Deposit Insurance Policies on
Bank Solvency in the EU-‐28
★
Author: Leonard A.J. Terwisscha van Scheltingaa, §
Supervisor: Dr. Lammertjan Damb
a MSc Economics, MSc Finance, University of Groningen, the Netherlands
b University of Groningen and CIBIF, the Netherlands
T H E S I S I N F O A B S T R A C T
Date: 26th of June 2014 JEL classification: F36 G21 G28 Keywords: Deposit insurance Bank solvency Coverage
Deposit insurance is a strong medicine to prevent bank runs with moral hazard as a side effect. This paper identifies this moral hazard effect for EU-‐28 countries by estimating the relation between different policies of deposit insurance coverage on solvency of banks. A panel data model with data between 1998 and 2013 shows that, on average, a 1 unit increase in deposit insurance coverage to GDP per capita ratio leads to a reduction in bank solvency of 1.8%. This effect is stronger for deposit insurance systems with full coverage.
★ Course code MSc Economics: EBM877A20. Course code MSc Finance: EBM866B20
1. Introduction
This paper estimates the impact of deposit insurance on bank solvency in the EU. Deposit insurance is a guarantee to reimburse deposits of depositors whose bank is unable to meet its debt obligations. It is either explicit or implicit. A country has explicit deposit insurance if a contractual obligation is present in which depositors are (fully) reimbursed when a bank is unable to meet its debt obligations. With implicit deposit insurance a contractual obligation is not present. However, when a considerable banking crisis occurs, governments will ultimately bail out banks to save their financial system. In this paper the focus is on explicit deposit insurance; therefore I further refer to this as ‘deposit insurance’.
Deposit insurance is a popular financial safety net that many countries introduce. The main reason is that deposit insurance can prevent bank runs (Diamond and Dybvig, 1983). Furthermore, it is also appealing for politicians because it protects small depositors without directly affecting the government budget. Furthermore, external pressure from multiple institutions increases the chance of adopting deposit insurance (Demirguc-‐Kunt, Kane and Leaven, 2006): the World Bank endorses it, IMF’s crisis management recommends it since the 1990s (Garcia, 1999) and the EU obliges it to its member states as of 1994 (European Commission, 1994). Moreover, during or just after a financial crisis, countries might introduce a deposit insurance system as part of a rescue plan to save their financial system (Demirguc-‐Kunt, Kane and Leaven, 2006).
During the last decades, the abundance of motives for introducing deposit insurance translated into a considerable increase in the amount of deposit insurance systems. For example, between 1980 and 2003 the number of countries with a deposit insurance system increased from 20 to 87 (Demirgüç-‐Kunt et al, 2005). This increase is also present in the 28 countries that nowadays form the EU: from 14 countries at the start of the EU in 1992, to all 28 countries as of 2003, when Malta introduced their deposit insurance system.
financial crisis between 2007 and 2009. The EU 2009 directive set new goals with respect to the coverage, raising the previous coverage of 20,000 euros of the (outdated) EU 1994 directive to 50,000 euros for the end of 2009 and 100,000 euros for the end of 2010. This led to a recovery of confidence in banks among depositors. Apparently the initial amount was too low.
Despite its positive features, deposit insurance comes with moral hazard as a side effect. An unintended consequence of increasing coverage is that it reduces the incentive of depositors to monitor banks. For the pre-‐crisis coverage of 20,000 euros in the EU, a rational depositor who intends to stall 80,000 euros at a bank will internalize the riskiness off a bank into the decision on which bank to choose. However, after the raise in coverage to 100,000 euros, the depositor will not internalize the riskiness of a bank anymore, but shifts focus more towards the interest rate that a bank offers. This leads the depositor to choose for a higher interest-‐paying bank engaged in riskier investments, rather than a lower interest-‐paying bank engaged in safer investments. To stay competitive, the low interest-‐paying bank needs to increase interest rates resulting in a new equilibrium in which interest rates are higher. To repay the increased interest rates, banks need to allocate more funds towards higher yielding but riskier investments. This has a decreasing effect on bank solvency: return on assets becomes less stable, leading to a decrease in the degree to which a bank is able to meet its obligations. Therefore, the resulting new equilibrium with higher coverage is more fragile and riskier compared to the former equilibrium with lower coverage.
I use a panel data model to identify the effects of deposit insurance on bank solvency in the EU-‐28. I regress the z-‐score, which is a solvency measure for banks, with different forms of deposit insurance coverage. This includes no coverage at all, limited coverage and full coverage. A dummy variable captures both the effect of the presence of a deposit insurance system as well as the effect of full coverage on bank solvency. I further dissect the full coverage dummy into two other dummies: a full legal coverage dummy and a dummy for full coverage provided by a political statement.
coverage results in lower solvency of individual banks. Furthermore, adopting a deposit insurance system results in lower solvency of individual banks. The negative relation is even stronger for deposit insurance systems with full coverage. Within full coverage systems, full legal coverage leads to the highest decrease in bank solvency. Full coverage by means of a political statement has a weaker decreasing effect on bank solvency.
This paper contributes to the literature because it uses a data continuous measure of deposit insurance rather than a dummy variable for the presence of it. Because deposit insurance is either absent or present, capturing the effect of deposit insurance by means of a dummy variable is a common procedure in the deposit insurance literature. It is easy to implement and does not require data on the amount of coverage to capture the effect of a deposit insurance system. However, all EU-‐28 countries nowadays run a deposit insurance system as of 2003. Capturing the effect of deposit insurance in the EU from 2003 onwards with a dummy variable becomes impossible. The advantage of using a data continuous measure is that it captures the marginal effect of a change in coverage, rather than capturing the effect of a presence of a deposit insurance system.
Another contribution to the literature is that I supplement the database by Demirgüç-‐ Kunt et al (2005) with respect to coverage data. To use a continuous measure of coverage, I update the database with 10 years of coverage of the EU-‐28 member states. I use sources of the EU, World Bank, IMF, national deposit insurance institutions and national legislation.
This paper continues as follows. The next section reviews the literature on deposit insurance as well as methods to assess its moral hazard effect. Section 3 provides an explanation of the methodology. Section 4 describes the data and sources and section 5 presents the results. Section 6 contains robustness checks and section 7 concludes.
2. Literature review
the central bank’s function of being the lender of last resort. Gropp, Hakenes and Schnabel (2011) show that government bailouts lead to excessive risk taking in the banking industry, especially among competitors. Even an increase in expectations of a bailout results in an increase in probability of financial distress (Dam and Koetter, 2012). Therefore, when a country introduces a deposit insurance system, it should carefully weigh the pros and cons of it.
2.1 Pros and cons of deposit insurance
The main benefit of deposit insurance is to prevent wasteful liquidations of bank assets caused by bank runs (Demirgüç-‐Kunt, Kane and Leaven, 2003). The theoretical literature seems to confirm this. The benefits of deposit insurance can outweigh the costs (Bryant, 1980). It can prevent bank runs (Diamond and Dybvig, 1983) and promote financial stability (Demirgüç-‐Kunt and Detragiache, 2002). Furthermore, deposit insurance protects small and uninformed depositors against bank failures (Lé, 2013).
The disadvantage of deposit insurance is that it comes at the cost of moral hazard. Since introducing deposit insurance moves part of the costs of bankruptcy from the depositors and banks to the government, moral hazard occurs among depositors as well as banks. From the depositor’s perspective, depositors are disentangled form the consequences of their actions (Calomiris, 1990; Gennote and Pyle, 1991; MacDonald, 1996): in their choice on where to stall deposits, depositors are less incentivized to base their decision on the financial health of a bank, and more incentivized to base it on the attractiveness of the offered interest rate. This lowers the market discipline through deposit interest rates (Demirgüç-‐Kunt and Huizinga, 2004).
insurance increases. Hence, if the deposit insurance premium has no relation to the expected cost of bankruptcy, it is optimal for banks to increase risk taking if deposit insurance coverage is increased (Kareken and Wallace, 1978).
2.2 Full coverage
Bank runs can be fully eliminated in a deposit insurance system with fully covered deposits. However, a disadvantage of this is that depositors lose the incentive to monitor their banks’ activities, leading to a loss in bank monitoring information provided by depositors. Battacharya, Boot and Thakor (1998) compare a deposit insurance system characterized by full coverage and limited coverage in a theoretical model. They show that a limited coverage system encourages market discipline, caused by increased bank monitoring of informed depositors. In the Diamond-‐Dybvig banking model, Hazlett (1997) finds that coinsurance and deductible schemes reduce the moral hazard problem relative to full insurance schemes. However, the cost of these alternatives is that they are less effective in preventing bank runs.
A number of studies confirm the theory that moral hazard problems are less present in limited coverage schemes relative to full coverage schemes. Using a sample of 4,109 publicly traded banks in 96 countries, Anginer, Demirgüç-‐Kunt and Zhu (2013) (ADZ (2013) from here on) find that full coverage has a magnifying effect on bank risk compared to limited coverage. This is consistent with Demirgüç-‐Kunt and Detragiache (2002), who find that full coverage might further intensify the moral hazard problem. Imai (2006) compares the change from full coverage to limited coverage in Japan in 2002. He finds that market discipline has increased after the transition to limited coverage.
2.3 Coverage of deposit insurance
EU. To join the EU, the (low income) accession countries are obliged to adapt the EU 1994 directive on deposit insurance schemes and impose 20,000 euros coverage. His findings suggest that if accession countries as a group had not joined the EU, the probability of adopting deposit insurance is low, as well as imposing a 20,000 euros coverage limit. He concludes that the obliged adoption of deposit insurance therefore is a cost for EU accession countries: the relatively high coverage of 20,000 euros may lead to reduced financial stability because of increased risk taking by banks. Dimitrova and Nenovsky (2008) find evidence of over-‐insurance of the accession countries as well. They argue that the compliance of accession countries to the EU 1994 directive has led to a coverage that is excessive relative to GDP per capita and bank capital. The excessive coverage increases moral hazard by distorting incentives of the poorly capitalized banks in the accession countries.
2.4 Empirical findings on the effect of deposit insurance on bank risk
In a recent study, ADZ (2013) investigate the specific link between deposit insurance and bank risk before and during the global financial crisis. They find that deposit insurance increases bank risk as well as systemic fragility in the years before the global financial crisis. During crisis time, however, deposit insurance has a decreasing effect on bank risk, and systemic fragility is lower in countries with a deposit insurance system. The overall effect remains negative, because their sample consists of a larger period without a crisis than with a crisis. In addition, good bank supervision can alleviate the unintended consequences of deposit insurance on bank systemic risk during good times, suggesting that imposing the right incentives is important for ensuring systemic stability.
capitalized compared to uninsured banks. Also, capitalization and liquidity are important factors that could determine bank failure.
2.5 Empirical methods to identify the effect of deposit insurance on bank risk
The banking literature contains various methods to examine the effect of deposit insurance on a bank’s insolvency risk. An important aspect is the methodology to estimate the risk of the value of a bank’s assets. Many studies use the option pricing model by Merton (1977) in which deposit insurance is modeled as a put option on the bank’s assets. For example, Leaven (2002b) uses the model to calculate implicit deposit insurance premiums for each bank, which then serve as a proxy for a bank’s riskiness. Merton’s (1977) model is attractive to use because it establishes a direct link between the value of the bank’s assets and the value of the deposit insurance contract. Furthermore, it uses market values of the bank’s assets and equity rather than accounting values. However, this study takes into account both listed and non-‐listed banks and therefore does not use Merton’s (1977) model. Another commonly used approach to estimate risk of a bank’s assets is to measure the volatility of a bank’s stock return. It is employed in ADZ (2013). Similar to the model of Merton (1977), its advantage is that due to its market based nature, highly frequent data can be used as input. A third model to assess bank risk is the z-‐score1. It is an ‘expected loss pricing’
model (Leaven 2002a) that estimates the expected default probability of a bank. The advantage of using the z-‐score is that it is based on simple accounting values rather than market values. It is therefore applicable to every bank.
The variable deposit insurance is differently treated in the literature. A commonly employed method is using a dataset with many counties that do or do not have deposit insurance, and then filter out the effect of deposit insurance by comparing the difference between them. An example is the study of Anginer, Demirgüç-‐Kunt and Zhu (2013). In a dataset of 96 countries, they use a dummy variable for both the existence of deposit insurance and the existence of full coverage. Subsequently, they regress these dummy variables on z-‐scores of banks. Huizinga (2005) uses data on deposit insurance of countries outside the EU to estimate the coverage that EU accession countries would
1 See for example: de Nicolo (2001), Dam and Koetter (2012), Anginer, Demirgüç-‐Kunt
have had (if any) if they were not forced by the EU to oblige to the EU 1994 directive on deposit insurance schemes.
Based on theory and empirical findings, I expect a negative relation between deposit insurance and bank solvency in the EU because of the moral hazard effect. Offering full coverage should magnify this effect even more. Most empirical studies have in common that they compare deposit insurance systems with different characteristics: periods of full coverage are compared with periods of limited coverage by using a dummy variable. In this paper I capture the effect of full coverage on bank insolvency with a dummy variable, but the variable for limited coverage is a data continuous variable.
3. Methodology
To identify the effect of deposit insurance on bank solvency in the EU-‐28, I regress deposit insurance coverage on a bank solvency measure in a multivariate panel data model. This section discusses the rationale behind the dependent, independent, control variables and the error term, followed by the econometric specification.
3.1 Dependent variable
The model contains the z-‐score as a dependent variable to reduce a potential selection bias. Z-‐score is a measure for bank solvency and is accounting based. This allows an assessment of bank solvency for every bank in the dataset, rather than only market based ones. I define the z-‐score as:
where z-‐score equals the amount of standard deviations returns may decrease before they exhaust a bank’s capital. 𝑅𝑂𝐴𝐴 is the return on average assets (π/AA) with π as net income. Specifically, 𝑅𝑂𝐴𝐴 is a bank’s net income in year t over average assets, defined as the average amount of assets on a 2 year rolling window: t and t-‐1. To overcome losing the starting year for which t-‐1 is not available, I create the 𝑅𝑂𝐴𝐴 for starting years by using only assets in t. 𝐸/𝑇𝐴 represents equity (E) over total assets (TA) and 𝜎!"## is
the standard deviation of the return on a bank’s average assets of all available years.
z − score =𝑅𝑂𝐴𝐴 + 𝐸/𝑇𝐴 𝜎!"##
Assuming that a insolvency occurs when a bank’s losses exceed its equity, the probability of an insolvent bank equals: P(ROAA<E/TA). I furthermore assume that
ROAA follows a normal distribution, the z-‐score then relates inversely to the probability
of bank insolvency (Roy, 1952). The z-‐score can be interpreted as the distance to insolvency: the higher the z-‐score, the more solvent the bank. Because of the moral hazard mechanism, I expect a negative relation between deposit insurance and z-‐scores of banks. The model contains the natural logarithm of the z-‐score because the distribution of the z-‐score is highly skewed.
3.2 Independent variables
Different forms of deposit insurance coverage enter as independent variable in the regression model. Starting with deposit insurance coverage, I scale coverage by GDP per capita. This transformation provides two main advantages compared with using unscaled coverage. First, it better reflects relative differences in the amount of coverage within a country and between countries. As a consequence, it better reflects the degree to which depositors are incentivized to monitor their bank. For example, in line with the former EU-‐1994 directive on deposit guarantee schemes, Austria has offered coverage of 20,000 euros between 1999 and 2008. During that time, due to real economic growth and inflation Austrian GDP per capita increased from 24,900 to 34,000 euros respectively. Austrian depositors in 2008 are therefore more likely to reach the coverage limit compared to Austrian depositors in 1999, implying that the former are more incentivized to monitor their bank than the latter. The same argument applies for differences in income between countries. Second, scaling coverage to GDP per capita results in a variable with more variance. Although coverage increased during the recent financial crisis, especially the decade before the financial crisis seems to have been a period in which governments lost attention on the topic of deposit insurance: many countries left their coverage unchanged. To obtain the most information out of the coverage values, scaling them to GDP per capita leads to more variance in the deposit insurance variable.
and ‘0’ otherwise. Subsequently, I split this dummy into full legal coverage and full political coverage.
The expectation is that a negative relation exists between of all dummies with respect to the z-‐score and that the effect of full legal coverage is larger in magnitude than full political coverage. First, in case of absence of deposit insurance, the bill in case of bankruptcy is implicitly for governments, because they will in the end save their financial system by saving a bank, and depositors who lose their money. Depositors and government therefore have an incentive to monitor banks. There is no moral hazard, and therefore no disruptions in market for deposits. Adopting a deposit insurance system transfers the bill in part from depositors to the government. The monitoring incentive partly shifts from depositor to government. Depositor’s incentive to monitor their bank decreases, resulting in an increased demand for riskier and higher yielding deposit accounts. Safe banks are forced to increase their interest rates to keep attracting deposits. The implication of this is that banks need to be involved in riskier investments to be able to repay their increased cost of money. The market becomes disrupted, higher interest rates. Banks become less solvent because of excessive risk taking.
In case of deposit insurance with full legal coverage, the bill in case of a bankruptcy is for the government in full. Therefore, the incentive to monitor is for governments in full. However, depositors might still have an incentive to monitor banks, depending on the credibility of the government. If a government is not credible and not able to repay deposits, it is likely that it raises taxes to repay, which indirectly shifts the bill to depositors. Therefore, in line with the Ricardian equivalence theorem (Ricardo, 1888) depositors could adjust their behavior depending on the method of financing by the government. I therefore expect that banks are less solvent in a deposit market with full coverage and a credible government, compared to the same situation but without a credible government. Nevertheless, a deposit insurance system with full legal coverage should have a decreasing effect on solvency of banks. This effect should be even stronger compared to limited coverage systems because the disruptive effect in the market of deposits is stronger.
Lastly, deposit insurance with full coverage by a political statement. The same arguments of full legal coverage count for full political coverage, except that depositor’s incentive to monitor banks now depends on the credibility of the political statement. As political statements are less binding compared to the law, the hypothesis is that a full political coverage system has a decreasing effect on solvency of banks, but the effect is less strong compared to full legal coverage. After all, it is not illegal for a politician not to keep your word, but it is illegal not to obey the law.
3.3 Control variables
To improve identification of the effect of deposit insurance on bank solvency, I follow ADZ (2013) and add bank specific and macroeconomic control variables to the model. Macroeconomic conditions influence the performance of all banks and the value of their pledged collateral, which affects a bank’s return on average assets and equity over total assets ratio and so solvency. Bank specific conditions correct for different characteristics of a bank that influence solvency. The model contains 4 bank specific control variables and 2 macroeconomic control variables. I transform these control variables by taking their natural logarithms if this leads towards a less-‐skewed distribution.
3.3.1 Bank specific control variables
First, the model contains the natural logarithm of total assets to correct for bank size. Ceteris paribus, larger banks are sooner systemically important than smaller banks. Therefore larger banks are subject to stricter supervision, which limits excessive risk taking. Furthermore, large banks that are too big to fail are able to finance themselves in a cheaper way because the market allows a smaller risk premium for them to pay. (Bijlsma, Lukkezen and Marinova, 2014) This positively influences return on assets, leading to a higher z-‐score. Because of stricter supervision and cheaper financing I expect a positive relation between the natural logarithm of total assets and z-‐score.
bank, I expect an increase in solvency if the share of deposits increases. Therefore, the expectation is that the variable is positively related to the z-‐score. The last bank specific control variable is loan loss provisions over net interest revenue, to correct for the share of provisions of a bank. Banks with non-‐performing loans handle this by setting funds aside in case the debitor is not able to pay back their loan. A high share of loan loss provisions therefore indicates a less solvent bank, leading to the hypothesis that it relates negatively to the z-‐score.
3.3.2 Macroeconomic control variables
First, I include the natural logarithm of variance of GDP growth as a proxy for economic stability. On average, firms in stable economies should have an easier time in paying back their loans on a continuous base than firms in unstable economies. Hence, banks in stable economies should have a lower standard deviation of return on average assets than banks in unstable economies. Variance of GDP growth should therefore relate negatively to the z-‐score. Second, I add the natural logarithm of GDP per capita to measure the economic development of a country. Well-‐developed countries often have a financial system that is more stable than less developed countries. The quality of bank supervision is higher as well. For this reason, the natural logarithm of GDP per capita should have a positive relation with the z-‐scores.
3.4 Error term
Possible omitted variables will end up in the error term. An example is the shareholder structure of a bank, which could be of influence on the amount of risk a bank takes (Leaven and Levine, 2009). However, due the scarceness of information with respect to shareholder structure of the the banks in the sample I do not include it in the model. Including it would greatly reduce the amount of observations, and possibly significance.
3.5 Econometric specification
non-‐crisis years. I use random bank effects because the focus of this paper lies on the total population of banks in the EU-‐28. Assuming a normal distribution of bank characteristic effects, random bank effects allow inference with respect to the solvency of all banks in the EU-‐28, whereas results of fixed bank effects only allow inference with respect to the solvency of banks in the sample. Furthermore, the assumption is that the sample of banks is a random draw of the bank population as a whole. This implies that the sample is indeed representative for the total bank population in the EU-‐28.
The econometric specification is as follows:
𝑧 − 𝑠𝑐𝑜𝑟𝑒!,! = 𝛽!+ 𝛽! 𝐷𝑒𝑝𝑜𝑠𝑖𝑡 𝐼𝑛𝑠𝑢𝑟𝑎𝑛𝑐𝑒 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 !,!+ 𝛽!𝐶𝑜𝑛𝑡𝑟𝑜𝑙!,!
+ 𝛼! + 𝛾!+ 𝑢!,!
(2)
where 𝑧𝑠𝑐𝑜𝑟𝑒!,! is the z-‐score of bank i in at time t. It depends on a constant 𝛽!, and on
𝐷𝑒𝑝𝑜𝑠𝑖𝑡 𝐼𝑛𝑠𝑢𝑟𝑎𝑛𝑐𝑒 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠!,!, representing the type of deposit insurance policy. The
types of deposit insurance policies are: presence of deposit insurance Deposit Insurance
Presence, limited coverage Coverage/GDP per Capita, a full coverage system in form of Full Coverage and full legal and full political coverage Full Legal Coverage and Full
Political Coverage respectively. Furthermore, 𝐶𝑜𝑛𝑡𝑟𝑜𝑙!,! contains variables that influence
a bank’s solvency on both bank specific and macroeconomic level. 𝛼! is the random effect component for banks and 𝛾! the fixed time effect.
The coefficient of interest is 𝛽!. Its value quantifies the relation between the different
compared to deposit insurance that is rooted in the law. Therefore, the hypothesis is that the coefficient of Full Legal Coverage is larger in magnitude compared to the coefficient of Full Political Coverage.
4. Data and descriptive statistics 4.1 Selection procedure and sources
The dataset comprises macroeconomic variables, bank specific variables and deposit insurance variables of all 28 EU countries. Macroeconomic variables are from Eurostat. The availability of bank specific variables in Bankscope forms a bottleneck for the timespan of the dataset. Bankscope offers bank specific data starting in 1998 with a yearly frequency; therefore the time span equals 16 years (from 1998 till 2013). Deposit insurance coverage data for the years 1998 till 2003 stem from the deposit insurance database by Demirgüç-‐Kunt et al (2005). I further update this database with respect to the coverage amounts for the years 2004 till 2013 by using sources of multiple financial institutions. They include country reports of the IMF, EU directives on deposit insurance schemes of 1994 and 2009, national legislation, documents of the Financial Stability Board, financial sector assessments by the World Bank and IMF and data from national deposit insurance institutions. Table A.1 in the appendix shows for all three types of variables their data source and appropriate year. The unbalanced panel forms another bottleneck for the number of observations. With respect to deposit insurance, years are missing, especially the period between 2004 and 2005 for accession countries. With respect to bank specific data, gaps in the dataset exist because banks go bankrupt, or simply do not report their data to Bankscope.
4.2 Descriptive statistics
Table 1 provides descriptive statistics for the variables. The maximum number of observations equals 69888. This is the case for the Deposit Insurance Presence dummy variable and the three types of full coverage dummy variables. However, the variable Deposits and Short Term Funding/Total Assets restrict the maximum number of observations to 15,219.
4.2.1 Dependent variable
The mean of the natural logarithm of z-‐score equals 3.39, and is lower to the mean of 3.50 of ADZ (2013). An explanation for this could be that they include banks of 96 countries, of which many are developing countries that have a less stable economic environment or a lower quality of supervision. However, the standard deviation is 1.15, which is slightly higher than the 1.08 of ADZ (2013). The minimum value of -‐6.58 indicates that a bank has a z-‐score between 0 and 1. In other words, the probability of bank insolvency is than 1 standard deviation.
4.2.2 Independent variables
The Coverage/GDP per Capita ratio has a mean of 2.19, indicating that on average, a country in the dataset offers a deposit insurance coverage of 2.19 times the average per capita income. To get a better sense of this, the average GDP per Capita in the EU-‐28 for the years between 1998 and 2013 is 27 thousand euros. Multiplying this by 2.19 yields 59 thousand euros of deposit insurance coverage for an average country with the average coverage to GDP per Capita ratio. The maximum coverage/GDP per Capita is 21.74 and its minimum value is 0 euros for countries that do not have deposit insurance in one or more of the years between 1998 and 2013. This is the case for Cyprus, Malta, and Slovenia, which have deposit insurance as of 2000, 2003, and 2001 respectively.