Deposit Insurance and Bank Risk Taking

This paper estimates the effect of deposit insurance on the risk-taking behaviour of banks. As shown in the theoretical literature, deposit insurance may induce moral hazard and incentivise banks to take on more risk. In this paper we provide an experimental setup in which we exploit an increase in the coverage limit of deposit insurance in the U.S. in order to identify the difference in risk taking by banks that were affected and banks that were not. This difference comes from the fact that state chartered savings banks in Massachusetts had unlimited deposit insurance coverage at the time when it was increased for all other banks in the US. Given that all banks in the sample are subject to the same regulatory and supervisory requirements, and that they are similar in other characteristics, we can isolate the effect of such increase in deposit insurance. We find, contrary to the literature, that this increase in deposit insurance did not increase bank risk-taking, nor did it affect market discipline, evident through a lack of an effect on deposit rates.


Introduction
Deposit insurance is one of the standard measures to assuring stability of the financial system. By assuring the depositors that their savings will be available to them when their liquidity need arises, they do not react to information or speculation about bank liquidity by withdrawing their deposits. This allows the financial system to weather periods of tight liquidity and maintain a steady flow of lending to households and firms.
In the light of high contingent liabilities for the government, the savings and loan crisis of the 1980s and 1990s and, more recently, the 2000s financial crisis, have spurred a debate about the costs and benefits of providing deposit insurance.
Theory provides two important results regarding the benefits of deposit insurance. On the one hand, as shown in the seminal work by Diamond and Dybvig (1983), deposit insurance eliminates the self-fulfilling equilibrium where depositors find it optimal to withdraw their deposits in the belief that other depositors will do the same. This leaves only depositors with liquidity needs to withdraw their deposits. On the other hand, as shown among others by Cooper and Ross (2002), deposit insurance induces a problem of moral hazard. If no deposit insurance is in place, a bank making riskier investments incurs a cost through compensating the depositors with higher deposit rates. If depositors are insured against potential losses, they will increase their supply of deposits, abstain from demanding a premium for the riskier investments undertaken by the banks, and will be less careful in monitoring their bank. This allows a bank to take on riskier investments.
Empirical literature has thoroughly covered the effect of deposit insurance on risk taking and the extent of market discipline. Focusing on cross country comparison Demirguc- Kunt and Huizinga (1999) find that explicit deposit insurance reduces required deposit interest rates at a cost of reduced market discipline. Similarly, Demirguc-Kunt and Detragiache (2002), find that explicit deposit insurance increases the likelihood of a banking crisis, especially in an institutionally weak environment. Furthermore, the adverse affect is strengthened by the extensiveness of the insurance coverage. Nier and Baumann (2006) focus on the effect of government safety nets like deposit insurance on capital buffers. Their results suggest that safety nets lower the capital buffers, an indication of an increased solvency risk, and that competition among banks dampens this effect. More recently, Anginer et al. (2014a) estimate the effect of deposit insurance on bank risk, measured by the bank Z-score, and on systemic fragility measured as the marginal expected shortfall (following Acharya et al. (2012)). Their results indicate that deposit insurance increased bank risk and systemic fragility in the years leading up to the global financial crisis. However, systemic fragility was reduced due to deposit insurance during the financial crisis. Focusing instead on a forward looking risk measure (the CDS spread) but still applying a cross country comparison, Liu et al. (2016) find that banks in countries with explicit deposit insurance systems tend to have higher CDS spreads.
A positive effect of deposit insurance on measures of risk found in studies above is further confirmed by Boyle et al. (2015) using a conjoint analysis approach. They find that respondents from countries without explicit deposit insurance exhibit greater risk of withdrawing their deposits in a hypothetical scenario of a failure of a competing bank and a tendency to impose higher deposit interest rate, suggesting a presence of market disciplining. Furthermore, in a natural experiment setting based on the introduction of deposit insurance within the Russian banking sector, Chernykh and Cole (2011), find that deposit insurance increased the solvency ratio and the loan-to-assets ratio, which, they argue, indicates a higher solvency risk and credit risk. Similarly, focusing on a with-in country analysis, based on internal loan ratings, Ioannidou and Penas (2010) find that in periods following the introduction the deposit insurance in Bolivia, banks were more likely to initiate riskier loans. They also conclude that this results is due to lower market discipline imposed by large depositors. Finally, using similar variation in deposit insurance as we do in this paper, Lambert et al. (2017), find that following an increase in deposit insurance coverage, banks which have been more affected by that increase become riskier, i.e., their Z-score has decreased.
The literature on the effect of deposit insurance on risk taking, although vast, exhibits some limitations which, in our opinion, may lead to a failure to identify the causal relationship. Firstly, while a focus on cross country comparisons ensures some variation in deposit insurance, such variation cannot exclude endogeneity of bank behaviour and deposit insurance policies. Furthermore, such an analysis neglects other institutional differences between banking markets. Secondly, it is important to distinguish between different measures of riskiness of the bank, such as the Z-score, which are based on the performance of the portfolio, and measures of risk taking. This is mainly due to the fact that economic shocks, which contribute to establishing deposit insurance, can also affect the state of banks' portfolios, inducing endogeneity. Measures of risk based on new investment are not subject to this shortcoming, as they reveal the choice of banks at the moment of investment. This paper adds to the literature by addressing these sources endogeneity. Firstly, to make sure that deposit insurance policy is not endogenous, we use a natural experiment setting from the US banking regulation. We analyse an episode where a change in the insurance coverage limit affected some banks but not others in an exogenous way (see below). Secondly, we build a bank-level measure of risk taking based on newly issued mortgage loans. These two methodological contributions allow us to estimate a causal effect of deposit insurance on risk taking.
We exploit an increase in the coverage limit of deposit insurance in the US for identification. In the US, all banks have deposit insurance coverage provided by the FDIC (Federal Deposit Insurance Corporation). In addition, state-chartered savings banks in the state of Massachusetts are provided with unlimited coverage by a private insurer. Membership to this unlimited insurance scheme is mandatory for all savings banks chartered in Massachusetts since 1934. The unlimited nature of this coverage is key to our identification.
In order to measure risk-taking, we use mortgage application data from the Home Mortgage Disclosure Act. We estimate the propensity to lend for a given level of borrower risk, which we proxy using loan-to-income ratios of each loan application.
Contrary to the literature, we find no significant effect of an increase in deposit insurance coverage on bank risk taking. Banks are not more likely to grant a loan for any given loan-to-income ratio of the borrower after such an increase.
Furthermore, to explore the reason for the absence of a significant effect, we use the same natural experiment setting to test whether an increase in deposit insurance reduces vigilance by depositors as indicated by lower movement in the supply of deposits. Theory suggests that the incentive to increase risk taking stems from a relaxed deposit supply which does not penalise such risk taking. We find that an increase in deposit insurance does not reduce deposit interest rates, indicating a lack of a market-discipline effect. This result confirms prior findings of Schmukler (2001) and rationalises a lack of an effect on risk taking. This paper is organised as followed: section 2 presents the mortgage and balance sheet data used in the empirical analysis. Section 3 presents the methodology, the natural experiment setting and the construction of the risk taking measure. Section 4 presents the results of the estimation of the effect of an increase in deposit insurance coverage on risk taking. Section 5 presents the results of the estimation of the effect on deposit supply. Section 6 provides a robustness analysis. Finally, section 7 concludes.

Data
Our empirical strategy relies on a well defined measure of risk taking and on comparability of the treatment and the control group. For the purpose of building a risk measure (i.e. an elasticity of loan origination to Loan-to-Income) our main source of data are the Home Mortgage Disclosure Act Loan Application Registries which contains the data on mortgage applications. To ensure comparability of the treatment and control groups we use the FDIC Call reports which contain the bank balance data.
The Home Mortgage Disclosure Act (HMDA) obliges banks above a set threshold of assets to report on mortgage applications. The yearly Loan Application Registries of banks which meet the criteria to report, contain all mortgage loan applications, the properties of the applicant and potential co-applicant (ethnicity, race, gender, income), the loan properties (amount, type, purpose, rate spread for some, occupancy), the properties of the house (type, census tract, etc.), the properties of the census tract (income relative to the relevant Metropolitan Statistical Area, minority population, number of housing units, etc.), and the action taken (origination, denial and its reason, purchase, etc.) 1 .
For the purpose of estimating the elasticity of mortgage origination to the loan to income ratio of the applicant we define origination as an application which has been accepted and then either originated or refused by the applicant, a purchase of a loan, or a pre-approved request. We define a non-origination as an application denied by the bank or a denied pre-request. We ignore all applications withdrawn by the applicants or applications closed for incompleteness. We also restrict the analysis to mortgage loans with the purpose of home purchase or refinancing, and we exclude home improvement.
The loan to income ratio is computed as the total loan amount in the application over the total gross annual income an institution relied upon in making the credit decision 2 .
The second important source of data is the balance sheet data. This data is used to assure the comparability of the treatment and the control group and for the purpose of controlling for bank characteristics in the final regression. The source of balance sheet data are the FDIC Call Reports. They are available at quarterly frequency for the population of FDIC insured banks. The subject of our analysis are the state chartered banks, all of which are also FDIC insured. Table 1 provides some descriptive statistics on both the population of FDIC insured banks and the subpopulation of state chartered FDIC insured banks. State chartered banks are on average smaller than the the average of all FDIC insured banks, both in terms of assets and the number of employees. They are also on average less profitable and operate with a lower net interest margin. They are similar in terms of the structure of their liabilities, but they do differ substantially regarding their loan portfolio where the state chartered banks focus much more on residential mortgage lending, which is evident from their high share of residential loans in total loans.
As will be explained in more detail below, we build a treatment group to resemble the Massachusetts state chartered banks. The share of non-insured deposits in This table contains the means of some descriptive variables for all the FDIC insured banks, the state chartered banks and the Massachusetts state chartered banks by years. Assets are total assets in millions of dollars, # Empl denotes the number of employees, Cap.rat. denotes the leverage (capital to assets) ratio, LTD denotes the loan to deposit ratio, roa denotes the return on assets, net.int.mar denotes the net interest margin, lta denotes the loan to assets ratio. residential denotes the share of residential loans in total loans, dep. to ass. denotes the deposits to assets ratio, retail denotes the share of retail deposits in total deposits, and insured denotes the estimated share of deposits insured by the FDIC.
the liabilities of Massachusetts state chartered banks and in the treatment group, around 20% of deposits, provides enough leverage to affect bank behaviour.

Methodology
The estimation of the causal effect of deposit insurance on risk taking follows the methodology of estimating the average treatment effect on the treated using the difference-in-differences estimation, where we regard treatment as an increase in deposit insurance coverage. The validity of the methodology relies crucially upon three steps. First, to assure random assignment of treatment, we make use of a natural experimental setting, where the limit of the deposit insurance coverage was raised for some banks but not for all. Secondly, to assure that banks in the treatment and control group are comparable, we resort to nearest neighbour matching based on covariates which might affect risk taking. Finally, to get a proper estimate on of the effect on risk taking behaviour, we construct a risk taking measure, based on new loan issuances, which differs from conventional risk measures, based on the existing portfolio of loans.

Identification -natural experiment
Policy actions regarding deposit insurance coverage, which stem from build-up of risk and bank behaviour, can render biased estimation of their causal effect on risk taking. To estimate the effect, a control group, unaffected by the policy due to reasons exogenous to their behaviour, is needed. In this paper, we exploit an experimental setup from the deposit insurance system in the US.
In the US, deposit insurance is carried out by the FDIC (Federal Deposit Insurance Corporation), which was set up in 1933. In addition to this Federal Deposit Insurance, since 1934 state-chartered savings banks in the State of Massachusetts are covered by a private deposit insurance company, the DIF 3 (Depositors Insurance Fund), which offers unlimited insurance on deposits of member banks. Membership to this insurance scheme is mandatory for all state chartered banks in Massachusetts.
On October 3rd, 2008, the FDIC increased the statutory coverage from $100,000 per depositor to $250,000. This was intended to be a temporary measure, but the decision became permanent on July 21st, 2010. Since DIF members 4 in Massachusetts always had unlimited coverage, they are unaffected by this change.
The fact that the state-chartered banks from the State of Massachusetts are unaffected by the increase in coverage enables us to use these banks as a control group in our estimation of the treatment effect. To assure similar regulatory framework, we focus our analysis on state-chartered banks only, and thus define the treatment group as state-chartered banks in other states where the policy would have an effect.
Using this framework, we use a difference-in-differences approach to examine how bank risk taking changes for the treated banks as a response to higher deposit insurance coverage, in comparison to that of the banks in the control group.

Identification -matching
With the control group defined by the fact that state-chartered banks in Massachusetts are unaffected by an increase in deposit insurance coverage, we are assured that the immunity to deposit insurance coverage increases is exogenous to the behaviour of these banks and thus randomly assigned. Equal regulatory framework is ensured by focusing on other than Massachusetts state-chartered banks to use as a treatment group. However, to achieve comparability of the treatment and the control group, the treatment group needs to be narrowed further to banks which share the characteristics of the control group.
Matching on observables before treatment allows us to obtain a list of banks which share the characteristics of banks in the control group. For each bank in the control group, we therefore find three matching banks in the pool of treated 5 . The matching methodology is nearest neighbor matching. We match on the pre-treatment averages of balance sheet size, leverage ratio, capital to asset ratio and deposit to loan ratio. For balance sheet size we match exactly. These variables have been shown in the literature to explain risk taking by banks. It is therefore crucial that they are balanced across the treatment and the control group. Table 2 shows the mean of some key variables for the treatment and the control group before the treatment date, together with the p-values for the two sample t-test for the means 6 . This table contains the means of some descriptive variables for the control and the treatment group before the treatment date. assets are total assets in thousands of dollars. car denotes the capital-to-asset ratio. Leverage is defined as total debt over total equity. loantodep is total loans over total deposits. Loan composition variables consist of the ratios of residential loans, commercial loans and individual loans to total loans. Deposit composition variables consist of total deposits in thousands of dollars, and the ratios of transaction accounts (demand deposits, etc.) and non transaction accounts (savings deposits, time deposits, etc), deposits below the pretreatment coverage limit of $100k, and deposits by other banks to total deposits. 5 The choice of number of matches is based on the fact that there are 50 banks in the control group. We aim to have more than one match for each in order to have a larger number of observations. Results are robust to other numbers of matches. 6 The null hypothesis is that the means are equal, hence a p-value higher than 0.05 indicates that the means for both groups are not statistically different at the 95% confidence level Table 2 provides evidence that both groups are similar in observables in the pre-treatment period. Given the list of variables that the matching is done upon, it is unsurprising that they have similar balance sheet size, capital to asset ratios and leverage ratios. In addition to that, their balance sheet composition is very similar. A large share of the loan activity of both groups is allocated to residential loans (i.e. mortgage loans), and most of their deposit base comes from retail deposits.
The total amount of deposits held by both groups is not statistically different between the two groups. Likewise, the difference in the share of deposits which are held in transaction accounts (with higher liquidity) or non transaction accounts (with lower liquidity) between the treatment and the control group is statistically insignificant. This indicates that both groups face a similar liquidity structure in their funding sources, which is important towards their ability for maturity transformation. Furthermore, it is worth noting that from December 2010, to December 2012, transaction accounts that pay no interest were entitled to unlimited deposit insurance coverage by the FDIC, as commanded by Congress. This could pose an issue for our experiment if the share of such accounts was a large enough part of our banks' balance sheets. This is not the case as shown in Table 2.
Other variables, such as the loan-to-deposit ratio, the share of residential loans to total loans, net interest margin and share of retail deposits to total deposits are statistically different, but qualitatively similar. One variable that is especially important is the share of total deposits below the pre-treatment limit of $100,000. We could expect that depositors with larger balances prefer to deposit them in MA banks, due to higher coverage. This could pose another issue for our experiment if, after the treatment, these depositors decided to move their funds to another bank. In that case the treatment would indirectly affect the control group through deposit supply. We address this issue in the Robustness section and show that it is not the case.
Furthermore, due to an overlap of the timing of the policy implementation and the start of the financial crisis, any significant differences in the characteristics between the two groups, would potentially plague our results due to a different reaction of the groups to the crisis. This exercise, however, confirms that the balance sheet structure of both groups is similar enough to ensure that when facing a shock, such as the financial crisis, they would be affected in a similar manner.
Finally, we note that some significant banking regulation changes, such as the Dodd-Frank Act, took place during the period of our study. However, insofar as these changes were done at the Federal level, they affect both the treatment and the control group equally and this is picked up by the difference in differences estimator.

A measure of risk taking
To analyse bank risk taking it is important to rely on a risk measure which is not tainted by past choices and current shocks to their portfolio. If instead a measure of risk based on the current portfolio is used, any asymmetry in the severity of the financial crisis could affect different portfolios differently and thus meddle with the effect which stems from new choices. As discussed earlier, a measure of risk taking based on new issuances of loans does not have this problem. Variation in such a measure results from the choices of banks with regards to the expected performance of loans and borrowers at the point of issuance and not on the ex post performance of loans issued in the past. To this end, we construct a measure of risk taking on the issuances of new mortgage loans based on the Home Mortgage Disclosure Act (HMDA) dataset as it provides information on the population of mortgage applications to banks and other mortgage lenders, including detailed information on the borrower and loan characteristics. Given the high share of residential loans in the portfolios of state chartered banks, which are at the focus of our analysis, we take the risk of new mortgage lending as representative of risk taking on the entire portfolio.
To construct a measure of banks inclination to take risk, we estimate a propensity to originate the loan given the loan risk characteristics. DellAriccia et al. (2012) and Ignatowski and Korte (2014), among others, have shown that loanto-income ratios are a good proxy for riskiness of loans. Following this idea, we measure the risk connected to the issuance of a loan and to the borrower using the loan-to-income ratio (LtI) measure computed from the HMDA data set for every loan application. We construct a new measure of risk taking through the following model 7 . 8 Origin t,i,j = γ 0 t + γ 1 t,i LtI t,i,j + t,i,j where Origin t,i,j denotes a binary loan origination variable which takes the value Origin t,i,j = 1 if application in period t to a bank i by a borrower j is accepted and loan is originated, and takes the value Origin t,i,j = 0 if the application is rejected and the loan is not originated. Time effects, γ 0 t , capture the effect of the macroeconomic situation in period t for all banks, such as market liquidity, regulation, and monetary policy, among other things. Finally, for every bank i in every period t we also obtain an estimate of the sensitivity of origination to the 7 In order for γ 0 t to capture the macroeconomic conditions affecting the origination choices, we estimate the model for all banks reporting to the HMDA data set but of course only use the γ 1 t,i for banks included in the final regressions. This implies including all the loan applications in the HMDA reporting in the estimations. The number varies between 17 million applications and 40 million application which constrains us to estimating the model as a linear probability model risk associated with a loan γ 1 t,i based on loan-to-income ratio of all applicants j from 1 to J, which serves as a measure of risk-taking by banks. Figure 1 plots the risk sensitivity parameters for the banks included in the analysis over the years.
While we believe that the risk taking measure, as constructed above, has deficiencies, we also believe that our results are robust to those deficiencies. Firstly, as is documented in Ferrari et al. (2018), out measure predicts well the county level default rate. Furthermore, the results using implicit deposit interest rates give credence to the conclusions we are inferring from the risk taking measure.

Difference-in-differences
In estimating the causal effect of the increase in deposit insurance coverage on risk taking we use the difference-in-differences approach. This implies estimating the following model: where D T is a dummy variable that takes value 1 for the treated and 0 otherwise, and D af ter is a dummy variable that takes value 1 for the treatment period and 0 otherwise. Hence β 3 , the coefficient for the interaction term between the treated dummy and the treatment period dummy, is the average treatment effect on the treated, i.e. the causal effect of the policy change on risk taking behaviour.
The key assumption in the difference-in-differences estimator is that, in the absence of treatment, both groups would have followed similar time trends. Figure  2 plots the means of the risk taking measure for the control and the treatment group. The figure indicates that the two groups behaved similarly in the pretreatment period regarding their risk taking. This implies that we are evaluating the effect of the policy on a comparable set of banks and that the results of the estimation of equation 2 will provide the estimate of the causal effect (or as the plot suggest the absence of one).
In our exercise to determine the cause of no causal effect, we will apply the same methodology to estimating the effect of an increase of deposit insurance coverage on the price of deposit funding for banks. The validity of the common trend assumption in the pricing variable is confirmed in section 5 which also provides the results.

Results -risk taking
We begin by providing the results of the estimation of equation 2. Columns (1) to (4) provide the results using several specifications. Columns (1) and (2) provide the results for the OLS specification and bank fixed effects specification. Since our focus is on the state-chartered banks, to control for possible regulatory and other institutional or structural differences across states, we include also statefixed effects (see column (3)). Furthermore, to control for a different severity of the crisis, we add a specification with state-year fixed effects (see column (4)). The dependent variable across all specifications is our risk measure, the sensitivity of loan origination to loan-to-income ratio, constructed as explained in the preceding section. In all specifications, the coefficient of interest, D T D af ter , corresponding to β 3 in equation 2, is insignificant. This suggests that an increase in deposit insurance coverage limit does not cause banks to take on more risk. The estimation is consistent with the graphical evidence from Figure 2, where in the post-treatment periods, the means of the two groups did not diverge.
The results of our empirical analysis do not confirm the results from the previous studies. Using an experimental setup and a measure of risk-taking that is based on new loans instead of balance sheet data, we find no significant effect of an increase of deposit insurance coverage limit on risk taking by banks. This result is important for policy making insofar as, when faced with the trade off between the risk of bank runs or the higher systemic risk induced by moral hazard, the regulator should take into consideration that the moral hazard channel may not be an issue.
It is worth pointing out that the treatment in our experiment is an increase in deposit insurance coverage, hence these results apply to the intensive margin. While it may be true that implementation of deposit insurance may increase the problem of moral hazard, it appears that moral hazard is not a concern when we consider increases in coverage. It should be stressed, however, that although this policy is increasing the insurance coverage and not establishing deposit insurance from zero, it does affect depositors with deposits in excess of $ 100,000, who tend to be more informed and knowledgeable in financial diversification. This limits the argument that the policy along the intensive margin would not provide enough of an incentive to apply market discipline.

The Crisis
As pointed out by Anginer et al. (2014b), deposit insurance can induce moral hazard during normal times, but its stabilizing effect may outweigh the moral hazard effect during times of financial turmoil. This would suggest that in periods of stress, stable funding allows banks to invest in safer assets. Furthermore, periods of financial stress correspond to periods of intense supervisory monitoring, which would limit the scope of banks to take on excessive risk.
In order to test whether the treatment effect is different during the crisis, we add a dummy variable for the crisis period and an interaction term between the crisis period and the treated group to regression 2 9 .
Here, β 4 reflects the effect of the crisis on bank risk, and β 5 is the coefficient for the treatment effect of deposit insurance during the crisis. Table 4 shows the results of the analysis including the crisis period interaction as described above. Column (1) corresponds to the standard difference-in-differences regression reported in the first column of Table 3, while Column (2) includes the crisis specification.
The coefficients of interest, D T D crisis , corresponding to β 5 in equation 3 is not statistically significant. This implies that the effect is insignificant in both crisis and non-crisis periods. This result is also important as it indicates that overall result of no effect of deposit insurance coverage limit is not due to the fact that in crisis and non-crisis subperiods the effects have an opposite sign as per Anginer et al. (2014b).
It is worth pointing out that the timing of the crisis and the that of the policy announcements poses some difficulties. Recall that it was in October 2008 when the FDIC announced a temporary increase in the deposit insurance coverage, which was supposed to last until December 2010. In July 2010, the FDIC announced that this measure would become permanent. Arguably, October 2008 was the peak of the financial crisis, with Lehman Brothers filing for bankruptcy in September of that year. Furthermore, by July 2010 the crisis in the US was coming to an end. The coincidence in time of the policy announcements and the crisis period makes it hard to disentangle whether the different treatment effect during these years is due to the financial crisis, or to its temporary or permanent nature. We address this issue later on in the robustness section.

Results -pricing and market discipline
We have so far established that an increase in deposit insurance coverage does not cause an increase in risk taking, disproving one of the more commonly cited potential pitfalls in providing deposit insurance or increasing its coverage. As theory predicts, the moral hazard effect of an increase in the coverage of deposit insurance is that deposit supply increases and depositors reduce their monitoring of bank behaviour. Increased deposit supply and laxer market discipline should, given a level of demand for deposit funding, decrease the interest rate on deposits.
To test whether an increase in deposit insurance coverage affects the deposit supply and market discipline we use the difference-in-differences estimation. Theoretical prediction states that an increase in deposit insurance coverage decreases deposit rates as a result of the deposit supply rise. We therefore run the following regression: where i t,i refers to the deposit rate of bank i in period t. As before, the coefficient of interest is β 3 , which provides the estimate of the treatment effect of an increase in deposit insurance coverage on the deposit interest rates. Since we do not have the data on deposit rates for new deposit transactions, following Schmukler (2001), we compute the implicit interest rate, i.e. the ratio of total interest expense on domestic deposits over total domestic deposits for each bank each year. In using implicit deposit rates, we may be overlooking deposit composition effects. It cannot be excluded that deposits with different maturities are subject to different rates. Averaging interest expenses across deposit classes may therefore be an issue. Figure 3 plots the means of the implicit deposit interest rates over the horizon for the control group and the treatment group as defined in the previous section.
The figure provides several insights. Firstly, the two series clearly exhibit a common trend in the pre-treatment periods, suggesting that the series are suitable to perform the estimation on. Secondly, the lack of divergence in the post-treatment periods already hints at the conclusions which are confirmed by our estimation results in Table 5. The coefficient of interest, D T D af ter , corresponding to β 3 in equation 5, is statistically insignificant, indicating that the increase in deposit insurance coverage does not increase deposit supply which would result in lower interest rate.
This result provides an explanation for the result of no response of bank risk taking following an increase in deposit insurance coverage, presented in the previous section. After in increase in deposit insurance coverage, the depositors do not increase their supply of funding to the bank which does not result in laxer market discipline. Our results are in line with those of Schmukler (2001).
6 Robustness 6.1 Indirect treatment effect on the non-treated through the deposit supply Our results rely on the assumptions of matching and the difference-in-differences approaches. We have established a treatment and a control group of banks that are similar in observables. However, to assure their validity, we need to make sure that the control group is unaffected by the treatment. The control group in our experiment comprises of banks whose deposit insurance is unlimited, which implies that they are unaffected by the country-wide increase in the national coverage limit from $100,000 to $250,000. Although we see no way for the control group to be affected directly by the treatment, there remains a plausible indirect effect through deposit supply. In this case depositors in Massachusetts state chartered banks, whose choice of bank was based on the fact that their funds are insured, could withdraw their funds in excess of $100,000 and opt for a bank which now also offers insured deposits but is more convenient along some other dimension. Figure 4 shows the average total balance of deposits for each bank, and how they are split between those which lie above the insurance limit and those below. The two kinks in the amount of deposits above and below the limit correspond to the change in policy. Note that with the increase in coverage, deposits between $100,000 and $250,000 were above the limit before the change, and are below the limit after, so the amount of deposits below the limit increases merely by accounting. The same logic can be applied to deposits above the limit. However, the total amount of deposits does not deviate from its trend. It can thus be concluded that the banks in the control group did not experience any deposit flight due to the policy change.

Results for different treatment dates
As stated before, the increase in deposit insurance that happened in 2008 was intended as a temporary measure. However, it was made permanent in July 2010. In order to make sure that the absence of a treatment effect is not due to the temporary nature of this measure, we estimate the same model with 2010 as a treatment date.
Furthermore, in order to ensure that the lack of effect is not due to anticipation effects, we use 2007 as a treatment year, to account for the case that some information regarding an increase in deposit insurance coverage circulated prior to the implementation. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 6 shows that the treatment effect is still insignificant for these treatment dates. Again, we find no evidence of a moral hazard problem that would result in higher risk taking by banks after an increase in deposit insurance coverage.

Conclusion
In this paper we estimate the effect of deposit insurance on risk taking by banks. It is a well established theoretical result that deposit insurance disincentivises depositors to monitor bank behaviour since it assures depositors access to their liquid assets independent of bank risk. This incentivises banks to take on more risk.
To test these predictions we use a natural experiment setting from the US deposit insurance system, where state chartered banks in Massachusetts were immune to an increase in deposit insurance coverage limit by the FDIC due to unlimited deposit insurance provided in local regulation. Furthermore, to avoid any biases arising from using risk measures based on the existing portfolios, we use mortgage lending data to construct a measure of risk taking based on new issuances.
Our results indicate that an increase in deposit insurance coverage has no effect on risk taking by banks. Furthermore, in exploring the reasons of no effect on deposit insurance, we find that an increase in deposit insurance coverage also has no effect on deposit interest rates. This indicates that deposit insurance does not relax market disciplining by depositors through an increase in deposit supply. These results imply that the usually quoted negative implications of deposit insurance are not an issue in practice and should therefore be disregarded when considering different deposit insurance schemes.