Abstract
This paper reviews the market failures that may justify the need for liquidity regulation, assesses whether liquidity regulation is a necessary complement to Lender of Last Resort (LOLR) policies, capital regulation and prompt corrective action, surveys the available evidence on the net benefits of liquidity regulation, and concludes by outlining research directions useful to improve bank regulation design.
1. Introduction
Illiquidity and freezes of many markets were the distinguishing features of the financial crisis of 2007–9. Drops in asset prices caused an erosion of banks’ capital, which induced banks to reduce lending and to tighten lending standards. Financial intermediaries with severe liquidity mismatches found it difficult to rollover their short-term debt and, in response, reduced leverage through fire-sales, which led to further price drops and reductions in lending. At the same time, prompted by expectations of limited future access to funding, banks hoarded liquid assets, leading to the virtual shutting down of interbank and other debt markets (see e.g. Brunnermeier, 2009).
The immediate policy response to this crisis was a massive injection of liquidity by central banks through Lender of Last Resort (LOLR)-type facilities. At the same time, the perceived inadequacy of the pre-crisis regulatory framework to cope with large financial shocks prompted significant bank regulation reforms embedded in the new Basel III framework. In addition to the introduction of heightened capital requirements, Basel III introduces new liquidity regulation restricting the liquidity composition of banks’ assets and liabilities.
Tirole (2011) surveys a large literature on liquidity, its measurement at an individual and aggregate level, as well as its role in crisis periods. Based on a selective review and update of this literature, this paper addresses three key questions regarding liquidity regulation. First, what are the market failures that may justify liquidity regulation? The identification of these failures is instrumental in how to best design regulatory tools to correct them. Second, is liquidity regulation a necessary complement to other policies to address identified market failures, such as capital requirements, prompt corrective action, or public provision of liquidity, such as LOLR facilities? An answer to this question may inform how best to map different bank regulations into specific policy targets. Third, what are the potential costs of liquidity regulation? An assessment of these costs is necessary to evaluate how liquidity regulations can be designed jointly with other regulations to maximise their benefits.
The remainder of this paper is composed of three sections. Section 2 reviews the market failures that can trigger liquidity crises and justify bank regulations. Section 3 describes the policies that can mitigate or eliminate such market failures, including LOLR policies, capital requirements, liquidity requirements, and their interactions. Section 4 briefly reviews the current Basel III liquidity regulation, surveys the available estimates of benefit and costs of the envisioned liquidity requirements, and outlines research directions instrumental in improving bank regulation design.
2. Liquidity crises as market failures
Diamond and Dybvig's (1983) (DD hereafter) seminal paper and its many extensions 1 identify two important special roles for banks and the associated fragility when all wealth in the economy is intermediated through a banking system. First, banks invest short-term depositors’ funds into long-term assets. This maturity transformation is welfare-enhancing, since it improves the productive efficiency of an economy by channelling funds to higher return long-term investments rather than to lower return short-term assets. Second, banks offer depositors demand deposit contracts. By pooling depositors’ funds, banks provide them with insurance against unexpected liquidity needs for immediate consumption, improving the economy's risk-sharing opportunities. Yet, when the banking sector holds costly-to-liquidate long-term assets financed by short-term demandable debt, runs may occur in the form of panics triggered by (uninformed) depositors’ expectations that all others will withdraw prematurely, or in the form of information-based runs prompted by negative signals about banks’ default risk. When runs spread through the entire banking system, they cause a liquidity crisis that impairs banks’ welfare-enhancing maturity transformation and risk-sharing services.
However, in an extension of the DD framework, where banks are exposed to both aggregate liquidity and credit shocks and are subject to a sequential service constraint on demand deposit payments, De Nicolò (1996) shows that no liquidity crisis would occur if banks offer deposit contracts with payoffs contingent on the occurrence of premature withdrawals of subsets of depositors. In essence, this specific design of deposit payoffs makes the deposit contract (almost) complete in the sense of having payments contingent on all states of nature relevant to attain an efficient allocation. Moreover, even if deposit contracts are incomplete, Allen and Gale (1998) show that liquidity crises can be part of an optimal arrangement, since the resulting changes in repayment terms following bank failures make the deposit contract effectively contingent on the realisation of the ‘crisis’ states. Thus, in these models the occurrence of a liquidity crisis is not necessarily associated with a market failure.
In extensions of the DD framework to general equilibrium economies where both banks and financial markets coexist, Allen and Gale (2004a, b) sharpen the identification of market failures potentially leading to liquidity crises. They show that if markets are complete – in the standard sense of markets where traded securities span all states of the world (i.e., Arrow securities) – then the resulting allocations are efficient. If both markets and deposit contracts are complete, liquidity crises do not occur. If markets are complete but deposit contracts are incomplete, liquidity crises do occur, but their occurrence is consistent with an efficient allocation, as in Allen and Gale (1998). Under incomplete markets, however, a liquidity crisis would force intermediaries to undertake fire sales of their long-term assets to honour debt holder claims. As a result, prices of long-term assets would fall below their fundamental values, leading to inefficient ‘fire sales’, which would further threaten intermediaries’ ability to repay debt, thus leading to their insolvency. Therefore, a key source of market failures leading to liquidity crises in this type of model is the incompleteness of financial markets. In a general sense, these models suggest that the best policies to attain optimal levels of maturity transformation and risk-sharing are those designed to introduce elements of ‘completeness’ in markets and contracts that are somehow unattainable by private agents. Importantly, in these models intermediaries might hold either excessive or insufficient amounts of investments in liquid assets depending on preference and technology parameters, leading, respectively, to excessive or insufficient maturity transformation relative to the optimum.
Research on the mechanics of liquidity crises spiked up in the aftermath of the 2007–9 financial crises, focusing on the connections between rollover risk, fire sales, market freezes, the role of collateral, and concentrated exposures. De Nicolò, Favara and Ratnovski (2012) identify fire sales externalities, strategic complementarities, and network externalities as the externalities at the root of liquidity crises emphasised by the literature. Fire sales externalities arise when banks’ investment decisions do not necessarily internalise the sharp declines in asset prices and the reduction in the value of collateral arising from generalised sell-offs of financial assets. These externalities extend to funding decisions, when banks fail to internalise their reduced capacity to rollover their debt during a liquidity crisis. 2 Strategic complementarities arise when banks choose strategies whose payoffs increase with the number of other agents undertaking the same strategy, leading to investment in assets with correlated risks, which further exacerbate the adverse effects of a liquidity crisis (see e.g. Acharya, 2009). Network externalities arise from bank interconnectedness, which may give rise to contagion risk arising from banks not internalising the potential for liquidity risk arising from concentrated exposures within a banking network. An important common feature of the models in this literature is the strict connection between banks’ funding and lending decisions, which suggests the existence of strong interactions of different bank regulations on banks’ choices.
Liquidity crises have been featured recently in the literature on general equilibrium macroeconomic models with financial constraints. In their survey of this literature, Brunnermeier, Eisenbach and Sannikov (2013) introduce a useful taxonomy of three interrelated notions of liquidity. Technological liquidity refers to the degree of reversibility of investment in physical capital. Market liquidity is associated with the degree to which physical capital, or claims on its payoffs, can be traded with limited impact on their prices. Funding liquidity is associated with the maturity structure of debt, as well as with the sensitivity of the margins required for collateralised debt to variations in the value of collateral. Liquidity crises can arise from the interaction of liquidity mismatches between the technological and market liquidity on banks’ asset side, and the funding liquidity on banks’ liability side. The implications of these interactions for liquidity crises are the focus of the models by Brunnermeier and Pedersen (2009) and Brunnermeier and Sannikov (2014). In Brunnermeier and Pedersen (2009), unexpected shocks leading to a fall in funding liquidity induce banks to recur to fire sales, which in turn cause a reduction in market liquidity of their assets. When banks’ assets are used as collateral for banks’ access to funding, an adverse feedback loop is put into motion, since decreased market liquidity and tighter funding liquidity amplify each other. Brunnermeier and Sannikov (2014) study the equilibrium dynamics of an economy where intermediaries monitor borrowers as in Diamond (1984) and Holmstrom and Tirole (1997) and are exposed to aggregate risks on both sides of their balance sheets. They show that in response to significant unexpected losses, banks recur to fire sales, adversely affecting asset prices and triggering amplification effects that, in reducing funding liquidity, lead to a liquidity crisis fuelled by mutually reinforcing declines of intermediaries’ and final borrowers’ net worth.
In sum, the literature identifies market failures as generally stemming from market incompleteness, incomplete contracts arising from agency problems and asymmetric information leading to a variety of financial frictions. Fire sales externalities, strategic complementarities and network externalities imply that banks’ investment and funding decisions fail to internalise the price effects and the amplification loops leading to debt markets freezes and ultimately to a liquidity crisis. Absent corrective policies, financial intermediaries take on excessive asset return risk and excessive liquidity mismatches, exposing themselves to excessive default and liquidity (rollover) risks.
3. Policies
Policies addressing liquidity crises can be classified in two categories: those that aim at mitigating the adverse effects of such crises once they occur, such as central bank policies, and those that aim at preventing such crises, such as regulation and/or prompt corrective action.
Central Bank policies
Central banks’ Lender of Last Resort (LOLR) policies have been viewed as essential in managing liquidity crises since Bagehot (1873). 3 Research on LOLR policies has intensified in the aftermath of the 2007–9 crisis, during which LOLR-type measures evolved from the easing of lending terms at central bank facilities to provisions of liquidity to bank and non-bank institutions, as well as to system-wide liquidity support to debt markets.
Bhattacharya and Gale (1987) is an early study providing a rationale for central bank interventions in interbank markets. In their model, banks subject to privately observed liquidity shocks can use an interbank market to hedge such shocks. However, banks’ reliance on the liquidity provided by an interbank market leads them to under-invest in liquid assets. A central bank can efficiently supply liquidity if banks’ asset choices can be perfectly monitored. But under asymmetric information about banks’ asset choices, a central bank is unlikely to acquire perfect knowledge of the quality of bank assets, and access to the LOLR will need to be designed so as to prevent banks from free-riding on the liquidity facility.
Rochet and Vives (2004) examine the role of LOLR policies, prompt corrective action, and liquidity requirements in a model where liquidity crises can occur as unique equilibria. They show that prompt corrective action is a necessary component of an efficient LOLR arrangement. Liquidity requirements can indeed prevent liquidity crises, but they are too costly in terms of foregone investment opportunities: this is an important point to which we return momentarily.
Some contributions have focused on the design of LOLR policies in the context of models where banks realistically offer nominal rather than real contracts. In Diamond and Rajan's (2006) model, money is essential as it facilitates transactions and must be used to pay taxes, and can improve risk sharing, since price adjustments introduce state contingencies in bank deposit contracts that improve the efficiency of consumption allocations. When a liquidity crisis occurs, a central bank buying bonds with money can reduce the incidence of bank failures, since it allows banks to continue to fund otherwise unfunded long-term projects. In a similar set up, Allen, Carletti and Gale (2014) show that a central bank policy of accommodating banks’ demand for money leads to first best efficiency when the banking system is hit by both aggregate liquidity and asset return shocks, as the adjustment of the price level supports the optimal level of real balances and consumption allocations. Yet, an accommodative monetary policy is not always sufficient to achieve efficiency, since it is ineffective in allowing the banking system to share bank-specific asset return risk. However, the assumption of fully flexible goods prices, which in these set-ups are necessary to support the role of central bank's monetary injections, is difficult to reconcile with the reality of sticky prices in the short and medium run.
Differing from the foregoing contributions, Stein (2012) focuses on central bank policies in the presence of fire-sale externalities. In his model banks create inside money by issuing deposits, while outside money is provided by a central bank. A fire-sale externality implies that banks under-invest in short-term assets and issue excessive short-term debt, thereby creating excessive inside money. Because of the resulting excessive liquidity mismatches, banks are forced to sell their assets at fire sale prices when they are hit by a liquidity shock. In turn, fire sales negatively affect all other banks through a decline in asset values. A LOLR policy can be successfully implemented to stem the adverse effect of the liquidity crisis. 4 However, in contrast to Rochet and Vives (2004), the LOLR might be more costly than liquidity regulation because of the need to tailor ex post interventions selectively due to the difficulty of distinguishing illiquidity from insolvency.
Indeed, distinguishing illiquidity from insolvency is one of the key potential costs associated with LOLR policies, particularly when the value of collateral assets needs to be assessed (see, e.g., Freixas and Parigi, 2014). Moral hazard induced by asymmetric information about banks’ investment choices presents a design challenge for LOLR policies, as emphasised in the papers by Acharya, Shin, and Yorulmazer (2011) and Farhi and Tirole (2012).
Acharya, Shin, and Yorulmazer (2011) investigate how LOLR policies affect bank liquidity choices ex ante. In their set-up, banks may either hold liquid assets in excess of what would be optimal to exploit gains arising from acquiring assets at fire-sale prices, or hold liquid assets at a level lower than optimal if they choose to pursue riskier investments. They show that in a liquidity crisis, LOLR unconditional support to failed and surviving banks reduces banks’ incentives to hold liquid assets. In contrast, support to surviving banks conditional on their liquid asset holdings increases banks’ incentives to hold liquid assets. Thus, this model suggests the existence of an important link between ex ante bank liquidity decisions and the contingencies under which (ex post) LOLR assistance is provided.
In Farhi and Tirole (2012), the efficiency of a LOLR policy depends on whether a central bank commits ex ante to an incentive compatible set of rules determining banks’ access to liquidity facilities. If the terms of LOLR liquidity provision are perceived to be stringent (lenient), then each bank has the incentive to hold higher (lower) levels of short-term liquid assets or issue less (more) short-term debt. However, LOLR policies may be subject to a time-inconsistency problem: a stringent LOLR policy set ex ante to induce banks to hold precautionary liquid assets might be expected by banks to become more lenient in a liquidity crisis through the expansion of access to LOLR facilities.
Summing up, the literature reviewed above points to the need to design LOLR policies that integrate ex ante regulatory tools capable of mitigating moral hazard and inducing banks to internalise some of the identified externalities. The design of appropriate (ex ante) regulations, and assessment of whether there is a specific role for liquidity regulation, are the key issues I now turn to.
Regulation
Asymmetric information about banks’ risk choices, uncertainty about bank asset values, and increases in uncertainty about bank default risk as triggers of liquidity crises, present the challenge of designing policies that control bank default risk ex ante. A classical regulatory instrument in controlling bank default risk is capital regulation. While the literature providing a rationale for capital regulation is large, there is not yet a consensus about what an optimal design of capital requirements is, and what a desirable level of capital requirement might be (see e.g. De Nicolò, 2015).
Are liquidity requirements a necessary complement to capital regulation? Admati et al. (2013) suggest that capital requirements might be a substitute for liquidity requirements, since they might target the same financial stability objectives at a lower cost. First, adequate bank capital would reduce the risk of information-based runs leading to liquidity crises. Second, capital requirements would lower the cost of providing LOLR assistance by reducing the cost of distinguishing illiquidity from insolvency. Most importantly, Admati et al. (2013) argue that the social costs of capital regulation are likely to be significantly lower than those associated with liquidity requirements. Forcing banks to hold excessive levels of liquid assets imposes on them the payment of an unnecessary liquidity premium, which is a social cost as it prevents banks from investing in more productive assets. In contrast, equity capital can be invested into lending or in marketable securities that earn returns higher than liquid assets. This argument essentially mirrors the conclusion of the model by Rochet and Vives (2004) reviewed above.
In her review of the sparse literature on liquidity measurement and regulation, Bouwman (2014) maps liquidity and capital requirements into the different sources of bank risks these regulations are supposed to address. Liquidity requirements may mitigate withdrawal or rollover risk on banks’ liability side, while capital requirements may mitigate bank default risk. Yet Bouwman recognises that in practice liquidity and capital requirements may affect banks’ joint investment and funding decisions, so that regulatory design should take into account the interaction of technological, market and funding liquidity risks mentioned in the models previously discussed.
Calomiris, Heider and Hoerova (2014) make an attempt to identify a specific role of liquidity requirements as complementary to LOLR policies and capital regulation. They study a finite horizon model of a banking system composed of heterogeneous banks which choose their risk profile (unobserved to outsiders) and offer demand deposit contracts. Liquid reserves act as buffers that reduce the vulnerability of banks arising from granting depositors the option to withdraw their funds. Runs can occur when triggered by uninformed depositors, or when prompted by depositors’ perceived deterioration in banks’ solvency. In the latter case, liquidity requirements are shown to provide incentives for banks to choose a conservative risk profile, since liquid asset holdings are observable and can signal asset quality. This would lower the probability of depositors withdrawing funds based on limited information about banks’ default risk. Yet, whether the functions attributed to liquidity requirements can be equivalently performed at a lower cost by suitably designed capital requirements remains an open issue.
Acharya, Mehran, and Thakor (2015) present a model where the design of capital regulation embeds provisions resembling a fairly special ‘liquidity’ requirement. In their model, banks face two moral hazard problems related to unobservable risk-taking choices by managers and risk-shifting by shareholders. They show that capital requirements that include a ‘special capital account’ composed of liquid assets that become unavailable to creditors in case of bank failure can effectively deal with both types of moral hazard. This arrangement however resembles more a form of contingent capital requirement than a liquidity requirement.
As noted, Rochet and Vives (2004) and Admati et al. (2013) concluded that liquidity requirements are likely to be more costly than appropriate LOLR policies, while Stein (2012) suggests the reverse is likely to be the case. Recent papers by Goodhart et al. (2012; 2013) and De Nicolò, Gamba and Lucchetta (2014), provide models in which multiple regulations come into play, and address the relative costs of different regulations.
Goodhart et al. (2012; 2013) consider a three-period general equilibrium model in which multiple regulations, including liquidity requirements, are evaluated in terms of their effects on welfare under a variety of financial frictions. Using simulations, they find that liquidity requirements are more effective than capital requirements in mitigating the build-up of liquidity mismatches, but need to be lowered in downturns to prevent fire sales. However, these results are highly dependent on parameter constellations that are only loosely calibrated using actual data.
As observed by Tirole (2011), the finite horizon (generally three-period) models used in most of the literature may capture key trade-offs in a tractable way, yet infinite horizon models may deliver a more reliable assessment of the impact of bank regulations, since they would fully take into account the implied shadow costs faced by banks. The paper by De Nicolò, Gamba and Lucchetta (2014) (DNGL hereafter) provides a welfare evaluation and a quantitative comparison of capital, liquidity requirements, and prompt corrective action policies in an infinite horizon model of a banking industry.
In DNGL, banks financed by insured deposits and short-term collateralised debt dynamically transform short-term liabilities into longer-term partially illiquid assets whose returns are uncertain. Financial frictions are introduced in the form of costly equity issuance and constraints on collateralised debt. The impact of regulations is quantitatively assessed in terms of changes in bank lending and measures of bank efficiency and welfare relative to a benchmark (unregulated) case, with simulations performed with parameters calibrated on US data. The dynamics of the model and bank optimal choices are analysed along the business cycle – parameterised by the evolution of a systematic risk factor – and in steady state. DNGL find that in an upturn, the addition of liquidity requirements to capital requirements forces banks to use retained earnings to build up liquidity buffers rather than invest in lending, but in a downturn liquidity buffers are not significantly different from those held by banks voluntarily when only capital requirements are in place. This result indicates that capital requirements provide banks with incentives to create liquidity buffers in downturns, i.e. liquidity holdings are counter-cyclical. In steady state, DNGL find that adding liquidity requirements to capital requirements reduces bank lending, efficiency, and welfare. This occurs because banks are forced to use retained earnings to increase bond holdings or reduce indebtedness, rather than investing them in lending. Quantitatively, the declines in bank lending and value metrics of efficiency and welfare associated with liquidity requirements are found to be fairly large and robust to several configurations of parameters.
Furthermore, DNGL obtain an important result: a policy of prompt corrective action imposing capital requirements contingent on observed bank capitalisation dominates standard non-contingent capital and liquidity requirements in terms of bank efficiency and welfare. A key reason for this result is that in the DNGL model deposit insurance introduces incompleteness in bank deposit contracts, as deposit payments are not contingent on the realisation of states of financial distress. Non-contingent capital and liquidity requirements do not change the degree of completeness of deposit contract. By contrast, prompt corrective action introduces contingencies that act as substitutes for the missing contingencies in deposit contracts. Although DNGL focus on prompt corrective action implemented through contingent capital requirements, their result suggests that prompt corrective action implemented via restrictions on liquid holdings might be a regulatory design worth exploring.
4. Basel III liquidity regulation and beyond
The literature reviewed thus far suggests that various types of externalities justify policies aimed at reducing excessive asset and liquidity risk in the financial system. However, a well grounded theoretical rationale for liquidity requirements as necessary complements to capital regulation, prompt corrective action, and LOLR-type policies has yet to be established.
In practice, the goal of current Basel III liquidity requirements is to reduce liquidity mismatches in the banking sector so as to limit banks’ need to liquidate assets should a liquidity crisis occur. These requirements include a Liquidity Coverage Ratio (LCR) and a Net Stable Funding Ratio (NFSR). The LCR requires banks to hold fractional reserves of liquid asset to meet short-term (30 days) liquidity needs (see Basel Committee on Banking Supervision, 2013). The NSFR is the ratio of the available amount of stable funding to the required amount of stable funding over a one-year horizon. Stable funding includes customer deposits, long-term wholesale funding, and equity. The required amount of stable funding is calculated by weighting funding by maturity (see Basel Committee on Banking Supervision, 2014). This ratio effectively imposes an upper limit on short-term debt with the objective of reducing funding liquidity risk.
These regulations are likely to affect bank business models, debt markets and monetary policy operations significantly. While the LCR is likely to have a direct impact on banks’ liquidity demand and their recourse to short-term central bank financing (see e.g. Bech and Keister, 2013), the NSFR involves changes in banks’ structural funding composition, indirectly affecting money markets and participation in monetary policy operations. Specifically, the NSFR is likely to induce a reduction in money market volumes and increase the attractiveness of longer-term central bank refinancing operations. An assessment of the impact of the NSFR regulation is planned to be performed during an observation period. The results of this assessment might induce central banks to adjust their existing policy frameworks in order to preserve their effectiveness (Committee on the Global Financial System, 2015).
The current consensus in the regulatory community is that these liquidity requirements have two key benefits: they give policymakers sufficient time to assess bank liquidity positions and arrange appropriate responses in times of stress, and force banks to maintain precautionary liquidity cushions limiting liquidity mismatches and funding liquidity risk. However, an assessment of the cost of these requirements on bank lending and real activity is still uncertain, as current estimates vary significantly depending on the assumptions and models adopted.
Based on a variety of models developed at central banks, MAG (2010) and Angelini et al. (2011) estimate a relatively small negative impact of liquidity requirements on lending and real activity. However, IIF (2011) and EBA Banking Stakeholder Group (2012) estimate significantly larger negative effects. As noted, in the context of a calibrated dynamic partial equilibrium model, DNGL estimates imply that the negative impact on lending, bank efficiency and welfare is fairly large. By contrast, in a general equilibrium model calibrated on US data, where banks do not perform any maturity transformation, Covas and Driscoll (2014) find that the introduction of liquidity requirements results in permanent reductions in bank lending and output smaller than those obtained by DNGL, but their model does not evaluate regulations either in terms of efficiency or in terms of welfare.
Two recent empirical studies focus on the combined impact of capital and liquidity requirements and on the link between bank solvency and liquidity. Distinguin, Roulet and Tarazi (2013) investigate the relationship between bank regulatory capital and bank liquidity using on-balance sheet data for European and US publicly traded commercial banks, and find that solvency and liquidity measures can be negatively or positively correlated depending on the type of banks and on how liquidity is measured. Using data for a sample of US Bank Holding Companies, Pierret (2015) finds a strong correlation between solvency and liquidity, suggesting that bank capitalisation is instrumental in ensuring access to liquidity in times of stress.
In conclusion, the theoretical and quantitative assessments of the benefits and costs of the current liquidity requirements, the links between solvency and liquidity, and the integration of liquidity regulation into bank regulation more generally, are tasks still in their infancy. Further developments in research and policy design are clearly needed.
The analysis of DNGL suggests that relative to capital regulation, liquidity regulation is fairly costly, consistent with the conclusions of Rochet and Vives (2004) and the arguments of Admati et al. (2013). However, the DNGL model, as well as any other model I am aware of to date, neither includes any of the externalities emphasised by the literature nor a LOLR-type facility. The introduction of these features in quantitative dynamic models of banking might significantly affect the evaluation of the net benefits of liquidity regulation.
Indeed, the importance of further developing quantitative models integrating policy analysis of multiple regulations hardly needs to be emphasised. As the literature on liquidity regulation is still at an early stage of development, progress in this area is clearly urgently needed to inform policy. As stressed by Brunnermeier and Sannikov (2014), the prevention and resolution of liquidity crises does not necessarily imply that strict financial regulations are best. In their model, policies requiring large transfers to and from the financial system or large open-market operations are best, and resemble some of the policy responses actually implemented during the 2007–9 financial crisis. However, they also show that small policy mistakes, even related to standard regulations, such as capital requirements, may have unintended consequences.
Dermine (2013) stresses the potential inefficiencies and attendant high welfare costs of liquidity requirements due to their hampering the key role of banks in providing welfare-enhancing maturity transformation, consistent with DNGL's quantitative findings. Assessing the robustness of these results in the context of quantitative models with multiple regulations and allowing for the presence of a LOLR appears to be an important research priority. More generally, adopting a more explicit mechanism design approach may be highly desirable. The DNGL result concerning the dominance of regulatory requirements as contingent ex ante prompt corrective actions over uncontingent requirements suggests that different designs of regulatory mechanisms could attain the desired financial stability objectives at lower costs.
Lastly, the limited knowledge of the benefits and costs of the currently envisioned liquidity requirements support the gradualism with which Basel III liquidity regulations have been planned to be introduced. Equally important is the on-going review of the results of the slow phasing in, which may admittedly require some modifications to the original Basel III liquidity requirements. A particularly important task that would usefully complement these efforts is the improvement of measurement of liquidity at a bank level, in the aggregate and across financial markets, as advocated by Tirole (2011) and along the lines suggested by Brunnermeier, Gorton and Krishnamurthy (2014).
Footnotes
2
An example of how these externalities may play a joint role in triggering liquidity crises is the model by Acharya, Gale and Yorulmazer (2011), who show that small shocks to assets’ fundamental values can significantly curtail banks’ debt capacity – defined as the capacity to borrow using assets as collateral – leading to fire sales and market freezes.
3
Holmstrom and Tirole (1998) provide general theoretical support for the public supply of liquidity by central banks, owing to their capacity to create money-like traded instruments. Bolton, Santos and Scheinkman (2011) provide a rationale for such public liquidity support in the context of market freezes and liquidity crises arising from asymmetric information.
4
Relatedly, Brunnermeier and Sannikov (2015) study a fully dynamic model with a central bank that pays interest on bank reserves. Banks create inside money as in 
, and their investments determine the money multiplier. They show that an accommodative monetary policy in downturns can mitigate the destabilising effect of aggregate liquidity shocks.