The role of capital in resolving agency conflicts between different groups of bank stakeholders.

1. Introduction
This essay discusses the role of capital in resolving agency conflicts between different groups of bank stakeholders, by discussing two papers. The first paper is a descriptive paper written by Berger, Herring, and Szego (2005), called “The Role of Capital in Financial Institutions”. The second paper, “Caught in Between Scylla and Charybdis? Regulating Bank Leverage When There Is Rent Seeking and Risk Shifting”, is a theoretical paper by Acharya, Mehran and Thakor (2013). Both articles examine agency conflicts between different bank stakeholders and how capital could be used to reduce these conflicts.
The rest is the paper is structured as follows. Chapter 2 discusses the paper of Berger et al., starting with the question why markets require financial institutions to hold capital, and why financial institutions differ from other companies. Next, we discuss why regulators require capital and why the market is not self-regulating, followed by the unintended consequences of regulation. Chapter 3 discusses the paper of Acharya et al., which starts by discussing two important moral hazard problems. Next, the base model is examined, followed by the extended model. Finally, we conclude by summarizing our findings and discuss the validity of both papers.

2. The role of capital in financial institutions
The theoretical paper of Berger, Herring and Szegö (1995), examines the role of capital in financial institutions, and focuses on how market generated capital requirements differ from regulatory capital requirements. The authors start with the Modigliani & Miller (M&M) proposition to determine market generated capital requirements, and discuss the departures of the frictionless assumption. One important friction, the safety net, will lead to agency problems and the paper examines how capital could be used to reduce these conflicts. Further, this section examines historical trends in bank capital, problems in measuring capital and some possible unintended consequences of capital requirements.

2a. Why do markets require financial institutions to hold capital?
We start with defining the market capital requirements as the capital ratio that maximizes the value of the bank in the absence of regulatory requirements. According to the frictionless model of Modigliani & Miller, leverage should differ randomly across sectors, since the capital structure could not affect firm value. We will first consider several market imperfections that helps explain the market capital requirements as taxes, financial distress costs, asymmetric information and transaction costs followed by a bank specific imperfection; the safety net.
Since interest payments are tax deductible and dividends are not, shifting payments to debt holders will increase firm value since it lowers the payments to the tax authorities. Increasing debt however, will increase the potential costs of financial distress. Therefore, an optimum debt level could be determined by offsetting the advantages of additional debt to the costs of additional debt in the presence of these two frictions. Second, the M&M propositions relies on the assumption that full information is available. Since commercial banks are specialized in lending to information-problematic borrowers, a relaxation of this assumption is needed. Banks will acquire information by the loan screening process, and augment this information when monitoring the loan. This private information creates asymmetric information between banks and financial markets. Ross (1989) shows that the market will draw inferences from the bank’s capital decisions. When it is less costly for ‘good’ banks to signal the market through higher leverage, a signaling equilibrium may exist where those banks with higher future expectations have lower capital ratios. The literature is, however, ambiguous since Acharya (1988) shows a signaling equilibrium with higher capital ratios.
When there is asymmetric information in combination with transaction costs, a problem might arise when managers have private information, causing shareholders to be reluctant in issuing new equity because it may sell at a discount. This discount is partly caused by the difference in transaction costs of issuing new debt or new equity and by the relative costs of internal and external finance.
Asymmetric information could also lead to agency problems between several stakeholders when the bank faces financial distress. Shareholders may find that actions that increase the value of all stakeholders do not necessarily increase their value. This may lead to attempts to shift wealth from creditors to shareholders, of which four agency problems are identified. First, this might lead to a moral hazard opportunity for shareholders when they substitute riskier assets for safer ones if assumed that creditors lack sufficient information. Second, when a bank is near default, shareholders may lack incentives to add new capital for positive NPV projects, since most of the proceeds will go to the creditors. Third, shareholders could continue operations beyond the point where the bank should be liquidated, in order to keep an option value for the claims and finally, shareholders could manipulate accounts to mask deterioration by downplaying its loan losses or by gains trading. As a consequence, creditors will demand a higher return to compensate for the expected value of these expropriations. In response, the bank could increase their capital ratio, to assure creditors that the bank is safe and to assure that shareholders and creditors are more closely aligned. Other agency costs arise from the conflict of interest between shareholders and managers. When shareholders cannot perfectly monitor the managers, more debt to the company could put pressure on the manager to generate cash flows. Managers are then better aligned with shareholders, giving them more incentives to work harder and make better investment decisions, since the manager will lose human capital value in the event of a default. Therefore, in contrast with the shareholder-creditor agency problem, higher leverage increases the discipline of managers and reduces agency costs. Compensating managers with shares and oblige them to hold the shares would further increase these incentives.
A bank specific deviation from the M&M proposition is the existence of the safety net for commercial banks. We want to consider how the safety net will affects market capital requirements in the absence of regulation. Governments will intervene when commercial banks collapse to reduce the lack of trust for depositors and to prevent bank runs. Since creditors know that the government will intervene, the safety net reduces the market capital requirements due to the lack of discipline. This could be simply eliminated when the deposit insurance premium reacts fully on risk. The risk adjustment was however only limited up to the time of writing (1995). Besides, the safety net may cause reduced costs of unsecured debt since the market believes that the bank is safer due to the insurance and that the bank could earn higher cash flows due to the implied subsidy by the government. As a consequence, the safety net introduces a new agency problem between the government and the bank.

2b. Why do regulators require financial institutions to hold capital?
Now we have discussed why the market requires capital, we will discuss why regulators require financial institutions to hold capital. First, society should be protected from the costs of financial distress, agency problems and the reduction in market discipline caused by the safety net. Second, although the main concern of the regulator is systemic risk, when multiple banks get into trouble, it is difficult to see whether the cause is systemic or idiosyncratic. Since the interbank market is important and large, a single default could start a chain reaction of defaults. Third, private companies would set the insurance fees based on the risk. However, the government does not. The regulator have some pressure possibilities as cease-and-desist orders, total withdrawal of insurance coverage, bank closure, limits on asset growth and brokered deposits and prohibition of dividend payments. However, these tools are blunt and uncertain and regulation is needed.
One obvious remedy to prevent financial distress or agency problems is to require banks to hold so much capital that a default is negligible. However, since we argued that there are deviations from the M&M proposition, the firm value would decrease and the weighted cost of financing would increase. These costs would be passed on to customers in the long run and therefore would impose social costs. According to Santomero and Watson (1977) capital regulation involves a tradeoff between the marginal social benefit of reducing the risk of negative externalities from bank failures and the marginal costs of diminishing intermediation. Setting these marginal benefits and costs equal creates ideal regulatory capital requirements. However, since this ideal system is prohibitively expensive, the regulator stipulates uniform, minimum ratios below which banks are subject to regulatory sanctions.
Not every instrument would decrease the agency problems and criteria have to be assessed. Instruments that qualify for regulatory capital should have three main characteristics. First, claims should be junior to those of the deposit insurer, so that they serve as a buffer to absorb losses before the government. Second, the instruments should count as ‘patient money’. It should not be redeemable without assured refunding so that it could be used as a stable source of funding during a possible bank run. Third, it should reduce the banks moral hazard incentives to exploit the protection of the safety net by undertaking excessive portfolio or leverage risk. The paper examines three instruments; equity, subordinated debt and uninsured deposits. First, equity is junior to all other claims and therefore serves well as a buffer to protect the government from losses. The maturity of equity is indefinitely long and since regulators often prohibit distressed banks to pay dividends, the funds cannot be redeemed during a crisis period. However, regulatory requirements to increase the equity ratio would reduce leverage risk, but the effect on overall bankruptcy risk and portfolio risk is ambiguous. Second, subordinated debt is junior to all claims other than equity and therefore serves as a buffer against losses to the government. Subordinated debt has a typically large maturity and is difficult to redeem in crisis periods. Although subordinated debt increases leverage risk, it may deter portfolio risk taking. Since the subordinate debt perspective is similar to the deposit insurer and cannot fall back on the safety net, subordinate debt holders will actively monitor the bank. Despite this theoretical framework, empirical research found only limited response on risk measures. Finally, we consider the use of uninsured deposits for regulatory capital. The first criterion does not hold since deposit holders have the same seniority as the deposit insurer. Second, uninsured deposits do not count as patient money since these deposits are typically redeemed in the event of a bank run. In conclusion, equity and subordinated debt broadly satisfy the criteria of regulatory capital, while uninsured capital does not.
When measuring these instruments of regulatory capital, the author poses some significant measurement problems. The numerator of this ratio is defined as ‘regulatory value of equity’, which depends on the market values of all on and off balance sheet assets and liabilities adjusted for limited liability. But these values are difficult to measure, as banks are specialized in lending to information-problematic borrowers and may be subject to manipulation by ‘gains trading’.
The denominator, defined as ‘variability of the regulatory value of equity’, is also hard to measure. As capital adequacy depends on the ratio of capital to the risk it is able to absorb, the denominator should measure the bank’s risk exposure. However, it’s hard to develop an accurate measure of risk exposure that is simple and can be uniformly applied across banks. Also, regulatory measures of risk exposure may be subject to manipulation by bank management.

2c. Unintended consequences of regulatory capital requirements
Although regulatory capital requirements reduce the agency problems, they also have some unintended consequences. Banks have to increase capital after the origination of a new loan to keep the capital ratio equal. Since this is costly, banks could also sell the loan contract to another party, along with the responsibility for monitoring and bearing credit risk. Off-balance sheet guarantees were not subject to capital requirements in the 1980s, so that banks could reduce the effective capital requirements by shifting from loans to these off-balance sheet guarantees. These guarantees also increase the capital ratio, since the guarantees are senior to creditors and therefore the cost of debt increases, but the increase in capital ratio is less lower than keeping the loans. Therefore, the regulatory capital requirements played an important role in the expansion of securitization in the 1980s. Second, in the early 1990s the regulators imposed risk-based capital (RBC) requirements, meaning that more risky assets have higher effects on the capital requirements. Since commercial loans are assigned a 100% risk weight and treasuries a 0% risk weight, there was a shift from commercial loans to treasuries large enough that some to be termed a ‘credit crunch’.

3. The Role of Capital in Resolving Agency Conflicts between Groups of Stakeholders
In the second paper, Acharya et al. formally structure two moral hazard problems in a model and extend this basic model with issues as shadow banking and sequential failures. They discuss optimal capital requirements from both the bank and the regulators perspective and discuss the lender of last resort consequences for the model. Finally they review the regulatory implications of their model and proposals. The proposals by Acharya et al. are aimed ‘’at increasing bank capital in a way that it does not compromise bank discipline by uninsured creditors and yet keeps in check bank incentives to take excessive leverage and risks that are correlated with those of other banks.’’
When ownership and control at companies are separated, agency conflicts often arise. Two moral hazard problems that banks face are rent seeking by managers who under provide their loan monitors and asset substitution where banks chooses excessively risky and socially inefficient portfolio’s to increase private wealth. Although there are ways to address these moral hazard problems, there is an inherent conflict between the two solutions, making it difficult to decide what the optimum should be. The rent seeking problem could be resolved by increasing debt. When debt levels are low, creditors are quite certain that they can recover their principal and creditors will lack incentives to monitor. High debt levels increase the risk for creditors and they are forced to closely monitor the bank, and liquidate in the case that managers shirk. Controversially, the asset substitution moral hazard problem can be solved by lowering the leverage ratio. This will increase the banks skin in the game and thus ensure that the bank will not take excessive risks.

3a. The base model
Acharya et al. start with a base model that consists of a bank’s owner who’s wealth is constrained in t=0 and needs external financing, either equity or debt. When provided with capital, the bank decides at t=1 whether to invest in a good loan portfolio (G) or an aggressive and excessively risky portfolio (A). The bank can also choose whether to monitor the loans or not. Given monitoring, the likelihood of success of the G portfolio at t=1 dominates the A portfolio, which in turn dominates the unmonitored portfolio (either G or A, since unmonitored portfolios are indistinguishable). However, at t=2, cash flows from portfolio A dominate G, which in turn dominate the not monitored portfolio.
From a social perspective, portfolio G with monitoring dominates any other choice, but the bank manager seeks optimal shareholder value net of his own monitoring cost. No external financier, debt or equity, can observe whether the bank chooses A or G and whether it is monitored. At t=1 it becomes clear whether the loans have been monitored. Creditors may choose to liquidate the bank and equity holders can fire the manager, both resulting in a new value of the bank (L), which is lower than the continuation value of the bank without monitoring. Besides the bank and the external financiers, the authors also introduce the regulator, the so-called Lender of Last Resort (LOLR). The regulator’s objective is to avoid the ex-post social cost of an industry collapse (Ʌ) and to stimulate the (socially) efficient portfolio choice by banks in t=0. The LOLR will bail out banks only when all banks fall together. 3b. Equilibria
Acharya et al. find a first best equilibrium that arises when the monitoring effort of the manager is contractible. In equilibrium, the bank will choose portfolio G with monitoring, and no scenario’s occur where creditors or shareholders will liquidate/fire. Unfortunately, a managers monitoring effort is not contractible so that a second best equilibrium has to be solved. To solve the second best equilibrium, we will first consider three lemma’s and some conditions.
First, lemma 1 says that for a large enough external investment, managers prefer to shirk as long as they are not threatened by liquidation of dismissal.
Second, when the debt level is too low (<D), creditors will unconditionally choose liquidation at t=1, as they prefer sure payoff relative to risky continuation. When debt is too high (>D0), creditors start to behave like shareholders and choose for risky continuation, since they are not so sure whether they will be repaid at liquidation because of the high debt levels. Only when the debt level at t=0 is in between these extremes, such that (DR∈(D,D0), creditors let their choice of liquidation depend on monitoring signals at t=1. Shareholders will not fire the manager since the liquidation value is smaller than the risky continuation value. Therefore, we can conclude that debt disciplines the manager, but equity does not.
Third, if the bank chooses G with monitoring, a certain repayment level must be promised to creditors at t=2 in order to raise an amount of D at t=0. Assuming that D > D (constraint 11), the second best equilibrium with private contracting involves the bank issuing debt between (D,D0). The manager monitors and the creditors never liquidate at t=1 in equilibrium.

3c. Proposition 1
Next Acharya et al. introduce the LOLR into the model and present proposition 1. Suppose that constraint 11 holds and that LOLR is perceived by banks as adopting a policy of bailing out all banks if they fail together, then two Nash equilibria exist. The first equilibrium is socially efficient, since all banks choose optimal debt ∈(D,D0), and choose G with monitoring. In the second equilibrium, all banks choose A and maximize leverage consistent with loan monitoring (looting equilibrium), since banks can socialize their incremental risk with the bailout subsidy. Therefore, this Nash equilibrium is socially inefficient. If banks are permitted by the regulator, they will choose the second, however the LOLR can easily eliminate this equilibrium by using a simple capital requirement that limits the banks debt. When constraint 11 does not hold and an inefficient equilibrium is to be avoided by the regulator, a simple capital requirement will no longer suffice. A capital requirement is needed that eliminates the social cost (Ʌ), ensures selection of loan portfolio G and ensures that the manager monitors. Acharya et al propose that the LOLR allows a certain level of debt, but also equity must be raised. This ‘’special’’ equity must be invested in a riskless security from which payoff will flow to the shareholders in normal times, but in case of individual idiosyncratic failure will accrue to the LOLR. This special capital account has two advantages. First, in case of failure the LOLR can use the proceeds to fund its administrative costs or even transfer them to surviving banks by lowering taxes and second, the special capital requirements can be set as high as the LOLR deems necessary without hurting the monitoring incentives of creditors. In proposition 1 it is the combination of what happens in the portfolio success state and the non-systemic failure state that allows both risk shifting and rent seeking to be addressed with capital requirements.

3d. Extended Model
The previous approach shows that if capital requirements, including “special” capital requirements, are properly designed, correlated bank failures will not exist in equilibrium and there is no need for any ex post bailouts. However, the model ignored two related issues that complicate the bailout decisions of the regulators.
The first one is described as “regulatory arbitrage” in the shadow banking system. The authors assumes that there are two types of banks, which are impossible to distinguish for regulators and creditors: normal regulated banks and shadow banks. The latter which are not subjected to the capital requirements and thus are free in choosing their leverage. These two banks differ in their choice of portfolios. Where shadow banks are locked up in portfolio A and are unable to monitor their borrowers, normal banks, given their regulatory capital requirements, optimally choose portfolio G. The emergence of shadow banks lets the regulated banking system avoid the regulatory costs (capital requirements), but still enjoy the benefits of the bailouts in case of failure, as they form a threat to healthy regulated banks because of spillover risks. The analysis will show that with the existence of these shadow banks, the design of capital requirements for regulated banks must anticipate the adoption of a sub game-perfect bail out policy by the regulators.
The second issue is that the base model assumes that during a systematic shock, and banks choose portfolio A, they will all fall at the same time. However, in reality banks will fall sequentially. When the first bank falls the LOLR will not know whether the failure is due to an idiosyncratic or systematic shock, and whether the bank should be bailed out or not. This means the LOLR faces a trade off (noisy inference problem) in deciding whether to bail out or not. Bailing out in the early state bears the risk that the shocks weren’t systematic and so taxpayers money is wasted. However, not bailing out the early-failing banks is bearing risk that the shock was indeed systematic and the entire banking system fails. Acharya et al assume that a bailout will only occur in t=2 and if it prevents contagion to other banks. The LOLR will either bail out all failing banks at t=2 or do nothing and wait until t=3, when all banks will fail.
Under the restrictions that the expected value of the unmonitored loan exceeds the liquidation value, the following three lemma’s lead to the second proposition. Lemma 4 states that if creditors assume that normal banks choose portfolio G and the bank issues a certain amount of debt such that the debt repayment falls in the optimal interval (lemma 2), then creditors will only liquidate the bank if they notice managers are not monitoring.
Adding another restriction which ensures that the incentive compatible debt level lies within the positive internal of the two extremes (as described in section 3b). This leads to lemma 5: as long as the incentive compatible debt level holds, conditional on monitoring, managers of the normal banks prefer portfolio G over portfolio A. Lemma 6 shows that the repayment that normal banks must promise to raise the desired amount of debt at t=0 is equal to debt divided by the probability that the bank will fail and be bailed out. This promised amount is higher than in lemma 3. This is because with the presence of shadow banks, the whole banking system becomes riskier.

So adding these two issues to the model, the bailouts have now somewhat other implications then before. This is because when there are no bailouts and given the high likelihood of correlated failures, normal banks become unable to generate expected pay-offs high enough to fund their loan portfolio. This will make their participation dissatisfied. However, when the regulators bail out failing banks with probability 1, then the looting equilibrium will arise again. The LOLR now faces a more proper consideration in which there is a combination of selective bailouts and capital requirements. 3e. Proposition 2
We assume that the optimal debt level as defined by lemma 4, where creditors get the incentive to threat with liquidation if managers don’t monitor, is less than the value of the incentive compatibility constraint for the manager to prefer G over A (constraint 20). If with a certain probability the regulator will bail out and constraint 20 holds, then the regulator imposes a capital requirement that the bank requires to issue debt at a certain level. This level of debt satisfies the optimal repayment promise to the creditors. All financing needs in excess of this optimal debt level must be met with equity. If constraint 20 does not hold, then the regulator requires ‘’special capital’’ and allows any level of debt. Comparing both propositions, the LOLR now uses a state-contingent bailout policy in combination with capital certain requirements. The capital requirements preserve social-efficient leverage and portfolio choices, while the selective bailouts keep the banking system from collapsing.

3f. Proposition 3
In the last part of the paper, the authors endogenize the bailout probability in the model. As mentioned before, the LOLR’s decision to bail out failing banks depends on a tradeoff between the benefit and cost of the intervention. Conditional on the observing failing banks, the regulator’s belief of that the systematic failure state has occurred, is crucial in the trade off. The larger the size of failing banks, the more likely it is that this state has occurred. This leads to the last result. The LOLR bails out when they observe an adequate amount of failing banks, which exceeds a critical number of failing banks. This critical number of failing banks is tested by Bayesian statistical inference approach.
Looking at the important implications for regulatory requirements in the analysis, implementing two-tiered capital requirements (proposition 1 and 2) has numerous advantages. First, the requirement approach deals simultaneously with several moral hazard problems, like risk seeking and risk shifting. Second, shareholders/managers may face losing their “special capital” in bad states. This ensures the positive aspect of high capital is maintained. Third, shareholders are not required to have some cash capital when liquidity is very low. This also leads to the fact that dividends could be retained at time of no danger of failure. Fourth, there is no need for designing “crisis triggers”. Last important advantage, the two-tiered capital requirements in co-operation with selective bailouts satisfy the participation constraints of the normal bank. This means, the temptation to set up shadows banks will be eliminated. To conclude, the base model determines the optimal bank capital should hold to eliminate different kinds of moral hazard problems. It shows that under some conditions, there is an optimal capital structure that solves the mentioned bank-specific moral hazard problems. But when these conditions do not hold, the goal of having the market discipline of leverage clashes with the goal of solving the asset-substitution moral hazard. In this case, the capital structure of the bank must tolerate some form of inefficiency and the value of the bank is not maximized. However, when banking failures become correlated and the entire banking sector could become in danger, debt creditors may be protected by the LOLR. The base model shows that the anticipation of regulatory forbearance (LOLR) leads to inefficient choices of leverage. Special regulatory capital requirements can address this problem. The authors extend the model with two features, namely the presence of sequential failures and shadow banks. The extended model shows that both a bailout policy and special capital requirements are needed.

4. Conclusion
This paper examines the role of capital in resolving agency conflicts between different groups of bank stakeholders, by discussing two papers. First, Berger, Herring and Szegö (1995) help define several general and bank specific agency problems and discuss the role of capital in resolving agency problems. The authors distinguish three different agency problems: a shareholder-creditor, shareholder-manager and bank-government agency problem. The paper describes how these agency problems affect the capital ratio and how regulation could help. First, shareholders can shift wealth from shareholders to creditors, and could commit actions at near-default scenarios that negatively influence creditors, and shareholders could manipulate accounts to mask deterioration. The bank could increase their capital ratio to assure creditors that the bank is safe, and to assure that shareholders and creditors are more closely aligned. Second, the shareholder-manager agency problem occurs when shareholders are not able to perfectly monitor managers, and managers don’t use best efforts to increase the company’s performance. Increasing leverage could reduce these agency costs, since with higher leverage managers have to generate higher cash flows to pay interest obligations. Finally, since commercial banks are secured by the safety net, shareholders and creditors tend to lack discipline in controlling risk. Since the market is not self-sufficient in this case, regulatory capital requirements are needed. Although tailor made requirements are best, this is not feasible and a regulatory minimum capital requirement is used. These minimums are subject to several criteria defining assets and capital. Finally, the authors discuss that the imposed regulatory requirements had some unintended side effects. First, banks made a shift to off-balance sheet banking since this reduces the capital requirements. Second, due to the introduction of risk-based capital, there was a shift from commercial loans to treasuries, large enough to be classified as a credit crunch. The paper of Acharya et al. has developed a theory on how an optimal bank capital structure can address three bank-specific agency problems. First, the shareholder-bank agency problem called rent seeking by underperforming managers could be deterred by increasing debt, as also discusses by Berger et al. Second, the creditor-shareholder agency problem asset substitution can be addressed by lowering debt. The authors structure these problems formally in a model and mention a third agency problem between the LOLR and the bank, that arises because of ex post-regulatory intervention, the so-called safety net. Since this intervention is anticipated, the bank will lack discipline resulting in over-levering, selecting socially inefficient and excessively risky portfolios, and paying out surplus debt as dividends. There exists an inherent conflict between the solutions for the first two agency problems. To deal with this difficulty the authors recommend a two-tiered capital requirement, which extends our discussion as in the paper of Berger et al. Besides a simple minimum equity capital requirement, banks should form “a special capital account” that only accrues to creditors in case of en masse failure. In case of individual failure, the creditors cannot claim this capital and it will accrue to the LOLR. The two-tiered capital requirement proposed by the authors will on one hand preserve the market discipline of debt and deter managerial shirking on monitoring, even in the presence of the safety net. On the other hand the leverage will be low enough to induce the bank’s shareholders to avoid excessive risk. We believe Acharya et al.’s model is valid to a great extent. Various papers and studies confirm the solutions to the two moral hazard problems mentioned in the first part of the article. Also studies from the 2007-2009 crisis leads to similar conclusions. By setting a two tier minimum capital requirement they nicely capture both solutions. One could argue that the model is too simplified and that too many assumptions/conditions are set. However, we believe that most assumptions and conditions introduced are realistic and not a threat to the external validity of the model. Also the authors describe the regulatory framework in a number of scenarios, when certain key conditions hold or not. The policy implications seem realistic and relatively easy to harmonize internationally, or at least as easily as the current tier-1 capital requirements. The two-tiered capital requirement not only deals with the aforementioned moral hazard problems, it also eliminates the need for shareholders to infuse additional cash in moments of low confidence and liquidity. To avoid banks of off-balance sheet banking, the authors recommend employing a state-contingent bailout policy.

--------------------------------------------
[ 1 ]. D, defined in (10) as the incentive compatibility constraint for the manager to prefer G over A, given monitoring.
[ 2 ]. (Calomiris and Kahn (1991) and Diamond and Rajan (2001)), Bhattacharya, Boot and Thakor (1998), and Merton (1977) (Myers and Rajan (1998)