Financial Liberalization and Banking Crises: Cross-Country Analysis of Liberalization Policies By Apanard Angkinand University of Illinois at Springfield [email protected]Wanvimol Sawangngoenyuang Claremont Graduate University [email protected]Clas Wihlborg Copenhagen Business School [email protected]July, 2007 Abstract: Financial liberalization is often blamed for banking and currency crises in many countries. In this paper, we a sk whether the anecdotal evidence with respect to liberalization and ba nking crises is well founded, and whether explanations of banking crises can be improved by distinguishing among different types of financial liberalization. We use a data set for financial liberalization for42 countries during 1973-2002 and distinguish between liberalization with respect to behavioral liberalization (credit and interest rate liberalization), competitive liberalization (equity market, capital account liberalization, and banks’ entry and activity liberalization), and privatization. The interaction between liberalization and institutional factors such as strength of supervision and deposit insurance coverage is studied as well. J EL Cla ss if ic at io n : G21; G28; F3 Keywords : Banking Crises; Financial Liberalization; Capital Regulation and Supervision __________________________ __We would like to thank Arth ur Denzau, Tripon Phumiwasana, Thomas D. Wil lett, and participants at the 2006 Annual Meeting of the Asi a-Pacific Economic Association, and the 2007 Annual Meeting of the Western Economic Association for their useful comments.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Financial liberalization is often blamed for banking and currency crises in many countries. In this paper, we ask whether the anecdotal evidence with respect to liberalization and banking crises iswell founded, and whether explanations of banking crises can be improved by distinguishingamong different types of financial liberalization. We use a data set for financial liberalization for 42 countries during 1973-2002 and distinguish between liberalization with respect to behavioralliberalization (credit and interest rate liberalization), competitive liberalization (equity market,capital account liberalization, and banks’ entry and activity liberalization), and privatization. Theinteraction between liberalization and institutional factors such as strength of supervision anddeposit insurance coverage is studied as well.
J EL Classification: G21; G28; F3Keywords: Banking Crises; Financial Liberalization; Capital Regulation and Supervision
____________________________ We would like to thank Arthur Denzau, Tripon Phumiwasana, Thomas D. Willett, and participants at the2006 Annual Meeting of the Asia-Pacific Economic Association, and the 2007 Annual Meeting of theWestern Economic Association for their useful comments.
the extreme case, the aggregate index becomes observationally equivalent to indexes for
different dimensions of financial liberalization.
In Section 2 we discuss the role of government in banking while costs and benefits of
financial liberalization are summarized in Section 3. Since we focus on the relationship between
financial liberalization and banking crises, we review the existing empirical literature on the
effects of both domestic and international financial liberalizations on banking crises in Section 4.
Hypotheses are developed in Section 5. Empirical methodology and data are described in Section
6. Results of the empirical tests are presented in Section 7. Conclusions and policy implications
follow in the last Section.
2. The Role of Governments in Banking
A financial system is an important driving force for the growth in the economy, and
banks dominate the financial system in large parts of the world. Banks collect deposits from
individual savers and lend to firms or corporations. By doing so, banks enable consumption
smoothing for households facing volatile income streams. Banks also organize the payment
system by acting as clearing and settlement houses. If a debtor bank is unable to pay its interbank
balance, this deprives other institutions of expected funds and prevents them from settling. A
banking system also enables trading in, and sharing of, risk of firms and projects to allocate
financial resources according to wealth holders’ risk preferences. Hence, if a large part of the banking system fails, economic activity may be severely affected.
To ensure safe and sound banking, government in most countries has strongly regulated
as well as restricted banks’ operations. Banking restrictions take many different forms. For
instance, for fear that excessive capital inflows can promote unsustainable booms followed by
economic contraction, government implements capital controls. To avoid bank failures due to the
rising interest rates, particularly during the period of high inflation, government imposes interest
rate ceiling. Financial repression in the form of controls on interest rates as well as credit
allocation is also one way the government can raise funds at a relatively cheap interest rate rather
than raising tax rates (Giovannini and de Melo, 1993, Fry, 1997), and allocate credit in order to
achieve certain economic policies. Since governments in many countries have used banks to fund
budget deficits, governments have long treated banks as public utilities. Thereby, they have been
reluctant to let troubled banks fail. At the same time, the survival of the banks has been assured
by recapitalizations and/or massive subsidies (Honohan, 1997).
The analyses by Shaw (1973), McKinnon (1973), and Beim and Calomiris (2000) show
that financial repression results in restrained economic growth. With financial repression such asinterest rate controls and high reserve requirement, banks cannot operate efficiently. They have
little incentives to explore new opportunities or to become innovative. With interest rates set
below equilibrium rates, savings are discouraged. With high reserve requirement and direct
credit allocation to certain industries and related parties, the funds available to the private sector
decline. With a limited supply of funds, banks need to ration credit. Well-established clients can
borrow funds at moderate interest rates, while small-scale investors have to obtain funds in
expensive and unstable informal credit markets (Diaz-Alejandro, 1985).
Due to evidence of inefficiency of financial repression, changes in political structures and
economic conditions (Abiad and Mody, 2005), and pressures from their trading partners, many
countries have reformed their financial sectors by moving towards fewer financial restrictions,
particularly during the 1980s and the 1990s. The important objectives of financial reform are to
enhance the efficiency of a banking system, improve the efficient allocation of credits, and
achieve a higher economic growth.
3. Benefits and costs of financial liberalizationA government can liberalize a financial sector by removing domestic and/or international
financial restrictions. Domestic or internal financial liberalization typically entails removal of
interest rate controls, direct credit allocation schemes, or both. They can also allow banks to
expand their activities, and they can privatize state-owned banks to reduce political influences.
International or external financial liberalization refers to the removal of capital account controls
by allowing the flow of funds in and out of the country, possibly including direct investments by
foreigners in the domestic banking sector.
As noted, the benefits of financial liberalization include the stimulation of savings,
improved credit allocation and, indirectly, higher growth when financial resources are channeled
to more productive sectors. Domestic financial liberalization also reduces the financial
constraints on many firms. Leaven (2003) finds that the financing constraints have reduced for
both large and small firms over time. In fact, under financial repression large firms favored
access to external sources of funds, but the difference between big and small firms concerning
cost of borrowing has disappeared. He also finds that financial liberalization reduces
imperfections in asymmetric information costs of firm’s financial leverage.
Turning to international financial liberalization, capital controls discourage foreign directinvestment (FDI) directly in some instances or indirectly, because multinational firms operating
in a country face higher cost of borrowing locally than in international markets (Desai et al,
2006). Hence, the removal of capital control can lead to more FDI, which will bring in new
technology and management skill.
In addition, both external and internal financial liberalizations tend to improve the
financial infrastructure (Schmukler, 2004) and bank governance. The benefits from reducing
foreign bank entry restrictions is that foreign investors will monitor their banks’ activities, and
host governments have to adopt an international accounting standard, thereby increasing
transparency. If foreign banks dominate the banking sector, governments are less likely to bail
them out. A lower likelihood of bail out disciplines banks and reduces moral hazard.
The process of eliminating financial restrictions can also incur costs, particularly during a
transition period, and possibly more permanently if banks’ excessive risk-taking incentives are
strong. Diaz-Alejandro (1985) argued that financial liberalization is destined to cause crisis.
Many recent studies show that almost all banking crises have been associated with financial
liberalization (Kaminsky & Reinhart, 1999, Williamson et al, 1998). Caprio and Klingebiel
(1996) and Eichengreen and Arteta (2002) conclude that banks are more likely to fail in a
liberalized regime than under financial repression. One possible explanation could be that
financial liberalization exposes the risk and poor performance of pre-liberalization portfolios
(Caprio et al, 2000).
There are a number of financial crisis theories that associate free, unsupervised financial
markets with crises. For example, Minsky (1992) explains banking crises referring to three
different kinds of financial transactions-- hedging, speculation and Ponzi schemes. In a Ponzi
scheme banks must borrow or sell assets to repay interest. According to Minsky, speculation and
Ponzi schemes may dominate the markets under some circumstances, unless supervisors prevent
this from happening. Wolfson (2002) later expands Minsky’s theory of financial crises into the
global context. During the Asian crisis, for example, “too much” money flowed into emerging
markets. Large capital flows into the Asian countries fueled expectations of further profits
leading to greater inflows and the expansion of corporate debt collateralized by assets with
ultimately inflated values. These crisis theories describe how crises can arise in uncontrolled and
unsupervised markets but they are not exact in specifying how different aspects of liberalization
and government involvement contribute to, or reduce, the likelihood of banking crises.
There exist a number of arguments why different aspects of domestic liberalization may
increase the likelihood of crises. According to Alba et al (2000), when governments liberalize the
financial sector, initial conditions, the regulatory framework, as well as macroeconomic
conditions may contribute to or reduce the likelihood of banking crises. Financial liberalization
creates new business environments for both bank managers and supervisors. Pre-financial
liberalization, bank managers and staff were used to deal within a controlled environment.
During post-liberalization new types of risk arise and bankers must manage them. If the
managers or supervisors are not familiar with the new environment, they cannot monitor banks’
activities properly. As a result over-lending or excessively risky lending may occur (Noy, 2004).
Another argument for costs of financial liberalization is that the eliminations of interest
rate controls and direct regulation of credit can strengthen price competition, reduce profitability
in the banking sector and thereby increase the fragility of the system. Inefficient banks facing
narrow profit margins would be more vulnerable to fluctuations in economic activity (Noy,
2004). However, for an opposite argument the increasing competition can contribute to greater
efficiency and market discipline on banks’ willingness to hold capital as a buffer.
Liberalization of markets for securities tends to increase the volatility of interest rates in
the market place. There is an argument that interest rate volatility is associated with substantial
costs because it requires more sophisticated risk management and it reduces client’ ability to
service debt (Caprio et al, 20001; Alba et al, 2000). After liberalization, banks could find
themselves lacking skilled people to manage portfolios efficiently, since these skills are not
developed during periods of financial repression.
The arguments why external liberalization contributes to currency and banking crises are
commonly put forward in the debate about the merits and demerits of controls of capital account
1 Caprio et al (2000) argue that with a weak banking system, macroeconomic shocks can lead the increase in interestrates, which then will reduce client’ ability to service debts, and at the same time banks are encouraged to borrowabroad and lend locally.
He observed monthly data for 26 emerging countries and finds that the likelihood of banking
crises increases after financial liberalization.
Eichengreen and Arteta (2002) extend the analysis in Demirgüç-Kunt and Detragiache by
distinguishing between the effects of domestic and international financial liberalization on banking crises. International financial liberalization is captured by a 0/1 dummy, where 1 is
assigned for periods with an absence of capital account controls. They focus on the sample of 75
emerging markets and developing countries during the period of 1975-1997. In banking crisis
regressions they do not find that capital account liberalization contributes to banking crises, but
their finding with regard to domestic financial liberalization is consistent with those in
Demirgüç-Kunt and Detragiache. Furthermore, by interacting domestic financial liberalization
with international financial liberalization they find a positive effect of this interaction term on
banking crises, suggesting that capital account liberalization increases the likelihood of banking
crises for countries that liberalize interest rate controls.
Noy (2004) examines whether banking crises are caused by insufficient supervision as
well as the loss of monopoly power after countries liberalized their financial systems. He uses
proxies similar to those in Eichengreen and Arterta (2002) and finds similar results. Domestic
financial liberalization is highly correlated with the advent of banking crises but international
financial liberalization is not. Interaction between domestic financial liberalization and
supervision, as well as with other institutional variables, suggests that poor supervision and
institutions explain the adverse effects of domestic financial liberalization.
The three mentioned papers explained the likelihood of banking crises by focusing on the
liberalization of interest rate controls as well as capital account. Other aspects of domestic
financial liberalization have also been examined in other studies. For example, in studying the
relationships among financial liberalization, financial crises, and economic growth, Rancier et al
(2006) construct two dummies to proxy financial liberalization. One is from the dates of equity
market liberalization and another is from the patterns of private capital flows. Both financial
liberalization variables are associated with the higher probability of banking and currency crises
(or twin crises). Barth et al (2004) focus on restrictions on bank activity, entry restrictions and
privatization. They find that restrictions on banking activity and foreign bank entry increase the
likelihood of banking crises. Their finding that restrictions on bank activity and competition
increases financial fragility is consistent with the argument that liberalization improves credit
For each country, the following six dimensions capture liberalization of different kinds of
restrictions:
(1) Elimination of credit controls including reserve requirements,
(2) Elimination of interest rate controls,
(3) Elimination of entry barriers and expanding bank activities,
(4) Security market policy,
(5) Elimination of capital account restrictions,
(6) Reduction of state ownership of the banking sector,
The seventh dimension of financial reform does not refer to liberalization per se, but to
capital regulation and prudential supervision:
(7) Enhancement of capital regulations and prudential supervision of the banking sector
(CRS).
Each dimension is measured on a 0 to 3 scale: 0 (fully repressed), 1 (partially repressed),
2 (largely liberalized), and 3 (fully liberalized). For the banking regulation and supervision
dimension, 0 is not regulated and supervised, 1 is less regulated/supervised, 2 is largely
regulated/supervised and 3 is highly regulated/supervised.
In our empirical analyses, we test the impact of these dimensions of financial
liberalization as well as the total financial liberalization on the likelihood of banking crises. The
total Financial Liberalization index (FL) is constructed by summing the first six dimensions of
liberalization; therefore, this aggregate index has a scale of 0 to 18. The banking crisis episodes
data are taken from Caprio et al (2005)2. Descriptive statistics for banking crises and financial
liberalization variables are reported in Table 1.
Since each dimension of liberalization is highly correlated (see Table 2a for correlations),
analyzing the impact of individual type of liberalization on banking crises does not provide
2 Caprio et al (2005) compile the banking crisis data based on published financial sources and interviews withexperts. There are two kinds of banking crises; a systemic banking crisis is defined as the situation when much or allof bank capital is exhausted; a borderline banking crisis is identified when there is evidence of significant banking problems such as government intervention in banks and financial institutions.
significant contribution. We, therefore, use the following rationale to distinguish components of
liberalization into three types.
To enjoy faster growth, many governments have initiated liberalization by allowing
banks to set interest rates or to allocate credit more freely. We argue that these types of liberalization are nearly equivalent since freedom to allocate credit without freedom to set
interest rates on loans implies that the allocation will be determined by the interest rate structure.
Similarly freedom to set interest rates on loans while controlling credit allocation will result in
continued excess demands and supplies for some types of loans. Controls on deposit interest
rates while loan rates and credit allocation are free will also lead to excess demands for credit.
Credit and interest rate controls of all kinds can be viewed as restrictions on banks
actions and behavior. We consider these restrictions the most fundamental ones, meaning that
other types of restrictions matter less if banks cannot freely allocate available funds and set
interest rates. The first two dimensions, eliminations of credit and interest controls, refer to what
we denote “Behavioral Liberalization”. It refers to the score for these two dimensions added
together:
i) Behavioral Liberalization = (1) + (2)
Other types include restrictions in equity markets and on international capital transactions
implying restrictions on the set of available sources and uses of bank funds, as well as lesscompetition from non-bank financial institutions. Restrictions on the range of activities of banks
can be involved in and restrictions on entry have similar effects. The scores for the third through
the fifth dimensions are, therefore, added to capture liberalization of restrictions on banks’
choices of uses and sources of funds, and competition. We denote this aggregate index
“Competitive Liberalization”. The correlations among these three dimensions are as high as
between 0.6 to 0.7 (Table 2a). Thus, it is likely that not much information is lost when using this
aggregate variable instead of (3), (4) and (5) individually.
ii) Competitive Liberalization = (3) + (4) + (5):
We also consider government ownership a third type of restriction. Thus, a third kind of
The correlations between the three types of liberalization are also high raising questions
about the possibility of identifying effects by type. We return to this issue.
To capture the idea that there is a learning period after a financial restriction is relaxed,
and, therefore, that liberalization may have temporary effects, we include the change in types of liberalization over a five-year period as independent variables (ΔFL, or ΔBehavioral
controls may increase the likelihood of banking crises. This is a temporary effect of behavioral
liberalization, and is captured by the change in the behavioral liberalization variable.
The competitive liberalization and privatization may also have a temporary effect in
increasing the likelihood of banking crises. Liberalizing competitive restrictions can increase therisk exposure and the vulnerability of the banking system. Given the liberalization of behavioral
and competitive restrictions, we also expect that privatization may increase the likelihood that a
banking crisis occurs because governments’ willingness to support banks in distress is likely to
decline.
Hypothesis 1.b: Over the longer term we expect that the relaxation of financial
restrictions on banks reduces the incidence of banking crises.
In the long term, risk management skills develop, credit allocation becomes more
efficient, and bad loans should decline. In addition, with the presence of strong regulation and
supervision of banks as well as market discipline, we expect that behavioral and competitive
liberalizations reduce the likelihood of banking crises. In the long run, privatization can affect
banks’ behavior through improved governance. These are the permanent effects, which are
reflected by the level of financial liberalization variables.
An additional consideration is that effects of liberalization could depend strongly on the
effectiveness of supervision, institutional quality, and the extent of deposit guarantees behind banks.
Hypothesis 2: On the role of bank capital regulation and prudential supervision (CRS),
institutional quality, and deposit insurance:
Hypothesis 2.a: Stronger capital regulations and supervisions and (legal and political),
and higher institutional quality in a country are expected to lower the likelihood that
liberalization increases the likelihood of banking crisis.
Hypothesis 2.b: The stronger the protection of bank depositors and other creditors of
banks, the greater is the likelihood that liberalization will result in a banking crisis.
Additional hypotheses with respect to the interactions between, for example, behavioral
and competitive restrictions could be formulated, but the high degree of correlation between
different kinds of liberalization data prevents us from using interactions between types of
liberalization (see Table 2a).
6. Model Specification
The data set on financial reform covers 42 countries for the period 1973-2002. We study
the effects of financial liberalization for this total sample of countries, as well as for a group of
20 emerging market economies. These countries have liberalized to various degrees during the
period. The emerging markets group corresponds to the selection by the Economist3.
The hypotheses with respect to the likelihood of banking crisis are tested using the
following model specification:
,
, , , 1 , , 1 , , 1 ,,
ln ,1
i t
i t j j i t k k i t l l i t i ti t
P
L Z Q XP
where, , 1 , , 1 , , 1
, , ( )
1 prob( 1| , , )
1 j j i t k k i t l i ti t i t Z Q
P BC Z Q Xe
BCi,tis the onset of banking crisis dummy variable, which takes a value of one in the first
year that crisis erupts, and 0 if there is no crisis. The subscript i refers to a country and t indicates
time. Z j is j types of financial liberalization, which are discussed in the previous section.
X is an l-element vector of a standard set of macroeconomic control variables used in thereviewed literature. These variables are lagged one period relative to the onset of banking crisis
dummy. They include the real GDP growth rate, the ratio of current account to GDP, the ratio of
money supply to international reserves, the growth rate of the ratio of domestic credit provided
by banking sector to GDP, the inflation rate, the Northern interest rate, and OECD income4.
Following Arteta and Eichengreen (2002), the Northern interest rate is the weighted average of
the interest rates in Germany, USA, UK, Switzerland, France and Japan. The weights are the
3 Emerging market countries are Argentina, Brazil, Chile, Colombia, Egypt, Hong Kong, India, Indonesia, Israel,South Korea, Malaysia, Mexico, Pakistan, Peru, the Philippines, Singapore, South Africa, Thailand, Turkey,Venezuela. We exclude Taiwan due to the missing of many macroeconomic variables. Other countries in the sampleinclude Australia, Canada, France, Germany, Italy, Japan, New Zealand, US, UK, Bangladesh, Bolivia, Costa Rica,Ecuador, Ghana, Guatemala, Jamaica, Morocco, Nepal, Sri Lanka, Uruguay, and Zimbabwe.4 Since we control for the world interest rate, we do not include the real domestic interest rate in regressions. This isalso because the domestic interest rate correlates with the liberalization of interest rate control variables and has nomeaningful effect during the pre-liberalization when interest rates were controlled.
fractions of debt denominated in the relevant currencies. The OECD income is the GDP of the
OECD member-nations.
Another set of variables controls for institutional quality and differences in deposit
insurance systems (Qk
). We use the real GDP per capita to capture general institutional quality.This variable is highly correlated with variables like corruption and rule of law but it captures the
quality of the legal and political systems in a country more generally.5 A proxy for the coverage
of explicit deposit insurance captures the difference in deposit insurance systems. It measures the
maximum deposit insurance coverage in a country relative to the value of the average (per
capita) deposit. To test whether the effects of financial liberalization depend on institutional
variables, we interact financial liberalization variables with each of the Qk variables (that is, Z j ×
Qk ). The descriptions and sources of the variables used in regressions are reported in Data
Appendix.
To reduce the effects of any source of simultaneity bias, we lag all independent variables
one year and exclude crisis years following the onset of a crisis within the same crisis episode.
To reduce the problem of correlated error terms across countries and over time in the panel
regressions, we use robust and clustering standard errors of estimates by country to correct the
covariance matrix of the estimates for heteroskedasticity and autocorrelation among the
observations across time within each country.
In order to achieve the objective of investigating whether there is information in the data
for types of restrictions relative to an aggregate financial liberalization index, it would be
desirable to use both levels and changes in behavioral liberalization, competitive liberalization,
and privatization, as well as interactions between these variables and supervision (CRS), deposit
insurance coverage and GDP/capita. The number of variables becomes very high this way,
however, and, more seriously; behavioral liberalization, competitive liberalization, privatization
and CRS are very highly correlated as shown in Table 2a. Thus, independent effects of all the
dimensions of financial reform may not be identified.
To alleviate the problems of number of data points and correlation among types of
liberalization we follow the following procedure:
5 The correlations between the GDP/Capita and the rule of law or corruptions are 0.6-0.7.
The first set of regressions in Table 3 include only the total financial liberalization index
in level (FL) and change over the preceding five years (FL) along with the index for capital
regulation and prudential supervision (CRS). In table 4, interactions between FL and CRS,
deposit insurance coverage and institutional quality6 (GDP/Capita) are introduced. Thereafter,
we divide FL into the three types (behavioral liberalization, competitive liberalization, and
privatization) in level, change, and in interactions but in steps. Thus, in Table 5 we divide the
level of FL into the three types while the change and interaction variables remain aggregated. In
Table 6, the change in liberalization is divided into three types while the level and interactions
remain aggregated. Finally, in Table 7, the interaction terms are divided into interactions with the
changes in three types of liberalization, while the aggregate liberalization index in the level and
the change is used. Interactions between the levels of types of liberalization and institutional
variables do not appear meaningful considering the high correlations among the types in levels,
as well as between the levels and institutional variables, in particular CRS. This step procedure
allows us to study the contributions of the types of liberalization while saving on degrees of
freedom and reducing potential problems of multicollinearity. All regressions include the same
macro variables as well as the levels of CRS, deposit insurance coverage and GDP/capita.
7. Empirical results
Table 3 reports the empirical results for the effects of the aggregate index for financial
liberalization in the level (FL) and preceding change (FL) on the probability of banking crises
for all the 42 countries and for 20 emerging markets. In this regression without interactions the
period is also divided into two sub-periods, 1973-1989 and 1990-2002. This division may strain
the data set, since the number of banking crises in emerging markets during the second sub-
period is only 17.
Focusing on the full period and all countries (column 1), the coefficient for the FL is
positive and strongly significant, suggesting that financial liberalization increases the likelihood
of banking crises. The FL seems to be associated with an increase in the likelihood of banking
crisis as well, but the significance level is low (75 percent probability of being different from
6 We experimented with a Corruption index and Rule of Law index as indicators of institutional quality. They are both highly correlated with GDP/capita and do not seem to contribute additional information.
zero). The coefficient for regulation and supervision (CRS) is negative and strongly significant. 7
The coefficient for deposit insurance coverage is positive as well on a weak level of significance.
In column 2, the results for emerging markets are the same but at a lower level of significance
with respect to FL and CRS, while FL and deposit insurance coverage are insignificant.
Dividing into sub-periods (columns 3-6), the results are very different in the two periods.
While results with respective to FL and FL for all countries for the first sub period remain
similar to the results for the whole period, the results for emerging markets and the second sub-
period are very different. Furthermore, CRS is significant only in the second sub period. The
only variable that remains robust in terms of sign (in additional to macroeconomic control
variables) is deposit insurance coverage. It has a positive sign in all cases, but with a low level of
significance.
If the results for the sub periods can be considered valid at all, they indicate that the level
of financial liberalization had a positive impact on banking crises during the first sub-period,
while regulation and supervision were effective at preventing crises during the second sub
period. However, this conclusion must be hedged given the relatively few crisis episodes in the
second sub-period in particular. The results for GDP per capita are also surprising in the second
sub-period, wherein the coefficients are positive and significant for all as well as emerging
markets.
In Table 4 interaction terms between FL on the one hand, and CRS, deposit insurancecoverage and GDP/capita on the other, are included. The impact of FL appears entirely through
the interactions in this case, while the result for FL remains unchanged. The impact of CRS
also comes through the interaction term while the level alone actually switches sign and becomes
positive. Thus we can conclude that interactions are important.
The interaction terms in Table 4 tell us that the impact of liberalization on banking crises
depend on both CRS and the level of domestic institutions (proxied by GDP/Capita). The impact
of financial liberalization on the likelihood banking crises is reduced if there is effective capital
regulation and prudential supervision (CRS). The influence of CRS in reducing the impact of
liberalization becomes stronger if GDP/Capita is high according to the significant coefficients of
7The estimated coefficients in logit regressions do not indicate the magnitude of increase or decrease in the
probability of banking crisis. However, the sign of the coefficient indicates the direction of the effect.
temporary impact of privatization (reduce the state ownership of banks) is a reduction in the
likelihood of banking crises.
8. Conclusion
The evidence presented in Tables 3-7 does not lead to clear-cut and robust conclusions of
the effects of financial liberalization on banking crises. Results are sensitive to country group,
period, and the inclusion of interaction terms. This sensitivity depends to some extent on few
observations on banking crises in sub-periods, and to some extent on high correlation among
types of financial sector reform. Effects of reforms with respect to restrictions on financial
market and banking activities can be separated from effects of reforms of regulation and
supervision, however.
Conclusions which can be drawn with a modest level of confidence are that financialliberalization seems to be associated with the increased probability of banking crises, but this
effect can be counteracted with strengthened capital regulation and supervision. When
distinguishing the index of total financial liberalization into different types; however, we do not
find significant effects for the individual types (behavioral liberalization, competitive
liberalization, and privatization) on the likelihood of banking crises. In other words,
distinguishing among types of liberalization does not add much information relative to using an
aggregate index for financial liberalization, or there is no particular type of financial
liberalization that lead to a higher likelihood of banking crises than others.
We also investigate whether the likelihood of banking crises will be temporarily higher
after the relaxation of financial restrictions due to learning period for risk assessment and
management skills to be developed. We find that it is more meaningful to distinguish between
different types of liberalization when studying the temporary effects (captured by changes in
liberalization) and taking into account the impact of deposit insurance coverage. Although their
significance levels are not high, the results show that at a high level of deposit insurance
coverage the change in behavioral liberalization increases the likelihood of crises at a greater
extent, whereas the change in privatization lowers the likelihood of crises. A higher competition
and lower likelihood of bank bail out as a result of privatization could discipline banks and lower
The onset of banking crisis dummy, which is equal to 1 in a first year of
each banking crisis episode (both systemic and non-systemic bankingcrises), and 0 otherwise
Caprio and Klingebiel
(2005)
Real GDP Per Capita The log of real GDP per capita (constant 2000 US$). WDI
Real GDP GrowthRate
Real GDP growth (annual %) WDI
CA to GDP Current account balance (% of GDP) WDI
Growth of DomesticCredit
The log difference of the ratio of domestic credit provided by bankingsector (% of GDP)
WDI
M2 to Reserve The ratio of money and quasi money (M2) to gross international reserves WDI & IFS
Inflation The log difference of GDP deflator WDI
Northern Interest Rate The weighted average of the interest rate in Germany, USA, UK,Switzerland, France, and Japan. The weights are the fraction of debtdenominated in the relevant currencies
% of ∆ OECD income The growth rate of the average GDP of the OECD (high income) nations Authors’ calculation
Behavioral Lib. The aggregate index of the variables of elimination of credit controls(include reserve requirements) and interest rate controls. The scale is 0-6.
Authors’ calculation based on Omori (2006)
Competitive Lib. An aggregate index of equity market lib., capital account lib. and entry &activity lib. The scale is 0-9.
Authors’ calculation based on Omori (2006)
Privatization Privatization of state ownership of the banking sector. The scale is 0-3. Omori (2006)
CRS (CapitalRegulation &Supervision)
Enhancement of prudential regulations and supervision of the bankingsector Omori (2006)
FL Behavioral Lib. + Competitive Lib. + Privatization. The scale is 0 – 18. Omori (2006)
∆ Behavioral Lib. Change of Behavioral from the previous five years Authors’ calculation based on Omori (2006)
∆ Competitive Lib. Change of Competitive Lib. from the previous five years Authors’ calculation based on Omori (2006)
∆ Privatization Change of Privatizationfrom the previous five years Authors’ calculation based on Omori (2006)
∆ FL Change of Total Financial Lib. from the previous five years Authors’ calculation based on Omori (2006)
Deposit InsuranceCoverage The interval data of the ratio of deposit insurance coverage per deposits per capita. This index ranges from 1 to 10 (the variable is constructed based on the data from Demirgüç-Kunt et al, 2005)
Prob > Chi-Square 0.000 0.000 0.000 0.000 0.000 0.000The dependent variable is the onset of banking crisis dummy. All independent variables are lagged by one year. “∆” refers to a change over five years. Regressions are estimated using the logitmodel, and t-statistics are calculated using robust and clustered standard errors within a country. *, **, *** indicate the significance levels of 10%, 5%, and 1% respectively. # indicates thecoefficient value zero that falls outside one standard deviation of the estimate. The numbers in parentheses are p-values.
The dependent variable is the onset of banking crisis dummy. All independent variables are lagged by one year.“∆” refers to a change over five years. Regressions are estimated using the logit model, and t-statistics arecalculated using robust and clustered standard errors within a country. *, **, *** indicate the significance levelsof 10%, 5%, and 1% respectively. # indicates the coefficient value zero that falls outside one standard deviation of the estimate. The numbers in parentheses are p-values.
Table 5: The Effect of Different Types of Financial Liberalization onBanking Crises, 1973-2002
(1) (2)
All Emg
∆ in FL 0.0550 # -0.0378(0.2796) (0.6542)
Behavioral Liberalization -0.0999 0.1483
(0.5736) (0.6257)
Competitive Liberalization -0.0905 0.2332
(0.6279) (0.4017)
Privatization -0.2634 # -0.2325
(0.1423) (0.4632)
FL CRS -0.1362 # -0.0861
(0.1169) (0.4352)
FL Deposit Insurance -0.0071 0.0093 #
(0.4157) (0.2805)
FL GDP per Capita 0.0338 * -0.0056(0.0579) (0.8510)
CRS 1.2176 0.1193
(0.3516) (0.9438)
Deposit Insurance 0.1186 # -0.0434
(0.2281) (0.6315)
Real GDP per Capita -0.4056 * 0.1432
(0.0562) (0.7037)
Real Growth Rate of GDP 0.0207 0.0315
(0.6217) (0.6199)
CA (% of GDP) -0.0977 ** -0.0740 #
(0.0139) (0.2257)
M2/Reserves 0.0365 ** 0.0572 #(0.0409) (0.1057)
Growth Rate of Domestic Credit 0.5538 # 0.8325 #
(0.3115) (0.2406)
Inflation 0.8577 *** 0.4608 #
(0.0003) (0.2678)
Northern Interest Rate 0.0713 # 0.1544 *
(0.3103) (0.0771)
% change OECD Income -0.2425 * -0.2770 #
(0.0958) (0.2551)
Constant -1.5881 # -5.2303 *
(0.2958) (0.0547)
No of Obs 746 347
Wald Chi-Square 75.764 219.894
Prob > Chi-Square 0.000 0.000
The dependent variable is the onset of banking crisis dummy. All independent variables are lagged by one year.“∆” refers to a change over five years. Regressions are estimated using the logit model, and t-statistics arecalculated using robust and clustered standard errors within a country. *, **, *** indicate the significance levelsof 10%, 5%, and 1% respectively. # indicates the coefficient value zero that falls outside one standard deviation of the estimate. The numbers in parentheses are p-values.
Table 6: Change in Types of Financial Liberalization, 1973-2002
(1) (2)
All Emg
FL -0.1317 0.1487
(0.3664) (0.6021)Δ Behavioral Lib. 0.0444 0.0775
(0.7033) (0.6956)
Δ Competitive Lib. 0.1099 # -0.0762
(0.2837) (0.6043)
Δ Privatization -0.1085 -0.3002
(0.6505) (0.3743)
FL CRS -0.1394 * -0.0575
(0.0879) (0.6098)
FL Deposit Insurance -0.0071 0.0085
(0.4195) (0.3942)
FL GDP per Capita 0.0345 * -0.0101
(0.0613) (0.7677)CRS 1.3176 # -0.1719
(0.2778) (0.9178)
Deposit Insurance 0.1244 # -0.0056
(0.2233) (0.9591)
Real GDP per Capita -0.4051 * 0.1638
(0.0656) (0.6690)
Real Growth Rate of GDP 0.0187 0.0403
(0.6582) (0.5386)
CA (% of GDP) -0.1008 *** -0.0687 #
(0.0096) (0.2312)
M2/Reserves 0.0319 * 0.0477 #
(0.0895) (0.1360)Growth Rate of Domestic Credit 0.5505 # 0.7972 #
(0.3076) (0.2276)
Inflation 0.8740 *** 0.4728 #
(0.0002) (0.2449)
Northern Interest Rate 0.0555 0.0759
(0.4103) (0.3823)
% change OECD Income -0.2396 * -0.3002 #
(0.0991) (0.1962)
Constant -1.4270 -4.6548 *
(0.3432) (0.0716)
No of Obs 746.0000 347.0000
Wald Chi-Square 85.9500 321.7110
Prob > Chi-Square 0.0000 0.0000
The dependent variable is the onset of banking crisis dummy. All independent variables are lagged by one year.“∆” refers to a change over five years. Regressions are estimated using the logit model, and t-statistics arecalculated using robust and clustered standard errors within a country. *, **, *** indicate the significance levelsof 10%, 5%, and 1% respectively. # indicates the coefficient value zero that falls outside one standard deviation of the estimate. The numbers in parentheses are p-values.
(0.4905) (0.9319)Real Growth Rate of GDP 0.0175 0.0068
(0.6834) (0.9234)
CA (% of GDP) -0.0904 ** -0.0944 #
(0.0167) (0.1518)
M2/Reserves 0.0206 # 0.0492 #
(0.2785) (0.1480)
Growth Rate of Domestic Credit 0.6152 # 0.7809 #
(0.2412) (0.2143)
Inflation 0.7708 *** 0.5679 #
(0.0001) (0.1188)
Northern Interest Rate 0.0578 0.0792
(0.4056) (0.4297)% change OECD Income -0.2332 # -0.3001 #
(0.1159) (0.1409)
Constant -2.9859 * -3.2603 #
(0.0650) (0.1497)
No of Obs 746 347Wald Chi-Square 103.967 43.426
Prob > Chi-Square 0.000 0.003The dependent variable is the onset of banking crisis dummy. All independent variables are lagged by one year. ∆FL refers to a change over five
years. Regressions are estimated using the logit model, and t-statistics are calculated using robust and clustered standard errors within a country.