Market Discipline and Financial Safety Net Design

Asli Demirgüç-Kunt and Harry Huizinga*

First Draft: July 1999Revised: April 2000

Abstract: An important question is whether the financial safety net reduces marketdiscipline on bank risk taking. For countries with varying deposit insurance schemes, wefind that deposit rates continue to reflect bank riskiness. Cross-country evidence suggeststhat explicit deposit insurance reduces required deposit interest rates at a cost of reducedmarket discipline. Internationally, deposit insurance schemes vary widely in theircoverage, funding, and management. Hence, there are widely differing views on howdeposit insurance should optimally be structured. To inform this debate, we use a newlyconstructed data set of deposit insurance design features to examine how different designfeatures affect deposit interest rates and market discipline.

Keywords: market discipline, deposit insuranceJEL Classification: E43, G28

* Development Research Group, The World Bank, and Department of Economics, Tilburg University, andCEPR, respectively. The findings, interpretations, and conclusions expressed in this paper are entirelythose of the authors. They do not necessarily represent the views of the World Bank, its ExecutiveDirectors, or the countries they represent. We thank the Editor, an anonymous referee, Thorsten Beck,Jerry Caprio, Stijn Claessens, Patrick Honohan, Edward Kane, Tae Soo Kang, and Maria SoledadMartinez Peria for comments on an earlier draft, and Anqing Shi and Tolga Sobaci for excellent researchassistance.

Market Discipline and Financial Safety Net Design

Abstract

An important question is whether the financial safety net reduces market discipline onbank risk taking. For countries with varying deposit insurance schemes, we find thatdeposit rates continue to reflect bank riskiness. Cross-country evidence suggests thatexplicit deposit insurance reduces required deposit interest rates at a cost of reducedmarket discipline. Internationally, deposit insurance schemes vary widely in theircoverage, funding, and management. Hence, there are widely differing views on howdeposit insurance should optimally be structured. To inform this debate, we use a newlyconstructed data set of deposit insurance design features to examine how different designfeatures affect deposit interest rates and market discipline.

I. Introduction

In the last two decades, we have seen a series of banking crises around the world

where banks have become systematically insolvent. Banking crises have occurred in

developed and developing countries alike. Prominently, the Asian crisis of 1997

involved banking crises in Thailand, Indonesia, Malaysia and Korea, with banks

becoming insolvent after economic downturns and currency devaluations.

Systemic bank insolvencies involve huge costs to the banks themselves, their

customers and to governments. Bank failures may lead to the destruction of a bank’s

information capital garnered in previously nurtured bank-customer relationships. A

disruption of bank lending and of the payments system may also cause a reduction in

investment and other economic activity. Further, bank depositors potentially lose heavily

because of bank failures. Last but not least, governments tend to incur large costs in

remedying a banking crisis. To make financial system breakdowns less likely and to

limit their costs if they occur, all countries of the world have financial safety nets in

place. These nets are amalgams of policies including explicit or implicit deposit

insurance, the central bank’s lending of last resort, bank insolvency resolution

procedures, and bank regulation and supervision.

Bank safety nets are difficult to design and administer, because they have the

conflicting objectives of protecting bank customers and reducing banks’ incentives to

engage in risky activities. In several countries including the U.S., the financial safety net,

structured to reduce the vulnerability of the financial system, appears to have had quite

the opposite result. Indeed, Kane (1989) identifies the U.S. financial safety net, and

notably fixed-rate deposit insurance and belated bank closures, as the single most

important factor in explaining the catastrophic Savings and Loan crisis of the 1980s.

Similarly, Demirgüç-Kunt and Detragiache (1998) find international evidence that the

existence of an explicit deposit insurance scheme has contributed to banking system

fragility.

To restrain bank risk taking, financial safety nets generally rely on two

mechanisms: (i) market discipline, and (ii) bank regulation. Bank creditors can exert

market discipline by withdrawing their funds, or demanding higher interest rates from

riskier banks. In case of publicly traded banks, equity holders can also effect discipline.

Bank regulations, in turn, can directly restrict a bank’s operations, and prescribe

corrective action if bank solvency is jeopardized. The challenge facing policy makers is

to ensure that the financial safety net enables rather than undermines market discipline.

There is a real danger that regulatory forbearance policies and overly generous depositor

protection increase rather than reduce the excessive bank risk taking which has been the

root cause of many bank failures.

A substantial literature discusses the potential effects of safety net design and

implementation on market discipline. This literature proposes various design features

such as limited insurance coverage, co-insurance, and private deposit insurance that leave

some room for market discipline in an explicit (public) scheme of deposit insurance. See

Kane (1999) for a general discussion, and Ely (1986), Calomiris (1997) and Wall (1998)

for specific proposals concerning private deposit insurance and (uninsured) subordinated

bank debt. For lack of empirical evidence, this debate about deposit insurance design has

been entirely theoretical and hypothetical. Hence, we do not know whether deposit

insurance design features that work well in theory work equally well in practice. The

main purpose of this paper is to fill this gap in our knowledge.

First, we investigate empirically to what extent market discipline exists around the

world. There is a considerable body of literature on market discipline mostly for the U.S.

that generally finds support for the existence of some market discipline by bank creditors.

We contribute to this literature by extending the analysis to a large number of developed

and developing countries. To do this, we investigate the sensitivity of (i) bank deposit

interest rates, and (ii) deposit growth rates to bank risk. Specifically, we relate a bank’s

implicit cost of funds (measured as interest expenses divided by interest-paying debt) in

excess of the riskless interest rate to bank liquidity. This is done separately for 31

countries over the years 1990-1997. In addition, for a larger set of 43 countries we

examine whether market discipline entails that riskier banks are less able to attract

deposits. To this aim, we relate the measured growth rate of (real) deposits to bank

liquidity.

Secondly and more importantly, we examine whether differences in market

discipline across countries can be explained by different design features of financial

safety nets. Evidence of this kind should be useful to policy makers around the world, as

they grapple with the question of how to design a financial safety net without

undermining market discipline. To enable this work, we have collected detailed

information on the nature of deposit insurance for over 50 countries. This data shows that

there is considerable cross-country variation in key design features such as insurance

coverage, co-insurance, source of funding and fund management. On a cross-country

basis, we relate the extent of market discipline (in terms of bank deposit interest rates and

deposit growth) to whether there exists an explicit deposit insurance scheme, and, if so,

how it is constituted.

The remainder of this paper is organized as follows. Section 2 discusses the

deposit insurance information, and other data used in this study. Section 3 presents the

empirical evidence on market discipline regarding deposit interest rates and deposit

growth in individual countries. Section 4 examines how the existence of an explicit

deposit insurance scheme affects market discipline, and particularly how it affects the

relationship between bank interest rates and bank liquidity, paying attention to the

potential endogeneity of bank liquidity. Section 5 examines how specific deposit

insurance design features are valued by depositors, as reflected in required deposit

interest rates, and how they affect market discipline. Section 6 concludes.

II. The data

This study combines cross-country institutional data on deposit insurance design

and bank-level data on interest expenses, deposit growth and other derived variables.

Information on deposit insurance schemes in individual countries as of 1997 is

represented in Table 1. As reported in the table, we have put together this information

from a variety of sources. As also seen in the table, many countries have moved to

explicit deposit insurance schemes during the 1980s and 1990s, although there remain

ample countries that do not have explicit deposit insurance. In the absence of explicit

deposit insurance, there is implicit deposit insurance by national governments.1

1 In the U.S., for instance, deposit insurance has frequently been extended to uninsured credits, but less so

since the passage of the of FDIC Improvement Act of 1991 (see Benston and Kaufman, 1998).

As mentioned, a major cost of introducing explicit deposit insurance is that it

potentially undercuts market discipline. In principle, this effect could be dampened by the

adoption of features that enhance market discipline. The common element of such

features is to ensure that there will be a credible first layer of private loss in case of bank

failure. Such losses provide private parties with incentives to continue to monitor banks

and to remain vigilant.

One market discipline-inducing feature is co-insurance. Co-insurance here means

that depositors are contractually required to share in their bank’s losses (up to a

maximum percentage of deposits) regardless of deposit size. Relatively few countries,

Chile, Colombia, Poland and the United Kingdom among them, have co-insurance.

Another way to induce some depositor vigilance is to ensure that insurance

coverage is truly finite. Indeed, most countries in our sample specify an upper limit to

officially protected deposits, with the exceptions of Mexico and Japan where de jure

covered amounts are unlimited. One way to limit insurance coverage is to exclude

interbank and foreign currency deposits. This is common in most countries (see table).

An important way to engage private parties in the deposit insurance scheme is to

make them underwrite and manage some or all of the insurance. Table 1 reveals that there

is extensive variation across countries in terms of the source of funding of deposit

schemes, and in terms of fund management. Most explicit deposit insurance schemes

establish insurance funds, and most of the payments into the fund come from banks, or

jointly from banks and the government. There is also wide dispersion in the deposit

insurance premium that banks have to pay into the fund. Making deposit insurance

premiums risk-based is a sensible way to try to reduce bank risk taking. Yet, so far in the

sample only Finland, Peru, Sweden and the U.S. have adopted risk-based insurance

premiums. At least with hindsight, there appears to be no relationship between a

scheme’s deposit insurance premium and banking system risk. In the U.S. where risk-

adjusted rates range from 0 to 27 basis points, for instance, more than 90 percent of the

banks qualify for the lowest rate of zero. Finally, regarding fund membership, most

countries have compulsory bank membership of the insurance fund.

Our bank-level data are derived from bank balance sheets and income statements,

as available from the BankScope data base compiled by Fitch IBCA. The data set covers

all OECD countries, as well as many developing countries. For a list of countries

included in our work, see Table 2.2 Bank coverage is comprehensive for most countries,

with covered banks roughly accounting for 90 percent of all bank assets nationwide. The

sample covers the period 1990-1997, and includes about 2500 individual banks.

Table 2 also provides mean values for all bank-level variables used in the

empirical work. Interest Expense is the annual interest expense divided by bank debt,

excluding non-interest bearing debt if any. Interest Expense is thus an implicit interest

rate on all de jure insured and uninsured bank liabilities. In most countries, bank

liabilities that are not covered by explicit deposit insurance may enjoy de facto, implicit

insurance. Hence, the dividing line between covered and uncovered liabilities is difficult

to draw in practice. In the later empirical work, we subtract the government interest rate

from individual bank interest rates to obtain a cross-sectional measure of bank risk that is

adjusted for the nominal riskless rate.

2 Below we find that because of insufficient data we can not estimate regressions for all individual countries in Table II.

Deposit Growth is the growth rate of a bank’s customer and short term funding,

after dividing by the GDP deflator. On average, banks in Costa Rica, Nigeria, and the

United Kingdom were contracting during the sample period, while Ghana and Peru had

rather high annual real deposit growth rate of around 30 percent. Liquidity is the ratio of

liquid assets to total assets. A higher value of Liquidity presumably indicates lower bank

riskiness for depositors. Admittedly, this is not the best conceivable indicator of bank

riskiness. However, it is a bank level variable available to us for a large international

cross-section of banks.3

Overhead is defined as non-interest bank expenses divided by assets, reflecting

bank variation in employment as well as wage levels. Differences in overhead may

equally capture differences in banks’ product mixes and the quality of service. Despite

high wage levels, Overhead is lowest at around 1 percent for high-income countries, such

as Japan and Luxembourg. It is notably high at 3.6 percent for the United States, perhaps

reflecting the proliferation of banks and bank branches due to historical banking

restrictions.

Finally, Short-Term Debt/Total Debt is a bank’s customer and short term funding

divided by total interest–bearing debt. This variable will be a determinant of Interest

Expense, if customer deposits receive different interest rates from, say, marketable notes

and debentures. For most countries, Short-Term Debt/Total Debt is close to unity,

reflecting the importance of short term debt in total debt.

3 In an earlier version of this paper (Demirgüç-Kunt and Huizinga, 1999) we have also used bank

capitalization and profitability as additional indicators of risk. Although results we obtained using theseindicators were broadly consistent with our results here, they tended to be less robust. This may be

III. International evidence on market discipline

A. Do riskier banks pay higher interest rates?

In this section, we present empirical evidence on whether there is market

discipline on banks for a range of individual countries. In principle, depositors can

discipline banks that engage in excessive risk-taking by demanding higher interest rates

or by withdrawing their deposits. We start with the first issue. Several researchers have

investigated whether the cost of debt finance for banks reflects their apparent default risk

for the case of the United States (see Flannery (1998) for a recent survey). Researchers

(Baer and Brewer 1986, Hannan and Hanweck 1988, and Brewer and Mondschean 1994)

typically find that rates on large, partially uninsured CDs reflect bank riskiness. For the

1983-1991 period, Flannery and Sorescu (1996) similarly find that spreads on uninsured

bank debentures (relative to a constructed callable treasury bond) significantly reflect

bank risk in the years 1989-1891, a period when doubts arose whether federal regulators

would fully bail out debt holders of bank holding companies. Using similar data only for

1983-1984, Avery, Belton and Goldberg (1988) and Gorton and Santomero (1990) had

failed to find any such evidence. Relating time series of CD rates to bank-specific news

as reflected in stock prices, Ellis and Flannery (1992) similarly find evidence that bank

CD rates indeed reflect bank-specific risk. Finally, Cook and Spellman (1994) find that

risk premiums on fully insured deposits at Savings and Loans equally reflect bank risk

factors in 1987, since the FSLIC guarantor of these deposits was technically insolvent at

that time, causing doubts about the quality of its guarantees.

because bank capital and profitability are subject to accounting manipulation and tend to be overstated atweak banks whereas liquidity is measured better in the banks’ accounts.

In summary, the U.S. experience shows that there is evidence of market discipline

on banks by insured and uninsured debt holders alike. Specifically, insured deposit rates

may reflect bank risk factors, if doubts arise about the credibility of the insurance

coverage. The U.S. experience thus shows that the distinction between explicit and

implicit deposit insurance becomes blurred, if there is a perceived risk of repudiation of

deposit insurance obligations.

This subsection extends the evidence on market discipline through liability

interest rates to a total of 31 countries. Specifically, for each country we estimate the

following equation:

Interest i,t = α + X i,t-1β + εi,t (1)

where Interest i,t is the ratio of interest expense to interest-bearing debt minus the

government interest rate for bank i. The government rate is the Tbill rate where

available, otherwise the discount rate. X i,t-1 is a vector of lagged bank variables, and εi,t

is a stochastic term. The vector X i,t-1 includes Liquidity as an indicator of bank risk.4 In

addition, Overhead and Short Term Debt/Total Debt are included as controls. As higher

values of Liquidity represents less risk, we expect it to enter the interest regressions

negatively if there is market discipline. Overhead would enter the regression negatively if

banks with high overhead offer high quality services, and therefore can attract deposits at

lower rates. Alternatively, Overhead may simply reflect bank inefficiency, and cause

depositors to demand higher rates to compensate for this inefficiency. The Short Term

4 Results including bank capitalization and profitability are given in Demirgüç-Kunt and Huizinga (1999).

Debt/ Total Debt ratio can also in principle have either sign, depending in part on the

yield curve. Throughout, the right hand side variables are lagged one period.

A negative relationship between Interest and Liquidity in a regression such as (1)

may reflect market discipline, but there are several other potential explanations. Bank

interest expenses are negatively related to liquidity if a generally higher level of interest

rates causes banks to economize on liquid assets. This potential problem is mitigated,

however, as Interest is constructed as the difference between the bank’s implicit cost of

funds and the government rate. Banks may further endogenously increase deposits

(relative to other bank liabilities), thereby lowering their interest expenses and increasing

liquidity to the extent that liquidity reflects required reserves. On the other hand, a

negative relationship between Interest and Liquidity may fail to arise, if banks

experiencing problems increase liquidity to avert market discipline. These problems of

the potential endogeneity of Liquidity are addressed by instrumenting for this variable in

section 4. There we also take into account that market discipline can be exerted through

either price or quantity rationing of banks. This is done by modeling deposit interest rates

and deposit growth simultaneously as alternative means of disciplining banks. The simple

specification as estimated in this section should clearly be seen as no more than an initial

exploration of the correlation between interest and liquidity variables in individual

countries.

For each country, we first estimate (1) using Ordinary Least Squares (OLS).

Second, we estimate the equation using the means of all the bank variables. This provides

us with between estimates indicating how the dependent variable changes, if we compare

the independent variables across different banks. Third, we estimate the equation

controlling for bank means (using the deviation from bank means for each variable). This

provides us with within estimates telling us how the dependent variable changes with the

independent variables, as these independent variables change for the same bank over

time.

The OLS, between, and within estimates of (1) for each country separately are

presented in Table 3. In each case, we only report the coefficients for Liquidity.

Coefficients that are significant at 5 percent or less based on White’s (1980)

heteroskedasticity-consistent errors are bold-faced. The coefficient for Liquidity is

predominantly negative, which we take as evidence of market discipline. In the last row

of the table we also report results using data for all countries. In all three specifications,

Liquidity develops a negative and significant coefficient. Taking significant negative

coefficient on Liquidity to be evidence of market discipline, we see that many countries

have some market discipline, among them the United Kingdom and the United States. In

the sample, Finland, Nigeria, Pakistan are among the countries that have suffered banking

crises where depositors have been made whole. Not surprisingly, these countries are

among the countries where we fail to find evidence of market discipline. An absence of

market discipline can, of course, exist for various reasons. Depositors may simply fail to

monitor banks closely; alternatively, bank regulators are relied upon to close insolvent

banks promptly before depositor losses occur; or, as indicated, the safety net may be so

complete as to eliminate all risk for depositors. Differences in institutional factors such as

quality of regulations or legal standards for enforcing contracts may also explain differing

extents of market discipline. Regressions for individual countries do not allow us to infer

the underlying causes of any absence or presence of market discipline.

B. Do riskier banks attract fewer deposits?

This section considers whether deposit growth of banks varies with their apparent

default risk. A small literature has considered this issue for the U.S. and elsewhere.

Gordon and Pennacchi (1990), for instance, discuss why discipline in the market for bank

deposits will often occur through quantity adjustment. Focusing on New York city banks

in the 1920s and 1930s, Calomiris and Wilson (1998) provide further evidence that

depositors discriminate across banks on the basis of risk and apply market discipline

selectively to reward prudence. Similarly, Kane (1987) finds that depositors withdrew

funds only from those Ohio institutions that were covered by the Ohio insurance fund

when the latter was in crisis. Park (1995) and Park and Peristiani (1998) in turn observe

that riskier U.S. thrifts experienced smaller deposit growth during the 1980s. More

specifically, these authors find that deposit growth is negatively related to the estimated

probability of thrift default. Considering banks in Argentina, Chile and Mexico, Martinez

Peria and Schmukler (1998), finally, find that deposits are negatively related to lagged

bank risk factors derived from accounting data.

In this section, we provide evidence on the relationship between deposit growth

and bank liquidity for 43 countries. We estimate an equation analogous to (1), where the

dependent variable is real Deposit Growth. Real deposits are calculated as nominal

deposits deflated by the GDP deflator. The explanatory variables again include lagged

bank Liquidity.5 A positive coefficient on Liquidity indicates market discipline, as then

deposit growth is relatively higher at safer banks. As controls, we include Government

rate, Overhead, and AssetSize, defined as the log transformation of total bank assets to

GDP. Again, banks with high Overhead may grow relatively slowly or quickly, while the

Size variable introduces a possible scale effect on deposit growth. Scale effects may be

important if a policy of too-large-to-fail stimulates the growth of deposits at large banks

regardless of their measured risk factors.

As for the interest regressions, we leave endogeneity issues related to liquidity

and the simultaneous estimation of interest and deposit growth to Section 4. Again, the

deposit growth regressions of this section should be interpreted as partial correlations

between deposit growth and liquidity that allow us to explore whether there is any

evidence of market discipline.

As before, we obtain OLS, between, and within estimates for the deposit growth

equation for each country. The results are given in Table 4.

As seen in Table 4, the fit of the deposit growth regressions, as proxied by

adjusted R-squares, is quite poor. The low explanatory power and generally inconclusive

results may reflect that actual deposit growth rates result from several opposing forces.

On the one hand, insolvent or near-insolvent banks may wish to pursue a risky growth

strategy on the off-chance of overcoming their difficulties. Riskier banks may succeed in

attracting additional deposits if they pay interest rates that are high enough. Such a

growth strategy would maximally exploit the bank safety net.6 Alternatively, risky banks

may decide or be forced by regulators to scale down by selling off assets and reducing

deposit taking. Retrenchment may be the preferred strategy, if bank charter values are

considerable. The choice between growth and retrenchment strategies may further be

affected by agency problems between management and shareholders. Specifically, bank

5 For results on bank capitalization and profitability again see Demirgüç-Kunt and Huizinga (1999).6 Of course, portfolio choice in addition to growth may affect bank risk.

managers have to reckon that a bankruptcy may wipe out a considerable share of their

human capital.7

On balance, it is not clear whether one should expect riskier banks to have

relatively high deposit growth. Therefore, the revealed low explanatory power of deposit

growth regressions is not surprising. Indeed, in the last row, where we use data for all

countries, we see that the coefficient for Liquidity is not significantly different from zero

in either specification. This poor fit may be due to endogeneity problems we have not

addressed yet. All the same, we have found some preliminary evidence that markets

force riskier banks to pay higher interest rates, even though it is not clear whether riskier

banks are forced by the market to retrench.

IV. Market discipline and deposit insurance

One of the important factors that potentially affects market discipline is the

financial safety net. The safety net is the whole of financial regulations and institutions

that seeks to prevent or limit depositor losses in case of an (impending) bank failure. By

this definition, the safety net includes explicit and implicit deposit insurance, bank

insolvency resolution practices, regulatory forbearance policies, and the central bank’s

functioning as a lender of last resort. Thus, variation in the extent of market discipline

across countries should in part reflect international differences in the operation of the

financial safety net. In this section, we present the results of tests of the impact of explicit

deposit insurance, as part of the safety net, on market discipline via deposit rates and

7 Several authors have examined how agency problems may affect bank risk taking. Saunders, Strock and

Travlos (1990) find that ‘stock-holder controlled’ banking firms tend to take on more risk than‘managerially controlled’ firms. Gorton and Rosen (1995) instead argue that entrenched managers, whoown considerable stock in a bank, may take on more risk.

deposit growth. To our knowledge, this is the first attempt to systematically examine the

impact of the safety net on market discipline. To do this, it is necessary to combine cross-

country bank level data and institutional data that reflect differences in the financial

safety net. This greater degree of freedom also allows us to address certain endogeneity

issues we have not dealt with in the previous section.

Focusing first on the cost of bank liabilities, we estimate the following equation:

Interest i,j,t = α + Xi,j,t-1 β + Yj,t γ + δ D j,t + φ D j,t L i,j,t-1 + εi,j,t (2)

where Interest i,j,t is again the ratio of interest expense to interest-bearing debt minus the

government interest rate for bank i in country j, Xi,j,t-1 is a vector of variables for bank i in

country j lagged one period (including Liquidity), Li,j,t-1 is short-hand for Liquidity, Yj,t

is a vector of macroeconomic controls for country j, Dj,t is a dummy variable flagging the

existence of an explicit deposit insurance regime and, finally, εi,j,t is a stochastic term. The

macroeconomic controls are Inflation, real Gdp Growth, and real Gnp per capita.8 In

some specifications, we also include interaction terms of the deposit insurance dummy,

Dj,t with the liquidity variable.9

As before, negative estimates of β are interpreted as evidence of market

discipline. A negative (positive) estimate of δ would indicate that the existence of an

explicit deposit insurance scheme reduces (increases) bank interest rates. A reduction of

required deposit rates on account of deposit insurance suggests that on balance depositors

8 In Demirgüç-Kunt and Huizinga (1999) we estimate a specification where the dependent variable is the

ratio of interest expense to interest bearing debt and government rate is included in the vector of macrocontrols. The results are very similar.

value deposit insurance (taking into account any induced effects on bank risk taking and

the likelihood of bank failure). The interaction terms of the deposit insurance variable

with bank risk factors enable us to estimate whether an explicit deposit insurance system

reduces or enhances market discipline. Specifically, positive estimates of the φ coefficient

suggest that explicit deposit insurance reduces market discipline, since the absolute value

of β + φ becomes smaller (for a negative estimate of β), indicating a reduced sensitivity

of liability interest rates to bank risk.

We estimate three different specifications. First, we estimate (2) as above. As

indicated, bank liquidity is generally endogenous. Banks may, for instance, to some

extent try to avoid market discipline (higher interest rates) by increasing their liquidity.

Thus, it would be appropriate to instrument for liquidity using exogenous influences on

bank operations such as macro shocks. A second potential problem with (2) is the

possibility of a purely mechanical negative relationship between the average bank interest

rate and liquidity resulting from increased deposits, if the official reserve requirement is

binding. To address these problems, we perform a two stage regression where we first

regress Liquidity on all the exogenous variables in (2) plus the Government rate and the

Reserve rate. The Reserve rate is bank reserves (at the macro level) divided by bank

deposits (also at the macro level), as a proxy for the reserve requirement. This way, we

obtain a value of predicted Liquidity, which is hopefully purged of its endogenous

elements. Specifically, changes in predicted Liquidity will reflect changes in macro

reserve policy rather than changes in a bank’s required reserves (for a given reserve

policy). Efforts by individual banks to avert market discipline (in the form of higher

9 In Demirgüç-Kunt and Huizinga (1999) we also include bank capitalization and profitability as bank risk

factors and explore their interactions with deposit insurance. Those results are not significant or robust.

deposit rates) by increasing liquidity will similarly not register in predicted Liquidity. To

complete the two-stage procedure, we estimate a regression as in (2) after replacing

Liquidity by its predicted value.

The resulting estimation still ignores that market discipline can alternatively occur

through deposit quantity adjustment. Thus, we need to control for deposit growth, which

is itself endogenous. This is what we do in specification (3). As in the case of Liquidity,

we obtain predicted values of deposit growth by regressing deposit growth on all the

exogenous variables in the model plus bank asset size. The second stage now includes

both predicted liquidity and deposit growth.

For each specification we estimate two versions. Version (a) does not include

the deposit insurance dummy or its interaction term, while version (b) does include them.

In specifications 2 and 3 Liquidity is replaced by its predicted value, and the interaction

term is given by the insurance dummy multiplied by this predicted Liquidity.

We estimate all three specifications by OLS and for mean values of all variables

by the between estimator. Because the results are very similar in most cases, we only

report those of the OLS estimations but note the differences, if any, in case we employ

the between estimator.10

The results are given in Table 5.11 Liquidity is negative and highly significant in

all specifications. Instrumenting for liquidity, or introducing deposit growth into the

equation do not change this result, as is seen from specifications (2) and (3). This is also

10 We do not discuss within estimates as there is no variation in the deposit insurance dummy over time

for the countries in our sample except for Greece and Peru.11 Included banks are from 30 countries: Australia, Austria, Belgium, Bolivia, Denmark, Finland, France,

Greece, Ireland, Luxembourg, Mexico, Malta, Nigeria, Nepal, the Netherlands, New Zealand, Norway,Pakistan, Peru, the Philippines, Russia, Sri Lanka, South Africa, Spain, Switzerland, Taiwan, Thailand,Tunisia, the United Kingdom, and the United States.

the case when we use the between estimator.12 Overall, these results confirm our country-

level results indicative of the existence of market discipline. First stage Liquidity results

also provide interesting insights. 13 As expected, bank liquidity increases with bank short

term debt (deposits) and the Reserve rate, while a higher government interest rate leads to

lower liquidity. Also, banks in higher growth countries tend to be more liquid.14

Turning to our control variables, the negative coefficient on Overhead suggests a

trade-off between interest and non-interest expenses. Indeed, banks with lower overhead

probably pay more interest because they are less attentive to customers’ demands for

convenience. The coefficient on the Short Term Debt/Total Debt variable is negative and

12 Using the between estimator, relevant coefficient estimates for different specifications are:

(1a) (1b) (2a) (2b) (3a) (3b) Liq. -.032*** -.311*** -1.057*** -1.194*** -1.051*** -1.020*** (.011) (.059) (.115) (.130) (.116) (.113)Dep. Gro. -.118

(.178)D. Ins. -.019* -.158*** .022

(.010) (.022) (.025)Liq x .326*** .793*** -.000D. Ins. (.058) (.100) (.167)

adj.R2 .60 .67 .68 .74 .69 .69N 1610 1610 1462 1462 1435 1435

Heteroskedasticity-corrected standard errors are in parentheses.

13 The estimated first stage OLS equation is:Liquidity= -.053 -.213Overhead +.187Short Term Debt/Total Debt + .891 Inflation + .012 Growth +.017Gnp/cap -.024 Gov. rate (.021) (.126) (.013) (.060) (.001) (.009) (.006)

+.011 Reserve rate Adj. R sq. =.13 No of obs: 5731 (.031)

The estimated first stage between equation is:Liquidity= -.032 +.377Overhead +.194Short Term Debt/Total Debt + .318 Inflation + .006 Growth +.001Gnp/cap -.001 Gov. rate (.031) (.183) (.028) (.095) (.002) (.004) (.001)

+.135 Reserve rate Adj. R sq. =.14 No of obs: 2073 (.040)

Heteroskedasticity-corrected standard errors are in parentheses.

14 To further check that the relationship between bank interest rates and liquidity is not simply mechanical,we conduct additional tests. First, we replicate our results including only those banks that have aboveaverage liquidity in any bank-year cell. These are the banks for which any reserve requirement isprobably not binding. Second, we replicate our results for only those banks which have a ratio of liquidassets to short term debt exceeding the macro reserve rate to again exclude those banks where the reserverequirement is most likely to be binding. Our results are not significantly different in either case.

significant in the first specification, suggesting that on average short term bank debt is

less expensive. In specifications (2) and (3) where this variable is also entered in the first

stage liquidity equation, it loses significance in the second stage interest equation. Thus,

it appears that the amount of short term debt, which is mostly deposits, is important to the

extent that it induces bank asset liquidity.

Among the macro variables, we see that Inflation is negative and significant,

indicating that bank interest rates tend to reflect inflation to a lesser extent than the

government rate. The opposite is true for the growth and income variables. In

specification (3), we also incorporate deposit growth into the equation, which is shown to

be insignificant and not to change other results significantly.

Columns (b) add deposit insurance and its interaction with Liquidity to each

specification. The deposit insurance dummy enters all specifications with a negative and

significant coefficient. Using the between estimator, the results are similar although the

deposit insurance variables are not significant in specification 3 (see footnote 12).

Overall, this suggests that an explicit deposit insurance scheme causes depositors to

perceive their deposits to be safer, leading to lower implicit bank interest rates. To

estimate the size of the deposit insurance effect on deposit rates, we calculate the

predicted values of the implicit interest expense evaluated at mean values of predicted

Liquidity for the two cases of explicit deposit insurance (the insurance dummy equals

one) and implicit deposit insurance (the insurance dummy equals zero). For specification

(2b) and (3b), we find that explicit deposit insurance reduces the average deposit rate by

around .017.15

15 Our result is consistent with the finding of Bartholdy, Boyle and Stover (undated) that explicit deposit

insurance reduces the deposit rate by 25 basis points. These authors, however, consider a rather small

The interaction term of Liquidity and the deposit insurance dummy, as seen, enters

all three specifications positively.16 Thus explicit deposit insurance significantly dampens

interest sensitivity to bank liquidity, suggesting that explicit deposit insurance reduces

market discipline. For specifications (2b) and (3b), an F-test of the hypothesis of β + φ = 0

is rejected, indicating that deposit insurance reduces, but not fully eliminates market

discipline through deposit rates. Finally, note that the interaction terms do not materially

affect the significance levels of the various bank and macro controls.

Differences in institutional quality of the country can also have an important

impact on market discipline. So far, we have excluded indicators of institutional quality

from our regressions. To test the sensitivity of our results, we performed several additional

regressions. We include indexes of institutional quality such as rule of law, quality of

contract enforcement and bureaucratic quality and their interactions with Liquidity in our

baseline specifications. The results indicate that in countries with greater institutional

development market discipline is stronger, as the various interaction terms develop

negative and significant coefficients. However, adding these regressors to our

specification does not change our results on the deposit institution variables. These results

are not reported.

On the basis of the international evidence, we conclude that explicit deposit

insurance reduces bank liability rates for the banking sector.17 Also, we find evidence that

sample of 13 OECD countries. Another difference is that these authors use a representative CD rate for aparticular country so that no bank-specific (risk) factors are included.

16 The interaction term is not significant in specification (3) of the between estimator, but this specificationdoes not fit the data as well as specification (2) since the added deposit growth term is not significant andadjusted R square is lower.

17 To see whether deposit insurance on net makes funding cheaper to banks, one also has to take intoaccount the insurance premiums that banks have to pay. Our estimate for the reduction in the depositinterest rate on account of explicit deposit insurance of 170 basis points exceeds the deposit premium ascharged by the countries in Table 1. Hence, deposit insurance on net appears to provide a subsidy.

deposit insurance makes individual bank interest rates less sensitive to bank liquidity,

thereby reducing market discipline on banks.

Analogously to Table 5, one can also estimate the impact of deposit insurance on

the growth rate of deposits. Table 6 reports these results. 18 As in the case of individual

country regressions, the results are generally inconclusive and the model has low

explanatory power. In contrast to the interest regressions, estimating deposit growth and

interest simultaneously –as we do in specification 3- improves the fit of the deposit

growth model. As expected, banks with greater deposit growth are those that pay higher

interest rates. In these specifications, predicted Liquidity also develops a positive and

significant coefficient, indicating that less risky banks are able to attract more deposits

consistent with market discipline. 19 Nevertheless, adjusted R-squares are very low as

evidence of a rather poor fit of the model and the explicit insurance dummy and its

interaction enters the various specifications only with insignificant coefficients. We get

very similar results using the between estimator.20 Thus, we find no evidence that deposit

insurance on average causes higher or lower deposit growth.

According to Buser, Chen and Kane (1981), it is this subsidy that entices banks to accept the regulationsthat often come with deposit insurance. An important objective of deposit insurance, however, is toprevent destructive bank runs. If successful in this regard, deposit insurance can lower deposit rates to thebenefit of the banks even if there is no implicit subsidy forthcoming.

18 Included banks are from 51 countries: Argentina, Australia, Austria, Belgium, Bolivia, Canada, China,Chile, Colombia, Costa Rica, Denmark, Dominican Republic, Ecuador, Finland, France, Ghana, Greece,Guatemala, Haiti, Hong Kong, Honduras, Hungary, Indonesia, Ireland, Japan, Jordan, Kenya, Korea,Luxembourg, Malta, Mexico, Moracco, Nepal, the Netherlands, New Zealand, Nigeria, Norway,Panama, Pakistan, Peru, the Philippines, Singapore, Thailand, Tunisia, South Africa, Spain, Sri Lanka,Switzerland, the United Kingdom, the United States, and Zambia.

19 In Demirgüç-Kunt and Huizinga (1999) we see that bank capitalization is more strongly correlated withdeposit growth. However, this result does not survive when we address the endogeneity of bank equityand try to instrument for it.

20 The only significant difference is that the AssetSize variable is now positive and significant in allspecifications, indicating that larger banks have an easier time attracting deposits. This is what we expectif larger banks are perceived to be safer.

V. How do differences in deposit insurance design affect bank funding costs and

market discipline?

An increasing number of countries has adopted explicit deposit insurance systems

in response to bank system fragility around the world. As a case in point, the European

Union requires member states to have an explicit deposit insurance system in place since

1994.21 In its policy advice, the IMF has also advocated the adoption of explicit deposit

insurance systems.22 Notwithstanding the current movement to explicit schemes, the U.S.

experience has shown that a badly designed and administered deposit insurance scheme

can exacerbate bank moral hazard to increase risk (see Kane, 1989). Hence, the details of

deposit insurance design may be very important in shaping banks’ incentives to take on

undue risk.

So far in the paper, we have shown that the existence of explicit deposit insurance

(i) reduces the interest rates required by depositors, and (ii) also reduces market

discipline. In this section, we investigate whether several specific deposit insurance

design features (as presented in Table 1) affect bank deposit rates and market discipline.

In particular, we estimate a set of cross-country regression of the following

Interest i,j,t = α + Xi,j,t-1 β + Yj,t γ + δ D j,t + φ D j,t Li,j,t-1 + λD j,t F j,t + θ D j,t F j,t Li,j,t-1

+ εi,j,t (3)

where Fj,t is the value of a particular design feature, and other variables are as defined

before. Typically, a design feature is a variable that indicates whether or not the system is

implicit or explicit with a particular design feature. For instance, the co-insurance design

21 See Lars Fredborg (1995).22 For details, see Garcia (1999).

feature is zero for implicit schemes, 1 for explicit schemes without co-insurance and 2 for

explicit systems that feature co-insurance by depositors. Only the coverage limit is a scale

variable, and it takes the value 0 for implicit deposit insurance and the statutory coverage

limit divided by Gnp/capita if explicit. The coefficient on the design feature, i.e. λ,

indicates the impact of a particular design feature on interest rates, compared to an

explicit system without that feature. A negative sign indicates that the feature is valuable

to the creditors, as it lowers required deposit interest rates. A positive or insignificant sign

instead may well indicate that the design feature is not credible. As before, we also

include interaction terms of a particular deposit insurance design feature with bank

liquidity. Again, positive estimates of the θ coefficients suggest that the design feature

reduces market discipline.

The regression results regarding the impact of different design features on implicit

interest rates are reported in Table 7. Since potential endogeneity of liquidity and deposit

growth do not affect the results significantly, we take specification 1 in Table 5 as the

baseline specification to introduce the design features and their interaction with liquidity.

As with the deposit insurance dummy, design features hardly vary over time. Therefore,

we only estimate the pooled-OLS specification and the between specification based on

mean values. These are reported in panels A and B of Table 7. For brevity, only the

relevant coefficients are reported.

In the first column of Table 7, we include a dummy variable for the existence of

co-insurance. Since this is a design feature that introduces additional depositor risk, we

expect Co-insurance to enter the regressions with a positive sign. The estimated

coefficients are indeed positive but only significant at the 10 percent level in the OLS

specification, perhaps because only two counties in the sample (Ireland and the United

Kingdom) have co-insurance. To see the impact of co-insurance on market discipline, we

examine the interaction term, which is negative and significant. Thus, co-insurance

strengthens market discipline, as expected.

The next column in Table 7 investigates the impact of the explicit coverage limit

on interest rates and market discipline. The coverage limit is calculated as the statutory

coverage limit (as in Table 1) divided by GDP per capita. Ceteris paribus, one expects a

higher coverage limit to lower required deposit rates and weaken market discipline as

higher coverage should enhance depositor safety. Indeed, we see that a higher coverage

limit significantly reduces interest rates and weakens market discipline in both the OLS

and between specifications.

The next two columns in Table 7 refer to the insurance coverage of foreign

currency deposits and interbank deposits in explicit systems, respectively. The negative

and significant coefficients for these design features in the two equations suggest that

expanded coverage tends to reduce interest rates further. There is a difference, however,

between the two. Coverage for interbank deposits reduces interest expense at the cost of

weaker market discipline. Coverage for foreign currency deposits instead appears to

strengthen market discipline.

The next set of columns include dummy variables relating to the ex ante funding

of the scheme, and what sources the funding comes from. The funding variable takes on a

value of 0 if the scheme is implicit, a value of 1 if the fund is unfunded, a value of 2 if

the fund is unfunded but callable, and a value of 3 if the scheme is funded. The results

indicate that an ex-ante funded scheme leads to the lowest interest rates, probably

because it is the most credible. However, it also weakens market discipline the most.

These results hold regardless of the specification.

The next design variable represents the source of the funding. This sourcing

variable takes on a value 0 if the insurance system is implicit, a value of 1 if funds come

from banks only, a value of 2 if funds come from the government and banks, and a value

of 3 if funds come from the government only. Schemes funded only by the government

lead to the largest declines in interest rates, but also to the largest reductions in market

discipline. Thus, the more heavily the government is involved in the funding, the lower

are required deposit rates and the market discipline is weaker. This may reflect that

government funding is deemed more credible than bank funding. These results hold for

both OLS and between estimations.

The next two variables relate to the fund’s management. The first variable takes

on a value of 0 if the scheme is implicit, a value of 1 if the management is private, a

value of 2 if the management is jointly public and private, and a value of 3 if

management is entirely public. This management index enters with negative and

significant coefficients in both OLS and between specifications. This suggests that

publicly managed schemes are more effective in reducing interest rates. However, they

also tend to reduce market discipline (and increase moral hazard), as can be seen from the

positive and significant sign of the interaction term of the management index with

liquidity. Looking at the individual dummies for joint and private management, we see

that joint schemes have no additional impact on interest rates, while private schemes lead

to a lower reduction in interest expense (because their coefficient is estimated to be

positive and significant). Both private and joint schemes appear to put checks on the

moral hazards associated with deposit insurance, as their interaction terms with Liquidity

are negative and significant in both specifications.

Finally, we have a membership dummy variable, that equals 0 for implicit

schemes, 1 if there is an explicit scheme with compulsory bank membership, and 2 if the

scheme has voluntary membership. Ex ante, we may think that voluntary bank

membership in the scheme leads to compounded problems of adverse selection and moral

hazard, leading to higher deposits rates. This is what we see in both specifications. There

is also some evidence that voluntary membership strengthens market discipline, but the

interaction term is significant at the 10 percent level only in the between specification.

To sum up, we find that all explicit insurance design features, except co-

insurance, private or joint management of the fund and voluntary membership, lead to

lower required interest rates. In terms of market discipline, we see that co-insurance,

joint and private management and coverage of foreign currency deposits increase market

discipline, whereas all other design features reduce it.

VI. Conclusions

Policy makers around the world design and operate financial safety nets so as to

prevent costly bank insolvencies. However, designing and implementing an effective

safety net is a difficult task, since overgenerous protection of banks may easily introduce

risk-enhancing moral hazard, and destabilize the very system it is meant to protect. The

challenge facing policy makers is to provide depositor protection without unduly

undermining market discipline. The design of the safety net is therefore crucial in

providing the right mix of market and regulatory discipline of banks. A considerable

amount of theoretical and prescriptive literature exists on this issue. At this point,

empirical work is desperately needed to better inform policy recommendations. Empirical

work of this kind has been lacking because of an absence of available information on

safety net design across countries.

This paper makes two important contributions to the literature. First, for a large

number of developed and developing countries we analyze bank-level interest expense

and deposit growth data to investigate if there is any evidence of market discipline in

individual countries. Second, using cross-country information on deposit insurance

systems we investigate the impact of explicit deposit insurance and its key features on

bank interest rates and market discipline.

Our country-level results show that many countries around the world retain some

degree of market discipline, regardless of the different safety nets they may have. This

may be due to incomplete coverage or lack of credibility of existing schemes, or reflect

inherent costs involved in recovering funds even if schemes are fully credible.

To investigate the linkages between market discipline and deposit insurance, we

estimate regressions with cross-country data. Here our results show that the existence of

an explicit insurance lowers banks’ interest expenses and makes interest payments less

sensitive to bank risk, particularly to bank liquidity. Thus explicit deposit insurance is

found to reduce market discipline on banks by their creditors.

We also investigate whether specific deposit insurance design features matter,

since the various countries with explicit deposit insurance operate systems with vastly

different coverage, funding, and management. Here our results suggest that higher

explicit coverage, broader coverage, and the existence of earmarked insurance funds,

government provision of funds, and public management of deposit insurance lower

deposit rates. On the other hand, existence of co-insurance, private management of

scheme, and voluntary membership are features that increase required rates.

We also examine the impact of these features on market discipline. We find that

higher coverage, coverage of interbank funds, existence of ex-ante funding, government

provision of funds, and public management only reduce market discipline, while co-

insurance, coverage of foreign currency deposits, and private and joint management of

insurance schemes may improve market discipline.

This research has important policy implications. Deposit insurance is found to be

valued by bank creditors, since it leads to lower required interest rates. The increase in

perceived safety for depositors, however, comes at a cost of a reduction in market

discipline. Thus at a broad level, the adoption of an explicit deposit scheme involves the

trade-off between increased depositor safety and reduced market discipline on banks.

To probe deeper, it is necessary to evaluate the desirability of key deposit

insurance design features. Our evidence on this count suggests that co-insurance,

private/public joint management of the fund and coverage of foreign currency deposits

are desirable since they lead to stronger market discipline with a negative or insignificant

impact on interest rates. Higher explicit coverage limits, coverage of interbank deposits,

existence of an earmarked fund, government-only funding of the deposit insurance

scheme, and public management are found to lower required deposit rates, but at a cost of

lower market discipline. Thus, these features again presents policy makers with an

important trade-off between depositor safety and bank risk taking.

References

Avery, Robert B, Terrence M. Belton and Michael A. Goldberg, 1988, Market disciplinein regulating bank risk: new evidence from the capital markets, Journal of Money,Credit, and Banking 20, 597-610.

Baer, Herbert and Elijah Brewer, 1986, Uninsured deposits as a source of marketdiscipline: a new look, Quarterly Journal of Business and Economics 24, 3-20.

Bartholdy, Jan, Glenn W. Boyle and Roger D. Stover, undated, Deposit insurance, bankregulation and interest rates: some international evidence, mimeo, University ofOtago, New Zealand.

Benston, George J. and George G. Kaufman, 1998, Deposit insurance reform in the FDICImprovement Act: the experience to date, Economic Perspectives 22 (FederalReserve Bank of Chicago), 2-20.

Brewer, Elijah III and Thomas H. Mondschean, 1994, An empirical test of the incentiveeffects of deposit insurance, Journal of Money, Credit, and Banking 26, 146-164.

Buser, Stephen A., Andrew H. Chen, and Edward J. Kane, 1981, Federal depositinsurance, regulatory policy, and optimal bank capital, Journal of Finance 35, 51-60.

Calomiris, Charles, 1997, The postmodern safety net: Lessons from developed anddeveloping economies. Washington: American Enterprise Institute.

Calomiris, Charles and Berry Wilson, 1998, Bank Capital and Portfolio Management:The 1930’s “Capital Crunch” and Scramble to Shed Risk,” NBER Working PaperNo. 6649.

Cook, Douglas O. and Lewis J. Spellman, 1994, Repudiation risk and restitution costs:toward understanding premiums on insured deposits, Journal of Money, Credit,and Banking 26, 439-459.

Demirgüç-Kunt, A. and E. Detragiache, 1998, The determinants of banking crises indeveloped and developing countries, IMF Staff Papers 45, 81-109.

Demirgüç-Kunt, A. and H. Huizinga, 1999, Market discipline and financial safety netdesign, Policy Research Working Paper, No: 2183, World Bank, Washington,D.C.

Ellis, David M. and Mark J. Flannery, 1992, Does the debt market assess large banks’risk?, Journal of Monetary Economics 30, 481-502.

Ely, Bert, 1986, Private sector depositor protection is still a viable alternative to federaldeposit insurance, Issues in Bank Regulation winter, 40-47.

Flannery, Mark J., 1998, Using market information in prudential bank supervision: areview of the U.S. empirical evidence, Journal of Money, Credit, and Banking 30,273-305.

Flannery, Mark J. and Sorin M. Sorescu, 1996, Evidence of bank market discipline insubordinated debenture yields: 1983-1991, Journal of Finance 51, 1347-1377.

Garcia, Gillian G., 1999, Deposit insurance: a survey of best practices, IMF WP/99/54,Washington, D.C.

Gorton, G. and G. Pennacchi, 1990, “Financial Intermediaries and Liquidity Creation,”Journal of Finance, 45, 49-72.

Gorton, G. and R. Rosen, 1995, Corporate control, portfolio choice, and the decline ofbanking, Journal of Finance 50, 1377-1420.

Gorton, Gary and Athony M. Santomero, 1990, Market discipline and bank subordinateddebt, Journal of Money, Credit, and Banking 22, 119-128.

Fredborg, Lars, 1995, EC deposit-guarantee directive, International Financial LawReview.

Hannan, Timothy H. and Gerald A. Hanweck, 1988, Bank insolvency risk and the marketfor large certificates of deposit, Journal of Money, Credit, and Banking 20, 203-211.

Kane, Edward J., 1987, Who should learn what from the failure and delayed bailout ofthe ODFG? in Federal Reserve Bank of Chicago, Proceedings from a Conferenceon Bank Structure and Competition, 306-326

Kane, Edward J., 1989, The S&L insurance Mess: How Did it Happen?, Washington,D.C.: Urban Institute Press.

Kane, Edward J., 1999, Designing financial safety nets to fit country circumstances: aprogress report, mimeo, World Bank.

Martinez Peria, Maria Soledad and Sergio L. Schmukler, 1999, Do depositors punishbanks for “bad” behavior?, Policy Research Paper no. 2058, World Bank,Washington, D.C.

Park, Sangkyun, 1995, Market discipline by depositors: evidence from reduced-formequations, Quarterly Review of Economics and Finance 35, 497-514.

Park, Sangkyun and Stavros Peristiani, 1998, Market discipline by thrift depositors,Journal of Money, Credit, and Banking 30, 347-364.

Saunders, A., E. Strock, and N.G. Travlos, 1990, Ownership structure, deregulation, andbank risk taking, Journal of Finance 41, 871-895.

Wall, Larry, 1998, A plan for reducing future deposit insurance losses: puttablesubordinated debt, Economic Review, Federal Reserve Bank of Atlanta(July/August), 2-17.

White, Halbert, 1980, A heteroskedasticity-consistent covariance matrix estimator and adirect test for heteroskedasticity, Econometrica 48, 817-838.

Table I. Deposit Insurance System Features

Implicit denotes lack of an explicit scheme. Date established refers to data of statute enactment. Co-insurance is a dummy variable that takes on a value of one if depositors face adeductible in their insured funds. Coverage limit refers to the amount the authorities insure explicitly. Foreign currency deposits and Interbank deposits take a value of one if insurancecoverage extends to those areas. Funding takes a value of one if the scheme is funded ex-ante. Source of funding can be from government only (2), banks and government (1), or banksonly (0). The premium banks pay is given as a percentage of deposits or liabilities. Management of the fund can be official (1), official/private joint (2), or private (3). Membership of thefund can be compulsory or voluntary. Sources: “Deposit protection arrangement: a survey”, Alexander Kyei, IMF WP/90/134, Washington, D.C.; “Global Surveys”, Institute ofInternational Bankers, 1998, 1997, 1996, 1995, 1994; “Korea introduces bank deposit insurance scheme”, Dong Won Ko, International Financial Law Review, April 1997; "Bankingfailures in developing countries: an auditor’s perspective”, Javed Nizam, International Journal of Government Auditing, January 1998; “Belgium implements deposit guarantee-scheme”,Andre Bruyneel, and Axel Miller, International Financial Law Review, June 1995; “Incentive structure and resolution of financial institution distress: Latin American experience”, ThomasGlaessner, and Ignacio Mas, Latin America and the Caribbean Regional Office, World Bank, November 1991; “Reform of the Finnish deposit guarantee scheme”, Veli-Pekka Valori andJukka Vesala, Bank of Finland Bulletin, March 1998; "Japan: Stimulation package", anon., Oxford Analytica Brief, December 1997; “EC deposit-guarantee directive”, Lars Fredborg,International Financial Law Review, December 1995.

Countries Type:

Implicit (0),Explicit (1)

DateEstablished

Co-insurance

CoverageLimit

(NationalCurrency or

ECU)

ForeignCurrencyDepositsCovered

Interbank DepositsCovered

Funding:

Unfunded (0),Funded (1)

Source of Funding:

Banks only (0),Gov. and Banks (1),

Gov. only (2)

Insurance Premiumof Deposits or

Liabilities

Management:

Public (1),Joint (2),

Private (3)

Membership:

Voluntary (0),Compulsory (1)

Argentina 1 1979 &1995

0 20,000 US $ 1 0 1 0 .0036-.0072 ofaverage deposits

2 1

Australia 0Austria 1 1979 1 260,000 ATS 1 0 0 1 callable 3 1Belgium 1 1985 0 15,000 ECU 0 0 1 1 .0002 of deposits

from clients2 1

Bolivia 0Canada 1 1967 0 60,000 Can. $ 0 1 1 1 0.0033 of insured

deposits, max.1 1

Chile 1 1986 1 90% ofdemand

deposits up to120 Ch$

1 0 0 2 callable 1 1

China 0Colombia 1 1985 1 75% per

deposit or Col$10 Mil.

0 1 1 0 .003 of insureddeposits

1 1

Costa Rica 0

Denmark 1 1988 0 300,000 DKR 1 0 1 1 0.002 of insureddeposits, max.

2 1

DominicanRep.

1 1962 1 RD $ 8,000 1 0 1 1 .0018 of deposits 2 0

Ecuador 0Finland 1 1969 0 150,000 FIM 1 0 1 1 0.0005 to 0.0030 of

insured deposits3 1

France 1 1980 0 400,000 Fr 1 0 0 0 callable 3 1Ghana 0Greece 1 1995 0 20,000 ECU 1 0 1 0 0.00025-0.0125 of

eligible deposits2 1

Guatemala 0Haiti 0Honduras 0Hong Kong 0Hungary 1 1993 0 1,000,000 Ft 1 0 1 1 0.003 of insured

deposits, max.1 1

Indonesia 0Ireland 1 1989 1 90% of 20,000

ECU1 0 1 0 0.002 of insured

deposits1 1

Italy 1 1987 1 100% of first200 Mil. ITLand 80% of

next 800Mil.

0 0 0.004 to .008 ofinsured deposits

callable

3 0

Japan 1 1971 0 Full coverageuntil March

2001

1 0 0 1 .000048 of insureddeposits

2 1

Jordan 0Kenya 1 1985 0 K Sh 100,000 1 1 1 1 0.004 of deposits 1 1Korea 1 1996 0 Full coverage

till 20000 0 1 1 0.0005 of insured

deposits1 1

Luxembourg 1 1989 1 15,000 ECU 1 0 0 0 callable 3 1Malta 0Mexico 1 1986 0 No limit 1 0 1 1 0.003 to.005 of

covered liab. plus.007 callable

1 1

Morocco 0Nepal 0Netherlands 1 1979 0 20,000 ECU 1 0 0 Callable 1 1

New Zealand 0Nigeria 1 1988 0 50,000 N 0 1 1 1 0.00937of deposits 1 1Norway 1 1961 0 2,000,000

NOK1 0 1 1 0.00005 of assets

and .0001 of deposits3 1

Pakistan 0Panama 0Peru 1 1993 0 4,600 Sl. 0 0 1 1 0.0065 to .0145 of

total deposits2 1

Philippines 1 1963 0 P 100,000 1 1 1 1 0.002 of totaldeposits

1 1

Russia 0Singapore 0South Africa 0Spain 1 1977 0 15,000 ECU 1 0 1 1 0.0002 of

deposits,max.2 1

Sri Lanka 1 1987 0 1,470 US $ 0 0 1 1 .0015 of deposits 1 0Sweden 1 1996 0 250,000 SEK 1 0 1 1 0.005 of deposits and

.001 callable1 1

Switzerland 1 1984 0 30,000 Sw F 0 0 0 1 callable 3 0Taiwan 1 1985 0 1,000,000 NT 0 0 1 1 0.00015 of insured

deposits1 0

Thailand 0Tunisia 0UnitedKingdom

1 1982 1 75% of 20,000GBP

1 0 0 0 Callable 3 1

United States 1 1934 0 100,000 US $ 1 1 1 1 0 to 0.0027 ofdomestic deposits

1 1

Zambia 0

Table II. Bank Characteristics by Country

Interest Expense is the ratio of interest expense to interest paying debt. Deposit Growth is the percentage growth in realdeposits. Equity is book value of equity (assets minus liabilities) to total assets. Liquidity is defined as liquid assets tototal assets. Profit is given by before tax profits divided by total assets. Overhead is personnel expenses and other non-interest expenses over total assets. Short Term Debt/Total Debt is customer and short term funding to total interestpaying debt. Data source is the BankScope data base of Fitch IBCA. Figures are 1990-1997 averages.

Interest Expense Deposit Growth Liquidity Overhead Short Term Debt/Total Debt

Argentina 0.078 0.207 0.128 0.070 0.949Australia 0.068 0.109 0.097 0.025 0.915Austria 0.054 0.048 0.127 0.025 0.918Belgium 0.057 0.082 0.203 0.022 0.984Bolivia 0.090 0.131 0.138 0.046 0.984Canada . 0.084 0.196 0.020 0.978Chile . 0.051 0.310 0.031 0.918China . 0.189 0.309 0.014 0.840Colombia . 0.105 0.219 0.084 0.930Costa Rica . -0.038 0.268 0.051 0.952Denmark 0.042 0.041 0.226 0.035 0.968Dominican Rep. . 0.176 0.308 0.063 0.619Ecuador . 0.278 0.302 0.073 0.931Finland 0.058 0.075 0.165 0.018 0.761France 0.072 0.029 0.144 0.033 0.897Ghana . 0.301 0.487 0.069 0.980Greece 0.096 0.225 0.327 0.035 0.997Guatemala . 0.093 0.230 0.058 0.816Haiti . 0.132 0.501 0.042 0.942Honduras . 0.172 0.262 0.043 0.899Hong Kong . 0.043 0.364 0.021 0.921Hungary . 0.141 0.090 0.043 0.943Indonesia . 0.240 0.219 0.028 0.892Ireland 0.057 0.291 0.210 0.011 0.980Italy 0.060 0.007 0.349 0.035 0.887Japan . 0.006 0.133 0.014 0.952Jordan 0.052 0.040 0.326 0.025 0.985Kenya . 0.260 0.331 0.037 0.996Korea . 0.221 0.164 0.027 0.871Luxembourg 0.072 0.046 0.430 0.011 0.945Malta 0.042 0.187 0.207 0.014 0.993Mexico 0.228 0.123 0.220 0.043 0.956Morocco . 0.119 0.335 0.027 0.996Nepal 0.052 0.266 0.379 0.024 0.987

Interest Expense Deposit Growth Liquidity Overhead Short Term Debt/Total Debt

Netherlands 0.057 0.100 0.357 0.014 0.883New Zealand 0.065 0.102 0.101 0.027 0.974Nigeria 0.075 0.085 0.571 0.085 0.998Norway 0.064 0.074 0.058 0.023 0.903Pakistan 0.070 0.164 0.433 0.030 1.000Panama 0.062 0.117 0.195 0.020 0.948Peru 0.077 0.311 0.239 0.077 0.988Philippines 0.073 0.151 0.257 0.042 0.999Russia 0.137 0.113 0.449 0.083 0.987Senegal 0.020 0.087 0.000 0.048 0.981Singapore 0.034 0.107 0.313 0.012 0.990South Africa 0.116 0.150 0.164 0.039 0.925Spain 0.070 0.104 0.179 0.030 0.990Sri Lanka 0.084 0.113 0.207 0.035 0.877Switzerland 0.045 0.055 0.177 0.035 0.831Taiwan 0.054 0.204 0.148 0.015 0.999Thailand 0.082 0.160 0.094 0.019 0.961Tunisia 0.052 0.040 0.191 0.020 0.906United Kingdom 0.060 -0.038 0.319 0.027 0.931United States 0.035 0.071 0.126 0.036 0.826Zambia . 0.127 0.505 0.114 1.000

Table III. Market Discipline – Interest Rates

The equation estimated is Interest t = α + β1 Liquidityt-1+ β2 Overheadt-1+ β3 Short Term Debt/Total Debtt-1+ εt. The dependent variable is the ratio of interest expense to interest payingdebt minus Government rate. Liquidity is defined as liquid assets to total assets. Overhead is personnel expenses and other non-interest expenses over total assets. Short term debt to totaldebt is customer and short term funding to total interest paying debt. Data source for bank data is the BankScope data base of Fitch IBCA. Government rate data are from the IMF’sInternational Financial Statistics. The sample period for each country is 1990-1997, where available. Three different estimation methods are used. The first one pools cross-bank time-series data and uses Ordinary Least Squares. The Between Estimator takes mean values for each bank over the sample period. The Within Estimator uses pooled data but includes bank andtime fixed effects. The last row for each specification uses data for all countries. Significance tests are based on White’s heteroskedasticity consistent errors. Bold numbers are significantat levels equal to or less than 5 percent.

Pooled-OLS Between Estimates Within EstimatesLiquidity Adj.R-sq. N Liquidity Adj. R-sq. N Liquidity Adj. R-sq. N

Australia 0.039 0.06 132 0.105 0.11 39 0.034 0.10 126Austria 0.073 0.09 158 0.055 0.02 42 0.096 0.12 155Belgium 0.005 0.13 183 -0.012 0.29 46 -0.022 0.43 178Bolivia -0.044 0.21 41 -0.014 0.37 13 -0.185 0.63 40Denmark -0.008 0.16 243 -0.007 0.55 56 -0.004 0.76 241Finland 0.091 0.10 34 0.141 0.27 7 0.034 0.66 33France -0.028 0.05 832 -0.050 0.08 217 -0.040 0.15 818Greece 0.032 0.00 60 0.048 0.13 15 0.055 0.61 58Ireland 0.062 0.00 26 0.079 0.00 13 0.050 0.00 24Italy -0.001 0.31 69 -0.001 0.14 52 -0.016 0.39 67Luxembourg -0.035 0.07 632 -0.040 0.09 122 -0.034 0.08 629Malta 0.010 0.05 19 0.008 0.00 6 0.010 0.00 18Mexico -0.006 0.00 36 -0.018 0.19 18 0.041 0.05 30Nepal -0.221 0.28 12 -0.119 0.38 6 -0.033 0.99 10Netherlands 0.005 0.09 114 0.005 0.08 32 0.005 0.13 110New Zealand -0.037 0.32 25 -0.076 0.60 9 -0.030 0.61 23Nigeria 0.054 0.00 57 0.048 0.00 25 0.065 0.22 46Norway -0.053 0.07 73 -0.002 0.40 15 -0.036 0.10 74Pakistan 0.016 0.09 87 0.019 0.09 20 0.020 0.50 87Peru -0.383 0.25 78 -0.450 0.54 22 -0.131 0.85 77Philippines -0.127 0.24 83 -0.191 0.29 22 -0.050 0.72 79Russia -0.080 0.00 53 -0.273 0.00 40 -0.126 0.98 31South Africa 0.030 0.02 49 -0.010 0.00 15 0.014 0.29 47Spain 0.010 0.09 351 -0.002 0.09 81 0.015 0.24 336Sri Lanka -0.064 0.00 25 -0.077 0.76 6 -0.018 0.67 24Switzerland 0.016 0.07 783 -0.011 0.08 175 -0.018 0.25 766Taiwan -0.025 0.09 119 -0.046 0.27 27 -0.024 0.34 119

Pooled-OLS Between Estimates Within EstimatesLiquidity Adj.R-sq. N Liquidity Adj. R-sq. N Liquidity Adj. R-sq. N

Thailand 0.142 0.06 94 -0.015 0.00 16 0.036 0.73 92Tunisia -0.099 0.18 51 -0.068 0.00 10 -0.055 0.58 51United Kingdom -0.026 0.09 171 -0.032 0.11 54 -0.028 0.11 163United States -0.004 0.10 2335 -0.012 0.21 520 -0.010 0.33 2312

All -0.041 0.12 7026 -0.111 0.16 1744 -0.032 0.11 6864

Table IV. Market Discipline – Deposit Growth

The equation estimated is Deposit Growtht = α + β1 Liquidityt-1+ β2 Overheadt-1+ β3 Asset Sizet-1+ β4 Government Ratet+ ε,t. The dependent variable is the real deposit growth.Liquidity is defined as liquid assets to total assets. Overhead is personnel expenses and other non-interest expenses over total assets. Asset Size is given by log transformation oftotal assets divided by Gdp. Bank data are from the BankScope data base of Fitch IBCA, and the GDP deflator, GDP and Government rate data are from the IMF’sInternational Financial Statistics. The sample period for each country is 1990-1997, where available. Three different estimation methods are used. The first one pools cross-bank time-series data and uses Ordinary Least Squares. The Between Estimator takes mean values for each bank over the sample period. The Within Estimator uses pooled databut includes bank and time fixed effects. Significance tests are based on White’s heteroskedasticity consistent errors. Bold numbers are significant at levels equal to or less than5 percent.

Pooled-OLS Between Estimates Within EstimatesLiquidity Adj. R-sq. N Liquidity Adj. R-sq. N Liquidity Adj. R-sq. N

Australia 0.040 0.01 95 -0.053 0.11 39 0.091 0.02 95Austria -0.182 0.09 122 -0.319 0.07 44 -0.090 0.09 122Belgium -0.215 0.07 115 -0.182 0.00 45 -0.240 0.08 115Bolivia 0.706 0.02 27 2.433 0.00 13 0.611 0.00 27Canada 0.196 0.02 180 0.230 0.21 55 0.172 0.04 180Chile -0.153 0.04 89 -0.350 0.08 26 -0.037 0.05 86China -1.238 0.05 23 -1.180 0.00 14 -0.868 0.02 23Colombia 0.231 0.02 73 -0.009 0.00 22 -0.587 0.32 73Costa Rica -0.225 0.70 15 0.168 0.00 15 0.199 0.71 15Denmark 0.099 0.05 192 0.353 0.00 55 0.139 0.08 192Ecuador -0.379 0.00 11 -2.293 0.00 13 -1.096 0.00 11Finland -0.199 0.00 28 -0.091 0.23 6 0.152 0.02 28France 0.171 0.01 872 0.108 0.02 277 0.173 0.02 872Greece -0.217 0.00 13 2.148 0.24 10 0.199 0.00 13Hong Kong 0.074 0.15 53 -0.327 0.00 23 0.061 0.12 53Honduras 2.349 0.28 16 3.956 0.34 8 2.330 0.34 16Hungary -1.489 0.00 21 2.392 0.28 24 -2.118 0.18 21Indonesia -0.467 0.07 250 -0.233 0.00 94 -0.514 0.07 250Ireland -0.287 0.24 34 -2.986 0.48 16 -0.038 0.25 34Italy 0.567 0.00 66 0.320 0.00 52 0.567 0.00 66Japan 0.103 0.10 627 0.695 0.25 147 0.071 0.16 627Jordan 0.078 0.38 34 0.577 0.95 5 -0.080 0.33 34Kenya -2.022 0.89 5 1.010 0.00 8 -2.022 0.89 5Korea 1.421 0.19 98 1.455 0.26 28 1.507 0.22 98Luxembourg -0.082 0.00 57 -0.199 0.14 67 -0.082 0.00 57Malta 0.036 0.13 13 3.631 . 5 0.146 0.26 13

Pooled-OLS Between Estimates Within EstimatesLiquidity Adj. R-sq. N Liquidity Adj. R-sq. N Liquidity Adj. R-sq. N

Mexico 0.706 0.06 11 2.363 0.56 11 0.697 0.13 11Nepal 0.206 0.30 10 0.694 . 5 0.395 0.13 10Netherlands -0.337 0.14 131 0.151 0.01 43 -0.346 0.11 131New Zealand -1.036 0.51 16 1.855 0.26 8 -1.134 0.52 16Nigeria -0.204 0.00 21 -0.814 0.00 19 -0.082 0.00 21Norway 0.479 0.01 54 -0.362 0.09 14 0.217 0.09 54Pakistan -0.127 0.29 66 -0.111 0.16 20 -0.037 0.41 66Peru -1.475 0.08 71 -1.951 0.06 21 -1.331 0.03 71Philippines -0.408 0.24 84 -0.785 0.19 26 -0.485 0.20 84South Africa 0.149 0.02 53 0.190 0.00 23 0.337 0.02 53Spain -0.144 0.08 267 0.017 0.04 82 -0.153 0.07 267Sri Lanka 0.315 0.00 18 1.637 0.98 6 0.463 0.00 18Switzerland -0.012 0.03 529 0.071 0.20 177 -0.069 0.05 529Thailand -0.891 0.16 81 -2.021 0.53 15 -0.848 0.24 81Tunisia 0.237 0.13 40 -1.014 0.05 9 0.278 0.06 40United Kingdom -0.089 0.05 217 -0.150 0.00 81 -0.093 0.05 217United States -0.088 0.01 1814 -0.055 0.10 524 -0.081 0.02 1813Zambia -1.636 0.00 6 3.556 . 4 -1.636 0.00 6

All -0.012 0.01 6537 0.016 0.02 2193

-0.024 0.01 6537

Table V. Deposit Interest Rates, Market Discipline and Deposit Insurance

The estimated model is Interest [Bank= i, Country=j, Time= t]= α + β1 Liquidityi,j,t-1+ β2 Overheadi,j,t-1 + β3 Short TermDebt/Total Debti,j,t-1 + β4 Inflation j,t + β5 Growthj,t + β6Gnp/capj,t + β7Deposit Insurance j,t + εi,j,t. The dependentvariable is the ratio of interest expense to interest paying debt minus the Government rate. Liquidity is defined as liquidassets to total assets. Overhead is personnel expenses and other non-interest expenses over total assets. Short TermDebt/Total Debt is customer and short term funding to total interest paying debt. Inflation is the annual inflation ratefrom the GDP deflator. Growth is the annual growth rate of real GDP per capita. Gnp/cap is the real GNP per capita in$10,000. Deposit Insurance is a dummy variable that takes on a value one if an explicit deposit insurance schemeexists. Bank data are from the BankScope data base of Fitch IBCA, and macro data are from the IMF’s InternationalFinancial Statistics. The sample period is 1990-1997. For each specification, column b also includes an interaction termof the Liquidity variable with the deposit insurance dummy. In specification (2) Liquidity is replaced by its predictedvalue from the regression of Liquidity j,t = α + β1 Overheadi,j,t+ β2 Short Term Debt/Total Debt j,j,t + β3 Inflation j,t +β4Growthj,t + β5 Gnp/capj,t t + β6 Government ratej,t + β7 Reserve Ratei,t-1 + εi,j,t. Reserve rate is given by reservesdivided by deposits. In specification (3) DepGro, predicted deposit growth variable is added to specification (2).DepGro is the predicted value obtained from the regression DepGro j,t = α + β1 Overheadi,j,t-1 + β2 Inflation j,t +β3Growthj,t + β4Gnp/capj,t t + β5 AssetSizei,j,t-1 + εi,j,t. Asset size is given by the log transformation of total assets dividedby Gdp. White’s heteroskedasiticy consistent standard errors are given in parentheses.

(1.a) (1.b) (2.a) (2.b) (3.a) (3.b)

Liquidity -.009**(.004)

-.197***(.032)

-.118***(.023)

-.369***(.044)

-.119***(.023)

-.369***(.044)

Overhead -.253***(.053)

-.279***(.053)

-.253***(.028)

-.231***(.027)

-.277***(.110)

-.176*(.111)

Short term debt/total debt -.010***(.003)

-.011***(.003)

.009(.005)

.004(.004)

.008(.005)

.004(.005)

Inflation -.603***(.069)

-.554***(.059)

-.110***(.035)

-.118***(.035)

-.100*(.053)

-.141***(.052)

Growth .277***(.052)

.228***(.044)

.064**(.032)

.018(.029)

.028(.170)

.101(.162)

Gnp/cap .005***(.001)

.001(.001)

.011***(.001)

.012***(.001)

.012***(.003)

.010***(.003)

Deposit Growth .034(.158)

-.078(.155)

Deposit insurance -.014***(.005)

-.061***(.009)

-.062***(.008)

Liquidity xDeposit insurance

.209***(.034)

.288***(.043)

.288***(.043)

No. of obs. 5194 5194 3411 3411 3411 3411

No. of countries 30 30 29 29 29 29Adj. R-square .45 .49 .40 .41 .40 .41F value 703*** 629*** 375*** 301*** 321*** 268***

***,**, and * indicate statistical significance at 1, 5, and 10 percent, respectively.

Table VIDeposit Growth, Market Discipline and Deposit Insurance

The estimated model is Deposit Growth[Bank= i, Country=j, Time= t]= α + β1 Liquidityi,j,t-1 + β2 Overheadi,j,t-1 + β3 Inflation j,t +β4 Growthj,t + β5 Gnp/capj,t + β6 AssetSizei,j,t-1 + β7Deposit Insurance j,t + εi,j,t.. The dependent variable is the percentagegrowth in real deposits. Liquidity is defined as liquid assets to total assets. Overhead is personnel expenses and othernon-interest expenses over total assets. Inflation is the annual inflation rate of the GDP deflator. Growth is the annualgrowth rate of real GDP per capita. Gnp/cap is the real Gnp per capita in $10,000. AssetSize is log transformation oftotal assets to gdp. Deposit Insurance is a dummy variable that takes on a value of on if there exists an explicit depositinsurance scheme. Bank data are from the BankScope data base of Fitch IBCA, and macro data are from the IMF’sInternational Financial Statistics. The sample period is 1990-1997. For each specification, column b also includes aninteraction term of the Liquidity variable with the deposit insurance dummy. In specification (2) Liquidity is replacedby its predicted value from the regression of Liquidity j,t = α + β1 Overheadi,j,t+ β2 Short Term Debt/Total Debt j,j,t + β3Inflation j,t + β4Growthj,t + β5 Gnp/capj,t t + β6 Government ratej,t + β7 Reserve Ratei,t-1 + εi,j,t. Reserve rate is given byreserves divided by deposits. In Specification (3) predicted Interest, interest expense to interest paying debt minus thegovernment rate, is added to specification (2). Interest is the predicted value obtained from the regression Interest j,t =α + β1 Overheadi,j,t-1 + β2 Short Term Debt/Total Debt j,t-1 + β3 Inflation j,t + β4Growthj,t + β5Gnp/capj,t t + εi,j,t. White’sheteroskedasticity-consistent standard errors are given in parentheses.

(1.a) (1.b) (2.a) (2.b) (3.a) (3.b)

Liquidity -.085**(.032)

-.071(.068)

-.103(.126)

.009(.265)

.370***(.132)

.372(.251)

Overhead .735***(.192)

.780***(.194)

.412**(.214)

.475***(.219)

4.752***(.938)

4.716***(.952)

Inflation -.273***(.091)

-.265***(.092)

-.153(.130)

-.147(.132)

9.804***(2.175)

9.624***(2.218)

Growth 1.124***(.181)

1.023***(.182)

1.089***(.243)

1.005***(.245)

-3.867***(1.048)

-3.843***(1.066)

Gnp/cap .022***(.004)

.018***(.004)

-.023***(.004)

-.020***(.005)

-.111***(.019)

-.107***(.019)

AssetSize .002(.002)

.001(.002)

-.001(.001)

-.001(.002)

-.000(.002)

-.0012(.002)

Interest 16.846***(3.610)

16.534***(3.692)

Deposit insurance -.025(.022)

.007(.059)

-.014(.059)

Liquidity xDeposit Insurance

-.021(.076)

-.150(.272)

-.023(.274)

No. of obs. 5014 5014 4351 4351 4351 4351

No. of countries 51 51 43 43 43 43Adj. R-square .04 .04 .04 .04 .05 .05F value 35*** 27*** 29*** 23*** 31*** 24**

***,**, and * indicate statistical significance at 1, 5, and 10 percent, respectively.

Table VII. Deposit Interest Rates, Market Discipline and Deposit Insurance Design Features

The estimated model is Interest [Bank= i, Country=j, Time= t]= α + β1 Liquidityi,j,t-1 + β3 Overheadi,j,t-1+ β4 Short Term Debt/Total Debti,j,t-1 + β5 Inflation j,t+ β6 Growthj,t + β7Gnp/capj,t +β8Deposit Insurance j,t + β9Deposit Insurance j,t x Liquidityi,j,t-1 + β10Insurance design feature j,t + β11Insurance design feature j,t x Liquidityi,j,t-1 + εi,j,t . Log transformations ofvariables are taken. The dependent variable is the ratio of interest expense to interest paying debt minus the Government rate. Liquidity is defined as liquid assets to total assets.Overhead is personnel expenses and other non-interest expenses over total assets. Short Term Debt/Total Debt is customer and short term funding to total interest paying debt.Inflation is the annual inflation rate from the GDP deflator. Growth is the annual growth rate of real GDP per capita. Gnp/cap is the real GNP per capita. Deposit Insurance is adummy variable that takes a value of one if an explicit deposit insurance scheme exists. In each specification a deposit design feature and its interaction with Liquidity is alsoincluded. Design features are defined as follows: Co-insurance: 0 if implicit, 1 if explicit with co-insurance, 2 if explicit with no co-insurance. Explicit coverage limit: 0 if implicit,statutory coverage limit divided by Gdp/capita if explicit. Coverage of foreign currency deposits: 0 if implicit, 1 if explicit and foreign currency deposits are not covered, 2 if theyare covered. Coverage of interbank deposits: 0 if implicit, 1 if explicit and there is no coverage, 2 if there is coverage. Type of fund: 0 if implicit, 1 if explicit but unfunded, 2explicit, unfunded but funds are callable ex-post, 3 if explicit and funded ex ante. Source of funding: 0 if implicit, 1 if explicit and funded by banks only, 2 funding by banks andgovernment, 3 if funded by government only. Fund management: 0 if implicit, 1 if explicit and private management, 2 private-public joint management, and 3 if managed publicly.Individual dummy variables for joint (private) management take the value 1 if management is joint (private) and zero otherwise. Membership: 0 if implicit, 1 if explicit andcompulsory, 2 explicit but voluntary. Only the estimated coefficients for Liquidity and deposit insurance variables are reported below. Bank data are from the BankScope data baseof Fitch IBCA, and macro data are from the IMF’s International Financial Statistics. The sample period is 1990-1997. For each specification Panel A presents OLS estimates andPanel B presents between estimates using averaged bank data for the sample period. Heteroskedasticity consistent standard errors are given in parentheses.

Panel A: OLS estimatesIs there Co-insurance?Implicit (0)No (1)Yes (2)

ExplicitCoverageLimitImplicit (0)or coverage/gdp per cap.

Are ForeignCurrencyDepositsCovered?Implicit (0),No (1), Yes(2)

AreInterbankDepositsCovered?Implicit(0), No (1),Yes (2)

Type of Fund:Implicit (0),Unfunded (1)Unfunded butCallable (2),Funded (3)

Source of Funding:Implicit (0), Banksonly (1), Governmentand Banks (2),Government only (3)

FundManagementDummy:Implicit (0),Private (1),Joint (2),Public (3)

IndividualDummyVariables forJoint, andPrivateManagement

Membership:Implicit (0)Compulsory (1)Voluntary (2)

Liquidity -.195***(.032)

-.148***(.028)

-.183***(.031)

-.178***(.031)

-.188***(.031)

-.189***(.032)

-.189***(.032)

-.189***(.032)

-.195***(.032)

Dep. Ins. -.021**(.008)

.004(.007)

-.004(.007)

.007(.007)

.040***(.009)

.026***(.008)

-.003(.005)

-.021***(.006)

-.024**(.006)

LiquidityxDep. Ins.

.238***(.037)

.047(.028)

.290***(.041)

.034(.032)

.039(.034)

.102***(.032)

.125***(.032)

.240***(.035)

.223***(.039)

DesignElements

.008*(.004)

-.006***(.001)

-.006***(.002)

-.017***(.003)

-.020***(.003)

-.022***(.003)

-.006***(.001)

(J) -.003(.002)

(P) .013***(.002)

.012***(.003)

LiquidityxDesignElements

-.030***(.011)

.040***(.005)

-.055***(.011)

.121***(.016)

.068***(.010)

.067***(.011)

.037***(.005)

(J) -.087***(.012)

(P)-.076***(.010)

-.013(.011)

N, country 4852, 28 4291, 23 4972, 29 4557, 28 4972, 23 4972, 29 4972, 29 4972, 29 4972, 29Adj. R-sq. .50 .55 .51 .52 .51 .51 .50 .51 .50F value 477*** 519*** 517*** 494*** 511*** 511*** 503*** 428*** 494***

***,**, and * indicate statistical significance at 1, 5, and 10 percent, respectively.

Panel B: Between EstimatesIs there Co-insurance?Implicit (0)No (1)Yes (2)

ExplicitCoverageLimitImplicit (0)or coverage/gdp per cap.

Are ForeignCurrencyDepositsCovered?Implicit (0),No (1), Yes(2)

AreInterbankDepositsCovered?Implicit(0), No (1),Yes (2)

Type of Fund:Implicit (0),Unfunded (1)Unfunded butCallable (2),Funded (3)

Source of Funding:Implicit (0), Banksonly (1), Governmentand Banks (2),Government only (3)

FundManagementDummy:Implicit (0),Private (1),Joint (2),Public (3)

IndividualDummyVariables forJoint, andPrivateManagement

Membership:Implicit (0)Compulsory (1)Voluntary (2)

Liquidity -.305***(.058)

-.234***(.055)

-.278***(.057)

-.268***(.056)

-.286***(.056)

-.286***(.056)

-.290***(.057)

-.290***(.057)

-.303***(.058)

Dep. Ins. -.017(.014)

.002(.013)

.017(.013)

.032**(.013)

.085***(.018)

.064***(.015)

.012(.012)

-.033***(.010)

-.044***(.011)

LiquidityxDep. Ins.

.355***(.061)

.218***(.060)

.373***(.063)

.020(.069)

.021(.237)

.105*(.066)

.160**(.062)

.371***(.056)

.362***(.062)

DesignElements

.001(.007)

-.005***(.002)

-.020***(.004)

-.038***(.005)

-.038***(.005)

-.046***(.005)

-.015***(.002)

(J) .009(.005)

(P) .031***(.004)

.029***(.006)

LiquidityxDesignElements

-.034*(.018)

.015**(.006)

-.052**(.021)

.202***(.030)

.114***(.019)

.124***(.021)

.069***(.011)

(J) -.114***(.024)

(P)-.138***(.021)

-.039*(.011)

N, country 1520, 28 1357, 23 1548, 29 1431, 28 1548, 23 1548, 29 1548, 29 1548, 29 1548, 29Adj. R-sq. .68 .72 .70 .70 .70 .70 .69 .69 .68F value 320*** 358*** 360*** 340*** 355*** 360*** 342*** 288*** 332***

***,**, and * indicate statistical significance at 1, 5, and 10 percent, respectively.