanatory variable would have the least effect on the change in crisis probability. When
the initial crisis probability of a country is in the intermediate range which is close to 0.5, a unit
change in the same explanatory variable would have the greatest effect on the change in crisis
probability.
4. Empirical results
Table 4 reports the estimation results for the full sample, while Tables 5 and 6 present the esti-
mation results for the subsamples of developed and emerging markets, respectively. The dependent
variable is the dummy of banking crises that takes a value of 1 when a country is experiencing a
systemic banking crisis, and 0 otherwise. The P-values are in the parentheses. LR is the value of
the likelihood ratio test with the null hypothesis that the slope coefficients are 0. The LR statistic
follows a chi-squared distribution with degrees of freedom equal to the number of restrictions
imposed.
Table 4
Bank volatility and systemic banking crises: all markets
1 2 3 4 5 6 7 8 9
Constant −2.07 [0.000] −2.29 [0.000] −2.17 [0.000] −2.03 [0.000] −1.98 [0.000] −2.11 [0.000] −1.94 [0.000] −1.95 [0.000] −2.01 [0.000]
VOL 0.27 [0.686] 0.40 [0.571] 0.37 [0.603] 0.33 [0.631] 0.33 [0.640] 96.17 [0.000] 3.78 [0.893] −44.53 [0.002] 0.21 [0.794]
GROWTH −2.09 [0.417] −2.32 [0.374] −2.72 [0.297] −3.20 [0.224] −3.00 [0.264] −3.27 [0.213] −3.18 [0.233] −2.93 [0.268]
REALINT 3.33 [0.022] 2.13 [0.159] 2.24 [0.141] 1.55 [0.314] 1.71 [0.267] 0.96 [0.549] 0.25 [0.876] 1.34 [0.387]
INFLATION −2.17 [0.017] −2.67 [0.005] −1.70 [0.075] −2.06 [0.037] −1.16 [0.234] −2.39 [0.020] −3.07 [0.003] −2.29 [0.023]
VGROWTH 15.96 [0.000] 15.21 [0.000] 16.64 [0.000] 16.38 [0.000] 13.07 [0.002] 17.84 [0.000] 16.74 [0.000] 16.13 [0.000]
DEPRE −2.16 [0.131] −3.03 [0.045] −2.20 [0.150] −3.03 [0.051] −2.62 [0.088] −2.93 [0.053]
GCREDIT −9.88 [0.019] −11.82 [0.007] −10.99 [0.013] −11.10 [0.015] −11.58 [0.011] −12.22 [0.006]
M2RESERVE −0.01 [0.667] −0.01 [0.635] −0.02 [0.213] −0.01 [0.535] −0.01 [0.622] −0.01 [0.703]
GOV × VOL −96.00 [0.000]
ACCT × VOL −0.88 [0.901]
AUDIT × VOL 11.19 [0.002]
INS × VOL 11.85 [0.124]
No. of crisis 215 215 215 207 207 207 195 195 207
NOB 1917 1917 1825 1723 1723 1723 1632 1632 1723
LR (zero slope) 0.138 23.998 a 23.214 a 26.817 a 30.78 a 68.608 a 30.623 a 39.953 a 33.803 a
Correct prediction (%) 88.78 88.84 88.27 87.99 88.04 87.93 88.05 87.75 87.93
We estimate the following Logit model specification: L(i,t) = Ln((P(i,t) )/(1 − P(i,t) )) = α + β1 VOL(i,t−1) + β2 GROWTH(i,t−1) + β3 REALINT(i,t−1) + β4 INFLATION(i,t−1) + β5 VGROWTH(i,t−1) + β6 GCREDIT(i,t−1) +
β7 M2RESERVE(i,t−1) + β8 DEPRE(i,t−1) + β9 (BANK × VOL(i,t−1) ) + ε(i,t) , where L(i,t) is the Logit, namely the log of the odds ratio in favor of the occurrence of a systemic banking crisis at time t for country i. VOL, GROWTH,
REALINT, INFLATION, VGROWTH, GCREDIT, M2RESERVE, and DEPRE are the bank volatility, real GDP growth rates, real interest rates, inflation rates, growth volatility, domestic credit growth rates, M2-to-reserves ratios, and
the changes of exchange rates, respectively. BANK is the indicators of banking risk management, which represents the dummy variable of government ownership of banks (GOV), the banking accounting disclosure standards (ACCT),
the banking audit management (AUDIT), and the dummy variable of the existence of deposit insurance scheme (INS), respectively. ε(i,t) is the disturbance. We use the lagged independent variables to control for the simultaneous effect.
P-values are in the parentheses. a, b, and c indicate the significant level of 1%, 5%, and 10%, respectively.
363
364
Table 5
Bank volatility and systemic banking crises: developed markets
1 2 3 4 5 6 7 8 9
Constant −3.01 [0.000] −2.05 [0.000] −1.83 [0.000] −1.54 [0.000] −1.54 [0.000] −1.45 [0.000] −1.44 [0.002] −1.19 [0.006] −2.02 [0.000]
VOL 83.45 [0.000] 106.32 [0.000] 109.08 [0.000] 112.16 [0.000] 112.04 [0.000] 13.18 [0.742] 392.12 [0.004] 569.17 [0.000] −557.86 [0.027]
GROWTH −19.29 [0.114] −18.62 [0.127] −18.67 [0.115] −18.66 [0.116] −21.91 [0.082] −22.83 [0.064] −23.74 [0.055] −21.67 [0.074]
REALINT 0.31 [0.948] −2.44 [0.608] −6.44 [0.201] −6.47 [0.199] −10.56 [0.045] −9.29 [0.129] −11.32 [0.077] −2.52 [0.629]
INFLATION −30.39 [0.000] −31.15 [0.000] −26.44 [0.000] −26.46 [0.000] −32.98 [0.000] −32.29 [0.000] −42.12 [0.000] −30.59 [0.000]
VGROWTH −24.32 [0.055] −25.66 [0.042] −26.74 [0.040] −26.78 [0.040] −13.93 [0.303] −0.55 [0.965] 4.04 [0.752] 30.32 [0.069]
DEPRE 1.04 [0.749] −0.32 [0.923] −2.13 [0.548] −2.02 [0.587] −1.73 [0.653] −3.56 [0.320]
GCREDIT −45.62 [0.000] −45.70 [0.000] −44.93 [0.000] −70.64 [0.000] −80.88 [0.000] −50.25 [0.000]
M2RESERVE 0.00 [0.872] 0.00 [0.870] 0.01 [0.718] 0.01 [0.724] 0.04 [0.204] 0.01 [0.846]
GOV × VOL 214.66 [0.000]
ACCT × VOL −82.4 [0.012]
AUDIT × VOL −129.37 [0.000]
INS × VOL 730.04 [0.003]
No. of crisis 76 76 76 76 76 76 64 64 76
NOB 1220 1220 1136 1084 1084 1084 1021 1021 1084
LR (zero slope) 15.587 a 55.067 a 58.751 a 75.486 a 75.495 a 100.072 a 79.694 a 109.319 a 102.22 a
β7 M2RESERVE(i,t−1) + β8 DEPRE(i,t−1) + β9 (BANK × VOL(i,t−1) ) + ε(i,t) , where L(i,t) is the Logit, namely the log of the odds ratio in favor of the occurrence of a systemic banking crisis at time t for country i. VOL, GROWTH,
REALINT, INFLATION, VGROWTH, GCREDIT, M2RESERVE, and DEPRE are the bank volatility, real GDP growth rates, real interest rates, inflation rates, growth volatility, domestic credit growth rates, M2-to-reserves ratios, and
the changes of exchange rates, respectively. BANK is the indicators of banking risk management, which represents the dummy variable of government ownership of banks (GOV), the banking accounting disclosure standards (ACCT),
the banking audit management (AUDIT), and the dummy variable of the existence of deposit insurance scheme (INS), respectively. ε(i,t) is the disturbance. We use the lagged independent variables to control for the simultaneous effect.
P-values are in the parentheses. a, b, c indicate the significant level of 1%, 5%, and 10%, respectively.
Table 6
Bank volatility and systemic banking crises: emerging markets
Emerging market 1 2 3 4 5 6 7 8 9
Constant −1.39 [0.000] −1.38 [0.000] −1.31 [0.000] −2.41 [0.000] −2.36 [0.000] −2.31 [0.000] −2.24 [0.000] −2.36 [0.000] −2.34 [0.000]
VOL −0.30 [0.770] −0.04 [0.970] −0.06 [0.953] 0.12 [0.889] 0.12 [0.898] 140.59 [0.000] −24.47 [0.314] −44.20 [0.004] 0.15 [0.858]
GROWTH −2.35 [0.404] −2.75 [0.338] −2.82 [0.311] −3.32 [0.241] −2.89 [0.311] −3.21 [0.257] −3.02 [0.293] −3.42 [0.229]
REALINT 0.48 [0.759] −0.30 [0.857] 0.64 [0.698] 0.09 [0.957] 0.40 [0.821] 0.43 [0.802] −0.29 [0.867] 0.14 [0.935]
INFLATION −4.37 [0.000] −4.78 [0.000] −4.52 [0.000] −4.78 [0.000] −4.10 [0.002] −4.17 [0.002] −4.95 [0.000] −4.77 [0.000]
VGROWTH 20.00 [0.000] 19.67 [0.000] 28.43 [0.000] 28.31 [0.000] 26.81 [0.000] 24.99 [0.000] 24.88 [0.000] 28.70 [0.000]
DEPRE −1.91 [0.230] −2.16 [0.209] −1.75 [0.356] −2.17 [0.207] −1.85 [0.274] −2.17 [0.205]
GCREDIT −8.54 [0.075] −10.27 [0.042] −9.07 [0.072] −10.67 [0.034] −10.93 [0.030] −10.17 [0.043]
M2RESERVE 0.17 [0.000] 0.17 [0.000] 0.10 [0.006] 0.16 [0.000] 0.17 [0.000] 0.17 [0.000]
GOV × VOL −140.67 [0.000]
ACCT × VOL 6.11 [0.312]
AUDIT × VOL 11.04 [0.004]
INS × VOL 4.19 [0.665]
No. of crisis 139 139 139 131 131 131 131 131 131
NOB 697 697 689 639 639 639 611 611 639
LR (zero slope) 0.112 21.778 a 23.035 a 54.459 a 56.046 a 93.922 a 49.066 a 56.737 a 56.28 a
β7 M2RESERVE(i,t−1) + β8 DEPRE(i,t−1) + β9 (BANK × VOL(i,t−1) ) + ε(i,t) , where L(i,t) is the Logit, namely the log of the odds ratio in favor of the occurrence of a systemic banking crisis at time t for country i. VOL,
GROWTH, REALINT, INFLATION, VGROWTH, GCREDIT, M2RESERVE, and DEPRE are the bank volatility, real GDP growth rates, real interest rates, inflation rates, growth volatility, domestic credit growth rates, M2-to-reserves
ratios, and the changes of exchange rates, respectively. BANK is the indicators of banking risk management, which represents the dummy variable of government ownership of banks (GOV), the banking accounting disclosure standards
(ACCT), the banking audit management (AUDIT), and the dummy variable of the existence of deposit insurance scheme (INS), respectively. ε(i,t) is the disturbance. We use the lagged independent variables to control for the simultaneous
effect. P-values are in the parentheses. a, b, and c indicate the significant level of 1%, 5%, and 10%, respectively.
365
In the column headed 1 of each table is the estimation result when the lagged bank volatility
is the only independent variable. The coefficient of the lagged bank volatility is insignificant
but positive for the sample of all markets and negative for the subsample of emerging markets.
However, the coefficient of bank volatility for the subsample of developed markets is positive
and significant at the 1% level. The results show that bank volatility has a different effect on
the occurrence of banking crises. All other things being equal, a higher banking volatility will
indicate a higher probability of the occurrence of banking crises for developed markets; however,
this effect is negligible for emerging markets. This result indicates that the market forces could be
more effective in enhancing the safety and soundness of the banking system for developed markets.
In the columns headed 2–5 of each table are the results when the macroeconomic indicators that
are thought to be the determinants of banking crises are included in the regressions. The coefficient
of banking volatility for the sample of developed markets remains positively significant under
different specifications, which indicates that the effect of banking volatility on the occurrence
of banking crises is independent of those macroeconomic determinants of banking crises. The
coefficients for the full sample and the subsample of emerging markets are insignificant. The
developed markets could have different causes of systemic banking crises from emerging markets,
and the results of the full sample are mainly driven by data from the emerging markets of which 8
out of 12 markets have experienced systemic banking crises during their sample periods. Systemic
banking crises tend to occur in a stagnated economy for developed markets, characterized by
higher banking volatility, falling of inflation rates, lower growth volatility, and lower domestic
credit growth. The banking system is more vulnerable to systemic banking crises in an emerging
market with falling inflation rates, higher variations of GDP growth rates, lower domestic credit
growth and higher M2-to-reserves ratio.
The results are partly different from those of Demirguc-Kunt and Detragiache (1998) who find
that banking crises tend to happen in a macroeconomic environment with higher inflation and
higher domestic credit growth. We conjecture that the differences come from a few aspects: first,
their sample markets cover most of the Latin-American countries that experienced banking crises
during the period of 1980–1994, while our sample countries primarily include the recent banking
crises which occurred in the late 1990s. Secondly, although with the same methodology, they use
the contemporaneous annual independent variables, while we use the lagged quarterly explanatory
variables. They also admit the weakness of using the contemporaneous explanatory variables,
since the direction of causality is ambiguous: the movements in the independent variables may
lead to a banking crisis, equally they may be among the consequences of the banking crisis, or
both may be triggered by some unobservable factors. Thirdly, they use their own definitions to
identify the occurrence of banking crises; we use the database from Caprio and Klingebiel (2003)
to identify the episodes of systemic banking crises. Therefore the incident and period of banking
crises could be different. For example, in their sample, the US, Sweden, and Japan were identified
as having experienced a banking crisis during the periods of 1981–1992, 1990–1993, 1992–1994,
respectively; while in our sample, only Sweden and Japan are identified as having experienced
systemic banking crises during 1991, 1991–2001, respectively. The problem of identifying the
episodes and dates of banking crises is one of the key weaknesses of the current research on
banking crises.8
8 The definition of systemic banking crises and the identification of the dates of banking crises are somewhat arbitrary
in the research on banking crises. The dataset provided by Caprio and Klingebiel (2003) has been widely used by the
recent research, e.g., see Hoggarth et al. (2002) and Bekaert et al. (2005).
The results confirm the findings of Hardy and Pazarbasioglu (1998), who argue that it would
have been difficult to predict the recent Asian banking crises using the traditional macroeco-
nomic indicators, therefore country-specific circumstances should be recognized in assessing the
likelihood of banking crises. Hardy and Pazarbasioglu (1998) also use the region dummy to
control for the region effect, and find that banking crises in different regions exhibit different
patterns. Previous studies primarily focus on cross-country analyses; we separate the sample into
the subsamples of developed markets and emerging markets, and find that different indictors
may play a different role in determining the likelihood of banking crises in different markets.
Higher banking industry volatility significantly contributes to the occurrence of banking crises
for developed markets, but its role is negligible for emerging markets. A higher M2-to-reserves
ratio and higher growth volatility are more pronounced in inducing systemic banking crises for
emerging markets, while lower growth volatility and lower domestic credit growth are more
distinct for the developed markets. Therefore, the cross-country results need to be explained
prudently.
In the columns headed 6–9 of each table are the results when the interaction terms between
bank volatility and the banking risk management indicators are added into the regressions
sequentially. For the sample of all markets, the coefficient of the interaction term between
bank volatility and the dummy of government ownership of banks is negative and significant
at the 1% level, while the coefficient of bank volatility itself becomes significantly positive.
This indicates that, all other things equal, higher bank volatility in a market with higher gov-
ernment ownership of banks is less likely to have banking crises. However, we see different
pictures for the subsamples of developed and emerging markets. The coefficient of the inter-
action term is positive and significant at the 1% level for the developed markets, while the
coefficients of bank volatility and growth volatility lose significance. For the subsample of
emerging markets, the coefficient of the interaction term is significantly negative, while that
of bank volatility is significantly positive. The results of the full sample are mainly induced by
the data from the emerging markets. Therefore, government ownership of banks increases the
probability of banking crises for developed markets whereas it is the opposite case for emerging
markets.
In the column headed 7 is the result when the interaction term between bank volatility and the
index of banking accounting disclosure standards is added. Better banking disclosure improves
the ability of market participants to discipline the excessive risk-taking of banks. In a market
with more transparent banking disclosure, higher bank volatility is less likely to lead to banking
crises. We find that this result is primarily driven by the data from developed markets where
the coefficient of the interaction term is negative and significant at the 5% level. However,
the inclusion of the interaction term does not affect the result of the subsample of emerging
markets.
The column headed 9 of each table reports the results that include the interaction term between
bank volatility and the index of audit management. Effective auditing could ensure that banks abide
by the banking accounting standards, and identify problem banks in a timely way. Better auditing
management enhances the role of bank volatility in predicting banking crises, the coefficient of
the interaction term is positive and significant for the sample of all markets, and this result is
mainly induced by the subsample of the emerging markets. For the subsample of the developed
markets, on the other hand, the coefficient is negative and significant.
In the last column of Table 4, the coefficient of the interaction term between bank volatility
and the dummy variable of the existence of an explicit deposit insurance scheme is positive but
insignificant. In a market with an explicit deposit insurance scheme, higher bank volatility is more
likely to be associated with the occurrence of systemic banking crises, and this relationship is
significant for developed markets but negligible for emerging markets.
In summary, we find that the developed and emerging markets have different leading indicators
of banking crises. For the developed markets, bank volatility consistently leads the occurrence
of banking crises even being controlled for macroeconomic variables, which indicates that the
market forces are more powerful in improving the soundness of the banking system for developed
markets. We also find that those macroeconomic and banking risk management indicators have
different impacts on the onset of banking crises, therefore the cross-country results need to be
interpreted cautiously.
5. Conclusion
It has been well recognized that market discipline is important in ensuring the soundness
and safety of a financial system. The disciplinary strength of market forces derives from the
immense power of the price system to aggregate information (Crockett, 2002). Extensive liter-
ature has examined the prediction of banking crises using balance sheet information of banks
as well as macroeconomic indicators. However, empirical research using market information
of banks to predict banking crises is insufficient. This research contributes to the literature
by using the publicly available market information of banks to predict the probability of
systemic banking crises. We investigate whether banking industry volatility, among those lead-
ing macroeconomic indicators of banking crises, is a useful predictor of systemic banking
crises.
We address this issue using a Logit econometric model to analyze the data from 18 devel-
oped and 18 emerging markets. Our findings have two very important policy implications: first,
market forces may have a different impact on developed and emerging markets. We find that
the marginal effect of bank volatility on the probability of the occurrence of systemic banking
crises is negligible for the sample of all markets. However, when we repeat the same estima-
tions for the subsamples of developed and emerging markets, we find that bank volatility is
a good predictor of banking crises for the subsample of developed markets, and this result
is robust to the controls of macroeconomic indicators that are traditionally thought to be the
determinants of banking crises. This finding indicates that the market forces are more influ-
ential in promoting the safety and soundness of the banking system in developed markets.
Second, the leading indicators of banking crises may be more country-specific. We also find
that those macroeconomic and banking risk management indicators have a different impact
on the probability of banking crises for emerging and developed markets. Therefore, the tra-
ditional cross-country results of the research on banking crises need to be interpreted with
caution.
Acknowledgements
Qiongbing Wu gratefully acknowledges the financial support of the Faculty Postgraduate

Research Scholarship from The Faculty of Commerce and Economics at The University of New
South Wales. We appreciate the helpful comments of Ike Mathur (the editor), an anonymous
referee, and the session participants at the 18th Australiasian Finance and Banking Conference
in Sydney, Australia. All the errors remain ours.
Appendix A. Summary information for individual markets
Markets No. of banks Sample period Crisis years Gov95 Gov70 ACCT AUDIT DINS
Australia 17 Q1/80–Q4/01 12.3 20.9 5 5 NO

Austria 12 Q2/88–Q4/00 50.4 70.8 3 8 1979
Belgium 9 Q2/80–Q4/01 27.6 39.9 4 7 1974
Canada 65 Q1/80–Q4/01 0 11 3 5 1967
Denmark 44 Q2/88–Q4/01 8.87 9.8 4 7 1988
Finland 5 Q3/88–Q4/01 1991–1994 30.7 32.1 5 5 1969
France 27 Q1/88–Q4/01 17.3 74.4 4 6 1980
Germany 32 Q2/91–Q4/01 36.4 51.9 3 8 1966
HK, China 12 Q2/86–Q4/01 0 0 N/A N/A NO
Italy 42 Q1/80–Q4/01 36 75.7 5 2 1987
Japan 99 Q1/80–Q4/01 1991– 0 6.9 3 3 1971
The Netherlands 9 Q1/87–Q4/01 9.2 7.8 3 5 1979
Norway 27 Q2/90–Q4/01 1987–1993 43.7 54.6 N/A N/A 1961
Spain 18 Q3/87–Q4/01 1.98 32.6 5 5 1977
Sweden 7 Q3/82–Q4/01 1991 23.2 20.8 5 5 1996
Switzerland 27 Q2/80–Q4/01 13.4 24.9 5 8 1984
UK 12 Q1/80–Q4/01 0 0 5 5 1982
United States 1236 Q1/80–Q4/01 0 0 5 6 1934
Argentina 8 Q2/93–Q4/01 1995, 2001– 60.5 71.9 3 8 1979
Bangladesh 18 Q3/92–Q4/01 Late 1980s–1996 95 100 2 8 1984
Colombia 14 Q2/94–Q4/01 53.9 57.7 N/A N/A 1985
Greece 13 Q3/88–Q4/01 77.8 92.7 4 5 1993
India 20 Q2/93–Q4/01 84.9 100 2 6 1961
Indonesia 22 Q4/90–Q4/01 1997– 42.9 74.9 4 6 NO
Israel 22 Q4/87–Q4/01 64.6 67.6 4 4 NO
Malaysia 17 Q2/91–Q4/01 1997– 9.93 20 5 7 NO
Peru 11 Q3/92–Q4/01 26.5 87.4 5 7 1992
Philippines 18 Q2/90–Q4/01 1998– 27.2 52.2 5 3 1963
Poland 12 Q2/95–Q4/01 84.3 100 4 5 1995
Portugal 7 Q3/90–Q2/00 25.7 100 4 5 1992
South Africa 9 Q4/95–Q4/01 0 0 4 5 NO
South Korea 14 Q1/85–Q4/01 1997– 25.4 56.6 5 4 1996
Taiwan, China 16 Q4/88–Q4/01 1997–1998 76.5 50.4 2 2 NO
Thailand 9 Q2/93–Q4/01 1997– 17.1 24.1 3 4 NO
Turkey 15 Q3/88–Q4/01 2000– 56.5 81.8 4 7 1983
Venezuela 9 Q4/94–Q4/01 1994–1995 58 82.9 4 6 1985
Gov70 and Gov95 are the percentages of government ownership of banks in 1970 and 1995, respectively; the data are
from Table 1 of La Porta et al. (2002); crisis years is the period when the country experienced systemic banking crisis;
data are from Caprio and Klingebiel (2003). ACCT is the indicator of bank disclosure standards. AUDIT is the indicator
of audit management for banks. ACCT and AUDIT are constructed from Barth et al. (2001). The last column indicates
the year when an explicit deposit insurance scheme was enacted, we construct a dummy variable DINS that takes a value
of 1 when a country has an explicit deposit insurance scheme during a system banking crisis; the data are derived from
Demirguc-Kunt and Sobaci (2001).
References
Baltagi, B.H., 2001. Econometric Analysis of Panel Data. Published by John Wiley & Sons Ltd., England.
Barth, J.R., Caprio, G., Levine, R., 2001. The Regulation and Supervision of Banks Around the World, A New Database.
The World Bank, Washington, DC.
Basel Committee on Banking Supervision, 2003. The New Basel Capital Accord (Consultative Document). Bank for
International Settlements.
Beck, T., Demirguc-Kunt, A., Levine, R., 2006. Bank concentration, competition and crises: first results. Journal of
Banking and Finance 30, 1581–1603.
Beck, T., Levine, R., 2002. Industry growth and capital allocation: does having a market- or bank-based system matter?
Journal of Financial Economics 64, 147–180.
Beck, T., Levine, R., 2004. Stock markets, banks, and growth: panel evidence. Journal of Banking and Finance 28,
423–442.
Beck, T., Levine, R., Loayza, N., 2000. Finance and the source of growth. Journal of Financial Economics 58, 261–300.
Bekaert, G., Harvey, C.R., Lundblad, C., 2005. Does financial liberalization spur growth? Journal of Financial Economics
77, 3–55.
Campbell, J.Y., Lettau, M., Malkiel, B.G., Xu, Y., 2001. Have individual stocks become more volatile? An empirical
exploration of idiosyncratic risk? The Journal of Finance 56, 1–43.
Caprio, G., Klingebiel, D., 2003. Episodes of Systemic and Borderline Financial Crises. World Bank Research Dataset.
The World Bank, Washington, DC.
Cetorelli, N., Gambera, M., 2001. Banking market structure, financial dependence and growth: international evidence
from industry data. The Journal of Finance 56, 617–648.
Crockett, A., 2002. Market discipline and financial stability. Journal of Banking and Finance 26, 977–987.
Demirguc-Kunt, A., Detragiache, E., 1998. The determinants of banking crises in developing and developed countries.
IMF Staff Papers 45, 81–109.
Demirguc-Kunt, A., Maksimovic, V., 1998. Law, finance, and firm growth. The Journal of Finance 53, 2107–2137.
Demirguc-Kunt, A., Maksimovic, V., 2002. Funding growth in bank-based and market-based financial systems: evidence
from firm-level data. Journal of Financial Economics 65, 337–363.
Demirguc-Kunt, A., Sobaci, T., 2001. Deposit insurance around the world: a database. World Bank Economic Review 15,
481–490.
Greene, W.H., 2000. Econometric Analysis. Prentice-Hall Inc., New Jersey.
Hardy, D.C., Pazarbasioglu, C., 1998. Leading Indicators of Banking Crises: Was Asian Different? IMF Working Paper
91.
Hardy, D.C., Pazarbasioglu, C., 1999. Determinants and leading indicators of banking crises: further evidence. IMF Staff
Papers 44, 247–258.
Hoggarth, G., Reis, R., Saporta, V., 2002. Costs of banking system instability: some empirical evidence. Journal of Banking
and Finance 26, 825–855.
Kaminsky, G., Reinhart, C.M., 1999. The twin crises: the causes of banking balance of payments problems. American
Economic Review 89, 473–500.
King, R.G., Levine, R., 1993. Finance and growth: Schumpeter might be right. Quarterly Journal of Economics 108,
717–738.
La Porta, R., Lopez-De-Silanes, F., Shleifer, A., 2002. Government ownership of banks. The Journal of Finance 57,
265–301.
Levine, R., 2005. Finance and growth: theory and evidence. In: Philippe Aghion, Steven Durlauf (Eds.), Handbook of
Economic Growth. North-Holland, Amsterdam, pp. 865–934.
Levine, R., Zervos, S., 1998. Stock markets, banks and economic growth. American Economic Review 88, 537–558.
Rajan, R.G., Zingales, L., 1998. Financial dependence and growth. American Economic Review 88, 559–586.

Banking Industry Volatility and BANKING CRISIS

Uploaded by

Document Information

Original Description:

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

Banking Industry Volatility and BANKING CRISIS

Uploaded by

Copyright:

Available Formats

Available online at www.sciencedirect.

Int. Fin. Markets, Inst. and Money 19 (2009) 351–370

Banking industry volatility and

JEL classiﬁcation: G15; G21; G28

Keywords: Banking crises; Volatility; Market forces

2. Data and summary statistics

2.1. Banking industry volatility

being exhausted) covering 93 countries since the late 1970s.

2.2. Other variables

2.3. Summary statistics of the data

(A) Macroeconomic determinants of banking crises

F. Moshirian, Q. Wu / Int. Fin. Markets, Inst. and Money 19 (2009) 351–370

(A) Developed markets

F. Moshirian, Q. Wu / Int. Fin. Markets, Inst. and Money 19 (2009) 351–370

(B) Emerging markets

Qiongbing Wu gratefully acknowledges the financial support of the Faculty Postgraduate

Appendix A. Summary information for individual markets

Australia 17 Q1/80–Q4/01 12.3 20.9 5 5 NO

You might also like