1
Master thesis for MSc Finance
The effect of deposit insurance on bank risk and banking crises
Supervisor: Dr. M.A. Lamers
2 Abstract
3 1. Introduction
In the recent financial crisis it has become evident that the banking sector is one of the Achilles heels of the European economy. The failure of banks, and the threat of failure by others, were the start of the financial crisis which fully aggravated in September 2008 in the United States. The bail out of the two American financial institutions Fannie Mae and Freddie Mac, followed by the failure of Lehman Brothers, was the starting point of the crisis which would bring the global economy to the edge of collapsing. The crisis made its appearance in Europe too and many countries relied upon deposit insurance to recover losses for depositors of failed and/or nationalized banks, often at great costs. With a large role for banks in their economies and high international integration, Europe could not avoid the catastrophe. Nonetheless, certain banks in the European countries were more financially sound than others and, from a perspective of systemic risk, some banks contributed more risk to the financial sector than other banks did. This paper aims to explain these differences in bank risk. For years, and in some cases even decades, governments and regulators have used deposit insurance as an instrument to maintain the stability of banks and the financial sector. However, there can be contradictions in measures and the goal they try to achieve. For example, some design features might improve depositor confidence and reduce the chance on bank runs but in the end increase bank risk because of distortions in market discipline. In addition, if banks are guaranteed to be bailed out through government intervention or through a deposit insurance scheme, there is an incentive for banks to take on extra risk or mismatch maturities of assets and liabilities (Diamond & Dybvig, 1983). This incentive is valid for all banks and therefore does not only create additional bank risk but also increased systemic risk since banks are induced to correlate their risk profile with other banks. Whether certain aspects of the deposit insurance scheme effect bank risk is one of the focus points this paper.
4 governments and regulators, it is of importance to understand how their policies may contribute to a more stable financial sector, and to the prevention and cost minimization of banking crises. Diamond & Dybvig (1983) argue that banking regulation and deposit insurance is desirable and is comparable to restrictive covenants.Before going into the economic arguments concerning the way deposit insurance schemes influence our topics of interest, an introductory rational is discussed to understand the context of deposit insurance schemes in the financial system.
With deposit insurance schemes, eligible depositors are guaranteed to be reimbursed in case of bank failure. Depending on policies, the type of deposits that are covered vary as does the amount of funds that is guaranteed. The latter can be done by setting a payout limit, bounding the reimbursement to a maximum or demanding coinsurance where the depositor takes a loss on the deposit first. When the deposits are insured, consumers profit from deposit insurance because it diminishes the risk of holding their deposits. This however, although regarded as positive, is not the main purpose of deposit insurance schemes. The European Commission describes the main objective of deposit guarantee schemes as maintaining financial stability (European Commission, 2010). To achieve this goal, several sub-objectives are pursued including enhancing the internal market, strengthening depositor confidence, ensuring a level playing field between banks, financing of deposit insurance scheme by banks and not taxpayers, and protecting a portion of depositor wealth in order to avoid bank runs (European Commission, 2010). In pursuing the goal of financial stability, there is a balance to be maintained between regulations and government interventions, and respecting market discipline. Furthermore, there can be contradictions in measures and the goal they try to achieve. For example, some design features might improve depositor confidence and reduce the chance on bank runs but in the end increase bank risk because of distortions in market discipline. These arguments will be examined in depth in the next chapter.
5 able to liquidate assets fast enough) to return to depositors if they withdraw en masse. Kleftouri (2014) concludes that “sudden deposits’ withdrawals can force even healthy banks to liquidate many of their assets at a loss and subsequently fail.” In the absence of deposit insurance, McCoy (2008) links bank runs to the prisoner dilemma that damages depositors through distributive justice. This entails depositors in the front of the line being returned their full deposits whereas depositors in the back are only partly compensated. In addition, banks must liquidate assets are fire sale prices to come up with funds when depositors withdraw en masse (McCoy, 2008). Deposit insurance encourages financial stability by functioning as a way to limit government guarantees. Kleftouri (2014) argues that during crises, governments and regulators are pushed to expand coverage of the deposit insurance. This increased coverage comes with great potential cost for taxpayers. A well-defined deposit insurance scheme with clear limits provides politicians and regulators reluctant to give in to this pressure a point of reference. The third way in which deposit guarantee schemes encourage financial stability is by supporting resolution mechanisms. A deposit insurance scheme will likely, though not necessarily, have guidelines on how to deal with bank failures. Kleftouri (2014) states “[…] a well-designed explicit deposit insurance system can facilitate an organized process for dealing with bank failures, enhancing the resolution mechanisms.” This leaves less room discretionary actions of politicians and regulators and clearly state the rules of the game. Lastly, deposit insurance schemes encourage financial stability by creating a level playing field. In the absence of deposit insurance, depositors prefer larger banks since these provide more security against losing their deposits. This may be due to their size and/or the assumption that governments cannot afford to let a large bank fail because of large economic consequences (Kleftouri, 2014). As a result, these large banks become too-big-to-fail. The too-big-to-fail argument is examined more in depth in the next chapter.
6 paper is related to their study but differs on various aspects. Firstly, I constructed a new dataset on deposit insurance for the year 2006 using existing sources supplemented with data obtained through personal communication. Additionally, this paper also includes research into the relationship between deposit insurance and the materialization of banking crises and the associated costs.
In this paper, I show that for the total sample period most of the design features of the deposit insurance significantly affect the liquidity of banks. For other indicators there is no convincing effect of deposit insurance variables. The reason why so few design features do not show a significant effect becomes apparent when separating the sample period into a pre-crisis period and a pre-crisis period. The signs of coefficients change during the pre-crisis. I show that especially systemic risk and the risk indicator for liquidity are affected by the deposit insurance design features during the crisis. Pre-crisis, this is mostly true for profitability and asset quality. The signs of coefficients vary among the categories for generosity as well as credibility and is different for the various risk indicators. This makes general conclusions about results challenging: therefore effect of the variables are considered individually.
Concerning banking crises, the results do not show a significant relationship with deposit insurance. Regarding output loss, there appears to be negative effect of coinsurance. Also private administration of the deposit insurance scheme is association with a less severe crisis.
7 2. Related literature
In this chapter, the relationship between deposit insurance and bank risk is examined first. Existing literature shows a positive association, the different arguments concerning this relationship are presented in the first paragraph. The second paragraph presents arguments from earlier studies on how differences in deposit insurance schemes affect bank risk. Furthermore, the effect of deposit insurance on systemic risk is examined. This chapters end with a concise overview of control variables that have proven to be relevant for the topic of this paper.
2.1. Deposit insurance and bank risk
There is substantial agreement in the literature about deposit insurance schemes increasing bank risk. Although a majority of the papers argue that deposit insurance has a positive effect on bank risk through decreased market discipline and increased moral hazard issues, there are studies that find a negative relationship. This negative relationship is due to the fact that with the introduction of an explicit deposit insurance scheme, implicit promises are replaced which in certain cases are assumed to be superior in terms of coverage.
The reduction in market discipline as a result of the introduction of a deposit insurance scheme is extensively documented in the literature. As a result of the introduction of deposit insurance schemes, depositors are assured that losses in case of bank bankruptcy will be compensated for. With no deposits to lose, protection of their funds decrease the incentive for depositors to “monitor and police bank risk taking” (Demirgüç-Kunt & Kane, 2002). Consequently, banks lose feedback from the market and will not be punished by depositors for taking on irresponsible high risk projects. Overly generous depositor protection, mainly deposit insurance, increase excessive bank risk taking leading to bank failure (Demirgüç-Kunt & Huizinga, 2004).
8 between market discipline and regulatory discipline. Ashcraft, (2008) shows that mixing regulatory capital and non-regulatory capital can also help to mitigate moral hazard issues through these restrictive covenants. The trade-off between regulatory capital and market capital is not particularly relevant for the banks studied in this papers because the sample only includes countries where membership to the deposit insurance scheme is compulsory.
Moral hazard problems arise due to implementation of a deposit insurance scheme. The moral hazard problem covers the issue of banks benefitting from taking risks but being isolated from large downsides by the deposit insurance and, as a consequence, take on riskier projects. Indeed, when the bank goes bankrupt, claims from depositors will be transferred from the bank to the deposit insurance. This gives the bank incentives to increase bank risk at the expense of funders of the deposit insurance scheme (Demirgüç-Kunt & Kane, 2002). This effect does not only hold for retail banks with deposits from individuals and small businesses but also for commercial banks who take their deposits from large(r) firms (Shiers, 1994). It is interesting to note that there seems to be a destabilizing effect during normal times leading to moral hazard problems, but a stabilizing effect during the recent crisis years (Anginer et al., 2014). In times of crisis, banks can be faced with limited investment opportunities and constricted funding possibilities which bounds them from excessive risk taking. The stabilization effect of deposit insurance can then dominate the negative moral hazard effects by enhancing depositor confidence and decrease the likelihood of bank runs.
9 banks are less inclined to take risk. The sample of their research includes 119 countries across multiple continents, this differs from the sample studied in this paper which only includes European countries.
2.2. Effect of differences in deposit insurance schemes on bank risk
Previous studies show that differences in deposit insurance schemes influence bank risk. Especially difference in generosity of the deposit insurance, it’s governance and credibility, and the applicable entry hurdles seem to have varying effect on bank risk.
Concerning the generosity of a deposit insurance scheme, regulations can separate deposits that fall within the scope of the deposit insurance – and are therefore insured – and those that are not. It is important that policy makers pay close attention to setting limits to coverage, so that parties with large funds are still at risk, causing them to act as a monitor to protect their funds. These parties include large depositors, subordinated debt-holders and other banks. Especially the latter plays an important role as monitors because of their high interbank deposits and their superior knowledge about banking and risk management. Further aspects that determine how deposit insurance influences bank risk include coinsurance, where losses are not only borne by one party (Demirgüç-Kunt & Kane, 2002). This spreads the risk between the insurer and the insured, and makes monitoring the bank a shared incentive. High generosity is associated with higher bank risk. These features impact the expectation of depositors about the amount of funds is received when the bank fails (Cull et al., 2005). In their research, generosity of a deposit insurance scheme is defined as, among other things, a combination of coverage per depositor, whether interbank deposits and foreign currency deposits are covered, whether the funding comes from government, banks or both, whether the program requires a deductible, and whether the program is managed by the government, the private-sector or both. The categorization is followed to a certain extent in this paper.
10 because it would cause them to pay, through membership contributions, for weaker banks going bankrupt. Making membership compulsory would prevent these kind of adverse selection problems and, additionally, increase the number of banks in the pool (Demirgüç-Kunt & Kane, 2002). Entry hurdles, under the definition of Cull et al. (2005), include variables covering whether participation is voluntarily or compulsory, whether the insurance is funded ex ante or ex post, the percentage of premium paid to insured deposits and whether premium payments are risk-adjusted.1 These entry hurdles can be a part of a deposit insurance scheme to combat adverse selection problems. Making membership compulsory, for instance, ensures that every bank is incorporated into the scheme and eliminates the chance of attracting mainly the weaker banks. In addition, it can serve as a means of making member banks bear the cost of risk they are contributing to the system, for example through raising the premiums with the risk of a bank’s portfolio (Cull et al., 2005).
It is important that the scheme has credibility to the extent that depositors rely on the promised insurance, as individuals and firms make decisions based on expected outcomes. This includes the way in which the scheme is funded and governed. Since the private-sector is usually thought of as better monitor than governments officials – and banks being “apt to solicit better information with which to monitor one another”, involving the private-sector in managing the scheme has been shown to be helpful. (Demirgüç-Kunt, Kane, & Laeven, 2008).
Based on the mentioned literature, I hypothesize the following: Hypothesis 1: Deposit insurance influences bank risk
Hypothesis 1a: Deposit insurance does not influence bank risk
As mentioned, deposit insurance can be partitioned into two categories in this paper. Variables representing the generosity of a deposit insurance scheme are examined as well as those representing the credibility of a scheme.
1
11
2.3. Effect of deposit insurance on systemic risk
Bank risk measures that focus on the risk of any one bank to fail is inadequate when studying crises. In the recent financial crisis, arguments on whether to bailout banks or not were often not only about one bank or one institution but what effect a bankruptcy of one bank would bring to other banks and whether it would eventually create a crisis that is systemic. A crisis is systemic if the failure of one bank works as a contagion causing failure of many banks, or if many banks fail together (Acharya, 2009). Anginer et al. (2014) therefore state that the focus should be on a bank’s contribution to the risk of the financial system as a whole.
One of the interesting paradoxes in deposit insurance thinking is that governments, especially those in advanced economies and in many developing countries, introduce a deposit insurance scheme to build stability in the countries’ financial system. The deposit insurance will give depositors less reason to claim their funds from the bank – which might lead to bank runs. Previous research indicates something different, however. Introduction of deposit insurance is aimed to bring stability to the entire system and not as such for one bank – unless failure of this bank threatens the health of the financial system. Nevertheless, literature shows that deposit insurance does not necessarily have this desired effect; it actually increases systemic risk. Besides the argument mentioned that are applicable to individual banks, there are reasons why deposit insurance schemes increase systemic risk, including incentivizing banks to correlate their risks.
12 banks (Acharya, 2009). In conclusion, deposit insurance may reduce depositor runs but it creates another risk: decreasing bank stability by incentivizing banks to take (correlated) risk.
Based on the literature, I hypothesize the following:
Hypothesis 2: Deposit insurance influences systemic risk
Hypothesis 2a: Deposit insurance does not influence systemic risk
2.4. Deposit insurance and other country level variables
The effect deposit insurance has on bank risk might differ across countries because of certain country level variables. To combat the potential omitted variable problem, control variables are included. In this paper, Anginer et al. (2014) is followed and the following control variables are included: the natural logarithm of GDP per capita, the natural logarithm of population, a variable that measures imports plus exports divided by GDP, a crisis dummy, and a variable that divides the market capitalization by GDP. These variables are included to capture effects of (1) economic development of country, (2) country size, (3) global integration, (4) whether the country is in crisis and (5) financial development and structure.
Cull et al. (2005) conclude that rule of law is an essential element which, if lacking, can cause generous deposit insurance schemes to have a negative effect on the stability. This is further support by evidence from Demirgüç-Kunt & Kane (2002) who argue that the relationship between deposit insurance and bank risk is influenced by whether there is a institutionally weak or strong environment. Especially countries with weak institutions benefit from deposit insurance schemes in the short term but are being damaged in the longer term because of the undermined market discipline.
13 3. Sample and data
In this chapter, the data underlying this research is examined. The first paragraph deals with the data source of bank level variables, including restrictions that are applied on the data. The second paragraph elaborates on deposit insurance data and how this information is collected, including self-collected data through personal communication. This chapter end with a description of the country level variables. These variables include control variables as well as data on banking crises and output losses.
3.1. Sample and bank level variables
For this study, bank level data is collected from two sources: Bureau van Dijk’s BankScope database and the Volatility Institute created by the New York University Stern School of Business. Data from Bankscope provides information on bank risk while data from the Volatility Institute provides a measure for systemic risk. The sample covers a total of 363 banks from 31 European countries. Bank level data in the years 2006 – 2009 is selected to cover a pre-crisis period (2006, 2007) and a crisis period (2008, 2009). Several restrictions have been applied in construction the sample. Firstly, European countries with no explicit deposit insurance scheme are excluded. Secondly, unconsolidated accounts are excluded. There are two reasons underlying this decision: to develop a comprehensive overview of a company’s operations and to have consistency between the banks and the deposit insurance regime they are subjected to. A bank will be covered by the deposit insurance scheme of the country where it is headquartered, a foreign branch will therefore not fall under the scheme of the country in which it does it business, but in the country where the headquarter is located. Lastly, banks that did not report on the bank risk measures are excluded. This possibly leads to a bias towards larger firms.
14 recommendation for national central banks and the European central bank, consistency in risk indicators is therefore desirable.
Asset quality is proxied by the ratio of loan loss provision to net interest revenue. The tier 1 ratio serves as a measure for capital adequacy. Profitability is measured by the return on average assets. The fourth and last indicator, liquidity, is proxied by the ratio of net loans to deposits & short term funding. These indicators of risk are accounting based. Accounting based data is considered to be backward looking (Apergis & Eleftheriou, 2012). This presents a drawback, changes in these risk indicators are delayed. To account for this drawback, both pre-crisis and crisis period comprise two years of data instead of just one year. In addition, a forward looking measure is used, namely the marginal expected shortfall.
As a result of the financial crisis, regulators are more and more interested in the stability of the financial system as a whole and not only in the stability of individual banks (Anginer et al., 2014). During the crisis, we have seen that it is not possible for certain banks to go bankrupt or experience major distress without causing a system failure and/or seriously damage to the economy of a country or even multiples countries. As Acharya, Pedersen, Philippon, & Richardson (2010) show, the Marginal Expected Shortfall (MES) as a measure for the contribution of systemic risk by an individual bank, has high explanatory power and predicts a firm’s impact on the recent crisis. For this reason, the MES is used in this paper as a measure of systemic risk.
15
3.2. Deposit insurance variables
16 Table 1.
Countries in dataset
Austria Finland Malta Spain
Belgium France Norway Switzerland
Bulgaria Germany Poland Ukraine
Cyprus Hungary Sweden
Czech Republic Italy Slovenia
Estonia Latvia Slovakia
a) more countries have been contacted and established contact with, but eventually dropped out of our sample due to limited availability of bank level data in BankScope.
The variables used in this study are categorized into two main groups: coverage and credibility. Cull et al. (2005) there is a third group to consider, namely entry hurdles. The variables in this category are not applicable to the sample being studied in this paper, however. All countries have a deposit insurance scheme in place and for the great majority, membership is compulsory for domestic banks. Coverage of interbank deposits has been excluded for our study because of the lack of variation between countries in this sample.
3.3. Country level variables
17 (2012), the World Bank Website, and the Global Governance Indicators as discussed by Kaufmann, Kraay, & Mastruzzi (2011).
Concerning country level control variables, banking concentration is defined as the share of the three largest commercial banks in a country of the total commercial banking assets. The log of a country’s population is taken to capture the effects of country size. The ratio of stock market capitalization to GDP is computed to control for effects in financial development and structure. The variable controlling for economic development consists of the natural logarithm of GDP per capita. Global integration is defined as the ratio of imports plus exports to GDP. The indicators for the institutional environment from Global Governance Indicators and include (1) government effectiveness, (2) control of corruption, (3) rule of law, (4) regulatory quality, (5) political stability and (6) accountability. Since Cull et al. (2005) show that only the indicator of rule of law is relevant for research on deposit insurance, only this variable is used as control variable.
4. Methodology
For testing the hypotheses, the Ordinary Least Squares method is used to estimate regressions which will be elaborated on in this paragraph. In case of estimating a regression where the dependent variable has a binary value, the logit model is applied. Binary dependent variables do not allow for OLS because there cannot be a normal distribution, binary variables can only take the values 0 or 1.
4.1. Deposit insurance and bank risk
To test hypothesis 1, the relationship between deposit insurance and bank risk is examined. For this reason, the following regressions is estimated:
Profitabilityi,t = α + β1 x pay_lim2006 + β2 x coin2006 + β3 x for_cur2006 + β4 x risk_adjust_p2006
+ β5 x fund_ex2006 + β9 x fund_priv2006 + β10 x adm_priv2006 + β11 x log(popt-1) + β12 xlog(gdp t-1) + β13 x bank_conc t-1 + β14 x glob_int t-1 + β15 x cap_gdp t-1 + β16 x ins_rul_lawt-1 +
18 Streamlining this equation reduces the regression to:
(1) profitabilityi,t = α + β1deposit insurance2006 + β2 Xi, t-1 + µi + η t + εi,t
Where X is a vector of control variables.
The coverage category is an indication of the intensive margin of deposit recovery, and includes information about how much of potential losses are expected to be retrieved through the deposit insurance scheme. The payout limit is taken as indication of how much the deposit insurance will distribute in case of bank failure. In the analysis, the log of this payout limit is taken to scale this number and assess what the impact is of 1% change. A dummy variable is included to evaluate the impact of coinsurance, 1 indicating presence of coinsurance and 0 in absence. In case of coinsurance, depositors will not receive recover their full deposits but will have to take a loss themselves first. The height of this coinsurance differs among countries. When the DI covers foreign currency deposits, a dummy is set to 1 and 0 otherwise. The coverage of foreign currency deposits entails covering deposits denominated in any other currency than the official domestic currency.
The credibility category comprises variables on the extensive margin, i.e. whether there is any expected payout at all to recover potential losses. A dummy variable that covers timing of funding is set to 1 in case of ex-ante funding and 0 in case of ex-post funding. In case of private funding, the funding type dummy is set to 1 and 0 otherwise. Privately funded schemes may easily run short of funds in case of systemic failures but encourages peer monitoring among institutions. The administration dummy is set to 1 if the deposit insurance scheme is administrated privately and 0 otherwise. The effect of risk adjustment premiums is also included by setting the dummy for this as 1 if banks make contributions to the fund based on their risk and 0 otherwise.
The variables concerning deposit insurance characteristics are bundled as well as country- and bank control variables in the following regressions.
(2) asset qualityi,t = α + β1deposit insurance2006 + β2 Xi, t-1 + µi + η t + εi,t
(3) capital adequacyi,t = α + β1deposit insurance2006 + β2 Xi, t-1 + µi + η t + εi,t
19 As discussed, systemic risk has received more attention in recent years as a results of the financial crisis. For this reason, not only bank risk as measured by the financial sound indicators is tested but also the marginal expected shortfall. The MES is an indicator of how much risk a bank adds to the financial system.
(5) marginal expected shortfalli,t = α + β1deposit insurance2006 + β2 Xi, t-1 + µi + η t + εi,t
4.2. Deposit insurance pre-crisis and crisis period
One of the main aims of this paper is to study the effects of deposit insurance and bank risk during the recent financial crisis. Consequently, the sample period is split into a pre-crisis period (2006, 2007) and a crisis period (2008, 2009). I examine the effect of variations in design of deposit insurance on bank risk and systemic risk during each of these periods and compare the results. As indicators for bank risk, the financial soundness indicators are used. Systemic risk is tested through the MES.
4.3. Deposit insurance and banking crisis
Bank risk itself does not necessarily come with major costs, a country-wide banking crisis usually does. Therefore, it is useful to not only uncover the effect of deposit insurance on risk but also if this eventually materializes into a banking crisis. Determining which aspects of deposit insurance schemes have significant effect on a banking crisis to come into existence, is a point of interest for governments and regulators. A logit model is estimated in this case because of the binary property of the dependent variable.
(6) banking crisisi = α + β1deposit insurance2006 + β2 Xi, 2006 + εi,t
20 sample (inherent to the limited amount of countries in Europe), compromises are made to be able to estimate the regression properly: only two variables are included to capture the effect of generosity and two for credibility. High correlating control variables are dropped from the equation. Variables capturing the effect of payout limit and coinsurance represent the generosity category while the credibility category is represented by private funding and administration. As control variables GDP, banking concentration, global integration and capitalization of stock market have remained. Other control variables have been dropped, due to our low sample size. I choose to drop variables that showed greatest correlation with other control variables. The correlation matrix of the full sample is presented in table 9. The correlations of the sample after selection is presented in table 10.
4.4. Deposit insurance and severity of banking crises
Not only the fact that a banking crisis materializes is interesting for our study, also the severity of a banking crisis in the sample period is a topic of interest. Whether certain characteristics of deposit insurance schemes can predict the presence of a banking crisis or not, it is still relevant to test for a relationship with the severity of the banking crisis. Perhaps a banking crisis might not be prevented by deposit insurance schemes but the severity of such a crisis is related. To test this, the following equation is estimated:
(7) output lossi = α + β1deposit insurance2006 + β2 Xi, 2006 + εi,t
For estimating this equation, the same variables are used that have been described in this chapter and the previous. Additionally, Laeven & Valencia (2012) define output loss as the cumulative loss in income relative to a pre-crisis trend. The output loss variable is a ratio of this loss to a country’s GDP.
4.5. Estimation technique
21 time-invariant, however. I am interested in how deposit insurance schemes - as they were constituted before the crisis - impact the independent variables in subsequent years. A drawback of this set-up is that estimating the regressions using cross-sectional fixed-effects is not possible, since using these fixed effects require variables changing over time. For this reason, cross-sectional fixed-effects nor cross-sectional random-effects are used. Firm-specific characteristics are used to differentiate banks, mainly the log of total assets. Time fixed-effects are used to account for unobserved changes in the variables over the years. The estimation results are based on robust standard errors, using Huber/White estimators of variance. The control variables are lagged by one year to mitigate any reverse causality problems.
5. Results
4.1. Descriptive statistic deposit insurance and bank risk
22 and administered privately dummy variables show more variation in the sample. The descriptive statistic of the control variables are given in Table 2.
Table 2.
Descriptive Statistics
Mean Median Maximum Minimum SD N
Net income / total average assets (in %)
0.763 0.654 3.449 -3.279 0.811 361
Loan loss provision / net interest revenue (in %)
20.756 16.000 182.267 -18.921 23.942 353
Tier 1 capital / total assets (in %)
9.786 9.040 32.740 5.130 3.453 331
Net loans / deposits & short term funding (in %)
92.838 86.750 663.155 9.142 49.307 356
Marginal Expected Shortfall
2.741 2.620 6.380 0.210 1.238 88
Log (Pay Limit) 10.491 10.127 12.407 7.728 0.778 361
Coinsurance 0.353 0.000 1.000 0.000 0.478 361 Foreign Currency 0.918 1.000 1.000 0.000 0.274 361 Risk-Adjusted Premiums 0.275 0.000 1.000 0.000 0.447 361 Funded Ex-Ante 0.583 1.000 1.000 0.000 0.494 361 Privately Funded 0.369 0.000 1.000 0.000 0.483 361 Privately administered 0.363 0.000 1.000 0.000 0.481 361 Log (Population) 16.852 17.453 18.222 12.915 1.165 361 Log (GDP) 27.274 26.958 28.912 22.580 1.232 361 Banking Concentration 68.779 66.403 99.868 28.802 16.864 361 Trade / GDP 0.817 0.747 1.755 0.476 0.326 361 Stock Market Capitalization / GDP 70.203 55.359 281.388 5.443 48.721 361 Institutional environment 1.220 1.427 2.000 -0.774 0.631 361
Log (Total Assets) 18.013 17.795 21.674 14.207 1.829 361
23 the years 2006 – 2009. Net income / total average assets represents the profitability, loan loss provision / net interest revenue the asset quality, tier 1 capital / total assets the capital adequacy and net loans / deposits & short term funding the liquidity of banks. These variables are expressed in percentages. The marginal expected shortfall is a measure for systemic risk. Log (Pay Limit) is the natural logarithm of the payout limit of a deposit insurance scheme. Coinsurance, foreign currency, risk-adjusted premiums, funded ex-ante, privately funded and privately administered are dummy variables. Log (Population) is the log value of population in millions. Log (GDP) is the log value of GDP in nominal constant US 2000 dollars. Banking concentration is the share of the three largest commercials banks in a country in the total commercial banking assets, expressed in percentages. Trade / GDP is calculated as (imports + exports) / GDP. Institutional environment is a variable that ranges from -0.774 to 2 and is an indication of how development of rule of law in a country. Log (Total Assets) is the log value of a bank’s total assets.
24 Table 3. Correlation matrix PROF ASS_ QUA CAP_
ADE LIQ MES LOG (PAY_ LIM) COIN FOR_ CUR RISK_ ADJU ST_P FUN_ EX_ ANTE FUND_ PRIV ADM_ PRIV LOG (POP) LOG (GDP) BANK_ CONC GLOB_ INT CAP_ GDP INS_ RUL_ LAW LOG (TOT_ ASS) PROF 1.00 ASS_QUA -0.37 1.00 CAP_ADE 0.11 -0.03 1.00 LIQ -0.01 0.11 -0.25 1.00 MES -0.25 0.30 0.21 -0.04 1.00 LOG(PAY_LIM) -0.12 -0.11 -0.17 0.23 -0.10 1.00 COIN 0.05 -0.08 0.15 -0.18 0.06 -0.41 1.00 FOR_CUR -0.03 0.07 -0.23 0.22 -0.01 0.32 -0.03 1.00 RISK_ADJUST_ P 0.06 -0.05 -0.18 0.29 -0.11 0.36 -0.38 0.18 1.00 FUND_EX_ANT E 0.20 -0.08 -0.06 0.11 -0.05 -0.44 0.29 0.04 -0.21 1.00 FUND_PRIV -0.17 0.15 0.04 -0.26 0.12 -0.20 -0.01 -0.32 -0.47 -0.10 1.00 ADM_PRIV -0.08 -0.10 -0.08 0.13 -0.04 0.62 -0.31 -0.08 0.41 -0.57 -0.21 1.00 LOG(POP) -0.05 0.10 -0.10 -0.02 0.03 0.19 -0.20 0.45 0.17 -0.41 0.05 0.07 1.00 LOG(GDP) -0.26 0.07 -0.15 0.02 0.11 0.39 -0.31 0.47 0.12 -0.52 0.12 0.21 0.89 1.00 BANK_CONC -0.19 -0.18 0.09 -0.01 0.12 -0.13 0.09 -0.26 -0.20 0.43 0.02 -0.21 -0.64 -0.42 1.00 GLOB_INT 0.02 -0.11 0.24 -0.06 0.10 -0.42 0.67 -0.30 -0.24 0.32 -0.09 -0.24 -0.64 -0.63 0.47 1.00 CAP_GDP -0.05 -0.17 0.16 -0.17 0.09 -0.12 -0.33 -0.47 -0.23 -0.27 0.38 0.01 -0.09 0.10 0.37 0.01 1.00 INS_RUL_LAW -0.34 0.02 -0.01 -0.06 0.12 -0.02 0.10 -0.04 -0.43 0.02 0.34 -0.11 -0.34 0.01 0.65 0.25 0.49 1.00 LOG(TOT_ASS) -0.31 0.21 -0.30 -0.03 0.36 -0.041 -0.14 0.10 -0.15 -0.18 0.32 -0.07 0.24 0.38 0.08 -0.13 0.34 0.38 1.00
25 the share of the three largest commercials banks in a country in the total commercial banking
assets, expressed in percentages. Trade / GDP is calculated as (imports + exports) / GDP. Institutional environment is a variable that ranges from -0.774 to 2 and is an indication of how development of rule of law in a country. Log (Total Assets) is the log value of a bank’s total assets.
5.2. Deposit insurance and bank risk
To test hypothesis 1, a relationship between deposit insurance schemes and bank risk, the impact of deposit insurance features on the indicators for bank risk is estimated. The results of the regression can be found in column 1, 2, 3 and 4 of table 4.
The significant coefficient of the payout limit indicates a positive relationship between the height of the payout limit and liquidity (p < .01). This indicates that a higher payout limit is associated with higher liquidity. Not finding a significant association with profitability, asset quality and capital adequacy is in line with literature. Davis & Obasi (2009) find no results with any of the financial soundness indicators which can be ascribed to the varying effect the variable has during non-crisis years and crisis years. Also coinsurance tested significant for the liquidity indicator (p < 0.1). The coefficient is negative, specifying a negative relationship between the height of coinsurance and liquidity. The coverage of foreign currency deposits has a positive coefficient for profitability (p < 0.01), a negative coefficient for capital adequacy (p < 0.1) and a positive coefficient for the MES (p < 0.01). This indicates that the coverage of foreign currency deposits increases systemic risk. The negative coefficient for capital adequacy is consistent with this, showing that banks become more unstable. On the contrary, coverage of foreign currency deposits is associated with stronger bank profitability. For the remaining indicators, the variables do not show a significant effect. Altogether, variables representing the generosity of the deposit insurance scheme do not show a convincing significant relationship with the majority of the financial soundness indicators.
26 funding, it is important for banks to signal that depositors do not have to worry about retrieving their funds by having larger reserves. In case of ex ante funding, this signal has less importance since the deposit insurance scheme has increased credibility. Both private funding and private administration of the deposit insurance scheme has a significant negative coefficient (p < 0.01) for liquidity. As with ex ante funding, these are variables on the extensive margin of the deposit insurance. A significant negative coefficient indicates a substitution in signaling robustness from liquidity to credibility of the scheme. Private administration furthermore tests positive with profitability (p < 0.01). Risk-adjusted premiums are significantly positively correlated (p < 0.01) with capital adequacy. Banks that pay a premium based on their risk profile show increased capital adequacy. This is a logical consequence of the incentive to lower a bank’s risk and pay a lower premium. Other indicators do not show any significant effects. Overall, and despite some significant effects there is no substantial evidence for a relationship between the credibility of a deposit insurance scheme and the majority of indicators of bank risk for the total period between 2006 and 2009.
Concerning the control variables, a most of variables are significantly related to dependent variables, especially the variables for population, stock market capitalization, total assets and to a lesser extent, banking concentration and the institutional environment variable capturing the effect of decent rule of law. This is in line with previous research and indicates a proper selection of control variables.
27 Table 4. Panel regression 2006 – 2009 (1) (2) (3) (4) (5) LOG(PAY_LIM) -0.033 (0.125) -2.276 (1.569) 0.347 (0.339) 14.802*** (4.003) -0.153 (0.146) COIN -0.041 (0.202) -3.575 (1.569) 0.631 (0.738) -17.689* (9.861) 0.273 (0.177) FOR_CUR 0.709*** (0.038) -1.642 (6.251) -2.529* (1.398) -9.761 (9.478) 1.029*** (0.373) RISK_ADJUST_P 0.058 (0.145) -3.240 (5.252) 1.376*** (0.351) 10.442 (8.899) -0.024 (0.110) FUND_EX_ANTE 0.110*** (0.042) -5.508 (3.499) -0.343 (0.389) -15.406*** (2.863) 0.166 (0.235) FUND_PRIV 0.255 (0.194) -5.508 (4.531) 0.590 (0.487) -30.003*** (1.312) 0.096 (0.125) ADM_PRIV 0.415*** (0.042) -9.801 (6.327) -0.115 (0.364) -19.841*** (1.324) 0.217* (0.117) LOG(POP) 0.398** (0.186) -5.718 (7.996) 2.144** (1.061) 14.298*** (5.368) -0.384*** (0.066) LOG(GDP) --0.289 (0.218) -14.362 (11.338) -1.291 (0.888) -9.542** (4.001) 0.349*** (0.102) BANK_CONC 0.013*** (0.003) -0.626*** (0.184) 0.031*** (0.004) 0.620*** (0.084) -0.007 (0.008) GLOB_INT 0.098 (0.077) -15.328*** (2.962) 2.289** (0.948) 8.796 (10.544) 0.697** (0.286) CAP_GDP 0.003** (0.002) -0.117** (0.054) 0.025*** (0.009) -0.137 (0.107) 0.010** (0.004) INS_RUL_LAW -0.263*** (0.028) 17.000* (8.986) 1.360 (0.998) 16.645* (8.561) -0.477*** (0.121) LOG(TOT_ASS) -0.094*** (0.029) 3.082*** (1.047) -1.380*** (0.110) 2.656*** (0.398) 0.226*** (0.033) Constant 2.080 339.566** 24.639*** -98.391*** -4.200 Observations (2.501) 1443 (156.934) 1409 (2.406) 1321 (26.121) 1423 (2.584) 350 R-squared 0.103 0.253 0.109 0.077 0.425
28 indication of how development of rule of law in a country. Log (Total Assets) is the log value of a bank’s total assets.
5.3. Deposit insurance and systemic bank risk
In order to test hypothesis 2, a regression is estimated to test for a positive relationship between deposit insurance and the marginal expected shortfall. The results of this regression are provided in table 4, column 4.
Of all variables capturing the different design features of deposit insurance schemes, only the variables coverage of foreign currency (p < 0.01) and privately funded (p < 0.1) are significant. They both have a positive coefficient, indicating increased systemic risk in the presence of these two variables. This is no convincing evidence of a relationship between deposit insurance and systemic risk. The results are not in line with earlier research. It is expected this is for the same reason as mentioned in the previous paragraph, differences in sample period. Non-crisis and crisis periods are tested separately later in this chapter. It is interesting to consider the effects of the control variables, though. Except for banking concentration, all control variables test significantly. The population (p < 0.01) and the variable representing the rule of law (p < 0.01) have a negative coefficient, indicating that countries with a larger population are associated with lower systemic bank risk as well as those with a more developed institutional environment. On the contrary, higher GDP (p < 0.01), higher global integration (p < 0.05) and higher capitalization of the stock market (p < 0.05) is associated with higher systemic risk. Similar results are found for a bank’s total assets (p < 0.01).
29
5.4. Deposit insurance and bank risk in non-crisis and crisis period
In the previous paragraphs, the entire period from 2006 until 2009 has been examined, pooling the pre-crisis years and the crisis years. Next, I will consider both periods separately to study the effect of deposit insurance features in two periods distinguished by the recent financial crisis. The effect of each design feature that has a significant effect on bank risk pre-crisis (2006, 2007) and pre-crisis (2008, 2009) is discussed in this paragraph, the results of all variables can be found in columns 1, 2, 3 and 4 of table 5 and 6. Table 7 gives a concise overview of variables testing significant in what period.
The payout limit is weakly significant (p < 0.1) in non-crisis period with a negative coefficient of -0.186 on profitability. This effect is reversed during the crisis period where the variable is highly significant (p < 0.01) and has a positive coefficient of 0.199. This indicates that in periods of crisis, a higher payout limit is associated with higher profitability. Coverage of foreign currencies is has a positive coefficient for both periods and is highly significant (p < 0.01). Thus, deposit insurance schemes that offer this covers are associated with more profitable banks. This relationship is true for periods of crisis as well as in the non-crisis period. The variable capturing the effect of private administration tests significantly (p < 0.01) positive in both periods. The choice of how the deposit insurance scheme is administered is often based on whether there are capable private bodies to take on this job (e.g. bankers associations) and previous experiences with banking failures.
30 parties than from governments who have more discretionary power and deeper pockets in case of emergencies.
With regards to capital adequacy, there are two deposit insurance design variables that have a significant relationship with the dependent variable in the non-crisis period as well as in the crisis period. The coverage of foreign currency has negatively associated (p < 0.01) with capital adequacy in the pre-crisis period but a positive coefficient in the crisis period. This effect is seen more often with certain deposit insurance design features. These results have to be interpreted as follows: coverage of foreign currencies are associated with lower capital adequacy during normal times and with higher capital adequacy during crisis years. Risk-adjusted premiums has a significant positive coefficient pre-crisis (p < 0.01) as well as during the crisis (p < 0.05). This indicates that bank are more likely to have higher capital adequacy when they pay premiums based on the riskiness they add to the deposit insurance scheme.
Liquidity is significantly associated with multiple design features during the non-crisis period and crisis period. The payout limit is in both periods highly significantly (p < 0.01) positive. With a coefficient 19.321 pre-crisis and 7.132 during the crisis, banks in countries with high payout limits appear to be associated with higher liquidity. Reason for this could be that banks that operate under deposit insurance schemes with high payout limits are able to attract larger amount of funds from depositors leaving them with more deposits which can be held in liquid assets. Variables representing the credibility of the deposit insurance scheme are highly significant (p < 0.01) and negatively associated with liquidity. This is true both pre-crisis as well as during the pre-crisis for the variables capture effects of ex-ante funding, private funding and private administration. Their coefficients, -16.321, -28.453 and -19.932 respectively, become even more negative during the crisis and change to: -18.973, -31.447 and -19.998 respectively. The mentioned results are partly in line with previous research mentioned in chapter 2.
5.5. Deposit insurance and systemic risk in non-crisis and crisis period
31 the marginal expected shortfall in the crisis period. For the non-crisis period, this is true for the variables capturing effects of the payout limit, coverage of foreign currency and private administration of the funds. Variables that test significant in both periods are discussed here.
32 Table 5.
Panel regression non-crisis (2006, 2007)
(1) (2) (3) (4) (5) LOG(PAY_LIM) -0.186* (0.099) -4.703*** (0.792) 0.605** (0.259) 19.321*** (3.811) -0.330*** (0.092) COIN -0.319** (0.124) 2.269*** (0.269) 0.593 (0.787) -8.938 (8.345) 0.219 (0.177) FOR_CUR 0.775*** (0.048) -1.541 (1.537) -4.788*** (0.066) -7.467 (7.518) 0.611*** (0.084) RISK_ADJUST_P -0.126 (0.104) 4.914*** (0.873) 1.881*** (0.072) -1.615 (8.694) -0.150 (0.101) FUND_EX_ANTE 0.072*** (0.026) -6.740*** (0.813) -0.884*** (0.085) -16.321*** (3.391) -0.070 (0.149) FUND_PRIV -0.052 (0.142) 1.332*** (0.415) 0.390 (0.756) -28.453*** (0.843) -0.007 (0.046) ADM_PRIV 0.359*** (0.001) 0.878*** (0.277) 0.336 (0.360) -19.932*** (0.843) 0.420*** (0.058) LOG(POP) 0.234*** (0.162) -2.605*** (0.737) 1.815** (0.769) 4.497*** (1.185) -0.312*** (0.004) LOG(GDP) -0.177 (0.107) 0.508 (0.682) -0.781 (0.597) -4.039*** (1.196) 0.244*** (0.037) BANK_CONC -0.005*** (0.000) -0.155*** (0.024) 0.026*** (0.003) 0.633*** (0.107) -0.005 (0.003) GLOB_INT 0.056*** (0.017) -9.653*** (0.463) 2.380*** (0.805) -7.787 (6.460) 0.167 (0.106) CAP_GDP -0.002*** (0.000) -0.049*** (0.001) -0.014*** (0.001) -0.113 (0.094) 0.006*** (0.002) INS_RUL_LAW -0.289*** (0.029) 1.257 (0.947) 1.623** (0.714) 3.590 (5.863) -0.274*** (0.010) LOG(TOT_ASS) -0.149*** (0.001) 0.511*** (0.024) -1.568*** (0.046) 3.572*** (0.145) 0.144*** (0.002) Constant 5.717*** 100.883*** 20.277*** -124.730*** 1.097*** Observations (0.297) 722 (2.192) 698 (1.105) 655 (15.395) 707 (0.114) 172 R-squared 0.053 0.155 0.089 0.081 0.237
33 indication of how development of rule of law in a country. Log (Total Assets) is the log value of a bank’s total assets.
Table 6.
Panel regression crisis period (2008, 2009)
(1) (2) (3) (4) (5) LOG(PAY_LIM) 0.199*** (0.076) 2.291** (0.900) -0.002 (0.553) 7.132*** (2.067) 0.264* (0.143) COIN 0.355 (0.331) -10.374 (7.047) 1.902*** (0.084) -37.920*** (0.127) 0.700*** (0.106) FOR_CUR 0.898*** (0.017) -4.001 (15.423) 1.104*** (0.025) -25.253*** (6.046) 1.949*** (0.490) RISK_ADJUST_P 0.207 (0.249) -11.149 (8.222) 0.747** (0.335) 23.660*** (5.554) 0.304*** (0.059) FUND_EX_ANTE 0.158 (0.135) -2.633 (9.666) 0.527 (0.529) -18.973*** (1.780) 0.829*** (0.283) FUND_PRIV 0.583*** (0.101) -13.825*** (3.896) 0.725*** (0.206) -31.447*** (0.205) 0.829*** (0.283) ADM_PRIV 0.527*** (0.065) -22.406*** (3.198) -0.107 (0.707) -19.998*** (1.693) -0.116*** (0.007) LOG(POP) 0.672** (0.307) 12.228 (12.809) 1.351*** (0.358) 20.368** (8.320) -0.576*** (0.151) LOG(GDP) -0.591** (0.289) -22.816 (15.968) -1.006** (0.502) -9.682 (5.940) 0.373*** (0.081) BANK_CONC 0.016*** (0.005) -0.955*** (0.063) 0.025*** (0.006) 0.778*** (0.077) -0.037*** (0.011) GLOB_INT 0.014 (0.267) -14.500*** (0.067) 0.523** (0.234) 29.206*** (10.274) 1.025*** (0.043) CAP_GDP 0.007 (0.004) -0.150 (0.100) 0.048*** (0.012) -0.329*** (0.049) 0.023*** (0.002) INS_RUL_LAW -0.238*** (0.081) 25.949*** (8.321) 0.060 (0.068) 31.607*** (8.899) -0.548* (0.329) LOG(TOT_ASS) -0.067** (0.030) 4.463*** (0.637) -1.324*** (0.086) 2.256*** (0.273) 0.269*** (0.004) Constant 1.339 436.939** 30.312*** -115.727*** -6.782*** Observations (2.962) 721 (208.622) 711 (0.054) 666 (3.703) 716 (1.39) 178 R-squared 0.084 0.203 0.161 0.089 0.379
35 This table presents an overview of variables that tested significantly pre-crisis and/or in the crisis period. (1) represents the profitability, (2) the asset quality, (3) the capital adequacy and (4) the liquidity of banks. These variables are expressed in percentages. (5) is a measure for systemic risk. Log (Pay Limit) is the natural logarithm of the payout limit of a deposit insurance scheme. Coinsurance, foreign currency, risk-adjusted premiums, funded ex-ante, privately funded and privately administered are dummy variables.
Table 7.
Significance in pre- and crisis period
(1) (2) (3) (4) (5)
36
5.6. Deposit insurance and bank crises
Next, I examine the relationship between deposit insurance and bank crises. The results of the regression can be found in column 1, 2, 3 and 4 of table 11. The descriptive statistic can be found in table 8. The correlations of the full sample are presented in table 9. As mentioned, multiple variables are dropped from the equation because of the limited size of the sample. The correlation matrix of the variables used is presented in table 10.
The results do not indicate a significant relationship between deposit insurance features and the materialization of a banking crisis. None of the variables show a significant effect. This result can be due to the limitations in sample but more likely, the possible effect deposit insurance scheme has on banking crisis does not dominate other more relevant causes of crises.
Two of the control variables show a weakly significant association (p < 0.01), namely the variables for GDP and global integration. Both have a positive coefficient, indicating that countries with higher GDP and more global integration are associated with a higher probability of banking crises coming into existence. No compelling conclusion can be drawn on the basis of these results, other than there to be no evidence for deposit insurance schemes affecting banking crises.
5.7. Deposit insurance and severity of bank crises
In contrast to the previous paragraph where no significant is demonstrated, results indicate significant associations between deposit insurance and the severity of the banking crisis. The results are given in table 11, column 2.
38
Table 8.
Descriptive Statistics
Mean Median Maximum Minimum SD N
Output loss 25.518 23.000 106.000 0.000 28.654 28
Banking crisis dummy 0.679 1.000 1.000 0.000 0.476 28
Log (Pay Limit) 10.277 10.001 12.430 8.824 0.776 28
Coinsurance dummy 0.393 0.000 1.000 0.000 0.497 28
Privately Funded dummy 0.464 0.000 1.000 0.000 0.508 28
Privately Administered dummy 0.250 0.000 1.000 0.000 0.441 28
Bank Concentration 68.455 68.100 99.531 26.163 19.185 28
Stock market capitalization 68.557 49.220 265.295 7.321 54.988 28
Global Integration 1.069 0.967 3.188 0.523 0.532 28
39
Table 9.
Correlation matrix all variables
40
Table 10.
Correlation matrix selected variables
OUT_LOSS BANK_ CRIS LOG (PAY_ LIM) COIN FUND_ PRIV ADM_ PRIV BANK_ CONC CAP_ GDP LOG (GDP) GLOB _INT RUL_ LAW OUT_LOSS 1 BANK_CRIS 0.462 1 LOG(PAY_LIM) -0.040 -0.004 1 COIN -0.160 -0.229 -0.148 1 FUND_PRIV 0.190 0.181 0.267 -0.016 1 ADM_PRIV -0.216 0.044 0.315 -0.296 0.124 1 BANK_CONC -0.274 -0.208 0.123 0.067 -0.092 0.049 1 CAP_GDP -0.124 0.361 0.007 -0.254 0.299 0.482 0.265 1 LOG(GDP) -0.152 0.415 0.235 0.010 0.106 0.248 0.058 0.396 1 GLOB_INT 0.094 -0.024 -0.187 0.160 0.171 0.043 -0.161 0.087 -0.532 1 RUL_LAW 0.022 0.295 0.302 0.119 0.263 0.384 0.591 0.591 0.390 0.078 1
41
Table 11.
Country level regression
(6) (7) LOG(PAY_LIM) -0.188 (0.700) -6.912 (8.119) COIN -3.192 (2.038) -23.608** (9.485) FUND_PRIV -0.455 (1.161) 6.613 (10.935) ADM_PRIV -1.143 (2.048) -26.045** (11.675) LOG(GDP) 1.735* (0.951) -8.211 (6.115) BANK_CONC -0.076 (0.052) -1.140*** (0.386) GLOB_INT 4.954* (2.917) -15.058 (12.527) CAP_GDP 0.025 (0.024) -0.161 (0.124) RUL_LAW 41.230** (15.918) Constant -41.531 383.378* Observations (26.502) 28 (221.218) 28 R-squared 0.432 0.490
42
6. Conclusion
Deposit insurance is one of the instruments for governments to maintain financial stability. Whether nations succeed in achieving this goal has been the focus of research in the last years. The way deposit insurance schemes are composed differs across countries and so does its effect on risk. Using a newly constructed database, I studied the effect of differences in these deposit insurance schemes on bank risk and systemic risk. Furthermore, the focus of this paper lies on the materialization of crises and the associated cost. This is a less researched topic in context of the recent financial crisis. The relationship between deposit insurance and the manifestation of a banking crisis during the latest crisis is examined, as well as the association with the depth of the crises. Governments and regulators have shown great difficulty to deal with the financial crisis, some countries fared better through it than others. Different approaches have been used in the different countries to ensure financial stability, but what design features make sense with respect to bank risk and banking crises?
I show that overall, there is a differing effect of deposit insurance design features on the various risk indicators. The DI features have the most impact on liquidity of the European banks, most notably those that capture the effect of the credibility of the scheme. Besides the liquidity, the various deposit insurance design features show little significant association with the risk indicators for the sample period 2006 - 2009. What stand out is that the payout limit, often adjusted during the crisis to restore trust of depositors, does not show a relationship with any of the risk indicators, except for liquidity. Furthermore, I show that several DI features have opposite effects over the years when examining the pre-crisis and crisis period separately. These features are spread through the variables covering the generosity of the scheme as well as its credibility. In particular the effect of the payout limit is conversed in the crisis. The effect of variables capturing the credibility of the deposit insurance scheme are stable pre-crisis and crisis for the liquidity indicator. Lastly, I show that the risk indicator asset quality, and in lesser degree profitability, show significant association with the deposit insurance features in the pre-crisis period. Especially liquidity and systemic risk show a significant relationship with DI during the crisis. Generosity of the DI has shown to increase the systemic risk.
43 on the output loss, as well as the private administration of the scheme. These are results that should be subject of future research. The great majority control variables used in this paper have proven to be significantly linked with the bank risk and systemic risk indicators. They are moderately successful in the estimation regressions for banking crisis and output loss.
This research has implications for future regulations. Governments and regulators should recognize how different design features have opposing effects pre-crisis and during crisis periods. Despite design features not having a significant effect on bank and systemic risk for a total period, the effects during a crisis might be significant and influence the financial stability. The effects of deposit insurance differs, even among variables within the generosity and credibility category. For this reason, I do not offer general advice, each feature has to be examined separately. Concerning the severity of a banking crisis, there appears to be negative effect of coinsurance. This indicates that regulators and governments could possibly decrease the cost of banking crises by introducing coinsurance to the deposit insurance scheme. Also private administration of the deposit insurance scheme is association with a less severe crisis.
7. Limitations and future research
Despite careful preparation and execution of this research, I am aware of multiple limitations. These include shortcomings in sample and methodology. The aims of this research are mostly reached but the following limitations should be taken into account.
First of all, the number of banks that of which the MES has been collected can be considered too small. The MES is not provided for all banks in the sample used in this paper. Especially comparing the number of banks with a MES value those that reported on the Financial Soundness Indicators there is a wide gap, this could give rise to a biased sample. This is for the reason that the MES is collected from a source that does not provide these figures for all the banks in the sample. Nevertheless, using a systemic risk measure is of added value for this paper because the recent crisis has shown that the systemic risk dimension is of importance when considering bank risk.
44 Future research might profit from using a Hausman-Taylor model. This model is not used in this paper due to the strict requirements and assumptions it entails (e.g. requiring at least as many exogenous time-invariant variables as there are endogenous time-invariant variables).
Thirdly, the sample period chosen is one of the limitations of this paper. Though appropriate results are found in the pre-crisis and crisis sample, the overall sample does not show as much links between deposit insurance and risk as is suggested by literature. Selecting a sample over a longer period of non-crisis years would bring my results more in line with literature.
Lastly, I made a good start in relation deposit insurance to banking crises and output loss but due to the limited amount of countries in Europe, the sample has been very small. Future research should study the relationship for a larger region, adding to the credibility of the results and making proper interpretation of the results possible.
45 8. References
Acharya, V., Pedersen, L., Philippon, T. & Richardson, M. (2010), Measuring Systemic Risk.
AFA 2011 Denver Meetings Paper
Acharya, V. (2009). A theory of systemic risk and design of prudential bank regulation.
Journal of Financial Stability, 5(3), 224-255.
Anginer, D., Demirguc-Kunt, A., & Zhu, M. (2014). How does deposit insurance affect bank risk? evidence from the recent crisis. Journal of Banking & Finance, 48, 312-321.
Apergis, N., & Eleftheriou, S. (2012). Credit risk: The role of market and accounting information-evidence from U.S. firms and a FAVAR model. Procedia Economics and
Finance, 2, 53-62.
Ashcraft, A. B. (2008). Does the market discipline banks? New evidence from regulatory capital mix. Journal of Financial Intermediation, 17(4), 543-561.
Brooks, Chris. Introductory Econometrics for Finance. 2nd ed. Cambridge: Cambridge University Press, 2008.
Chernykh, L., & Cole, R. A. (2011). Does deposit insurance improve financial
intermediation? evidence from the russian experiment. Journal of Banking & Finance,
35(2), 388-402.
Cihak, M., Demirgüç-Kunt, A., Feyen, E., & Levine, R. (2012). Benchmarking financial systems around the world. World Bank Policy Research Working Paper, (6175)
Costa Navajas, Matias and Thegeya, Aaron, Financial Soundness Indicators and Banking Crises (December 2013). IMF Working Paper No. 13/263.
Cull, R., Senbet, L. W., & Sorge, M. (2005). Deposit insurance and financial development.
Journal of Money, Credit & Banking, 37(1), 43-82.
46 Demirgüç-Kunt, A., & Kane, E. J. (2002). Deposit insurance around the globe: Where does it
work? Journal of Economic Perspectives, 16(2), 175.
Demirgüç-Kunt, A., & Huizinga, H. (2004). Market discipline and deposit insurance. Journal
of Monetary Economics, 51(2), 375.
Demirgüç-Kunt, A., Kane, E. J., & Laeven, L. (2008). Determinants of deposit-insurance adoption and design. Journal of Financial Intermediation, 17(3), 407-438.
Demirgüç-Kunt, A., Kane, E.J. & Laeven, L. (2013), "Deposit Insurance Database," Policy Research Working Paper #6934, Washington, DC: World Bank.
Demirgüç-Kunt, A., Baybars K., & Luc Laeven, (2005), "Deposit Insurance around the World: A Comprehensive Database," Policy Research Working Paper #3628, Washington, DC: World Bank
Diamond, D., & Dybvig, P. (1983). Bank Runs, Deposit Insurance, and Liquidity. Journal Of Political Economy, 91(3), 401-419
European Commission (2010), Proposal for EU of the European parliament and of the council on Deposit Guarantee Schemes [recast], COM(2010) 368 Brussels.
González, F. (2005). Bank regulation and risk-taking incentives: An international comparison of bank risk. Journal of Banking & Finance, 29(5), 1153-1184.
Hull, J. (2012). Risk management and financial institutions. John Wiley & Sons.
Kaufmann, D., Kraay, A., & Mastruzzi, M. (2011). The worldwide governance indicators: Methodology and analytical issues. Hague Journal on the Rule of Law, 3(02), 220-246.
Kleftouri, N. (2014). Meeting the rationale of deposit protection system. Journal of Financial
Regulation & Compliance, 22(4), 300-317.
47 McCoy, P.A. (2006), “The moral hazard implications of deposit insurance: theory and
evidence”, paper presented at the Seminar on Current Developments in Monetary and Financial Law, Washington, DC, 23-27 October, available at:
www.imf.org/external/np/seminars/eng/2006/mfl/pam.pdf (accessed December 2015).
Niinimäki, J. (2004). The effects of competition on banks’ risk taking. Journal of Economics,
81(3), 199-222.
Shiers, A. F. (1994). Deposit insurance and banking system risk: Some empirical evidence.
Quarterly Review of Economics & Finance, 34(4), 347.
