RDP 2001-07: A History of Last-Resort Lending and Other Support for Troubled Financial Institutions in Australia 2. Theoretical Approaches to Lender-of-last-resort Policy

The case for a lender of last resort rests on the need to compensate for two types of market failure: the existence of information asymmetries in banking markets and the systemic consequences of bank failure. A lender-of-last-resort facility, however, involves some costs, particularly moral hazard costs. The question, then, is how lender-of-last-resort policy can best be shaped so as to minimise these costs.[3]

2.1 The Case for a Lender of Last Resort

The structure of banks' balance sheets distinguishes them from other types of firms. On the liability side, bank deposits have a fixed nominal value and are paid out on a first-come-first-served basis. Most of their assets comprise loans that are not readily marketable. This combination of fixed-value deposits and illiquid assets makes banks susceptible to depositor runs and results in externality costs to the economy if a bank fails.

That deposits of fixed nominal amounts are paid out on a first-come-first-served basis creates the potential for coordination problems amongst depositors. Each depositor knows that if other depositors rapidly withdraw funds, banks will be forced to sell illiquid assets at a loss and, as a result, are unlikely to repay all deposits. Diamond and Dybvig (1983) argue, therefore, that any external event that causes depositors to believe others will withdraw their deposits may trigger a run.

A core element of banks' business is the provision of loans to those borrowers who are unable to obtain finance directly from debt markets. Banks specialise in lending to firms that lack sufficient capital to pledge as collateral or the widely known reputations that would allow them to raise funds directly in capital markets. By accessing inside information on firms, banks overcome the information costs that make debt issuance direct to the market prohibitively expensive for many companies (Fama 1985). Due to the inherently opaque nature of such assets, depositors are not readily able to observe the financial condition of banks. This creates an information asymmetry between depositors and bank management. Since depositors are not able to monitor, and therefore price, the riskiness of individual banks, they cannot effectively penalise risk-taking by demanding higher interest rates. Instead, they can only discipline managers by withdrawing their funds (Calomiris and Kahn 1991).

Systemic risk arises from the potential for problems at one bank to cause difficulties at other banks. Sources of contagion include asymmetric information problems and interbank exposures. Since depositors cannot monitor bank risk-taking, difficulties at one institution may cause depositors to doubt the liquidity of other banks, leading to a generalised flight to cash (Docking, Hirschey and Jones 1997). In addition, interbank exposures arise from the operation of the payments system (Rochet and Tirole 1996).

Banking illiquidity may impose externality costs on the broader economy. Due to the information-intensive nature of banking relationships, difficulties at an individual bank may disrupt the supply of finance to firms that are unable to obtain financing through other means. For those firms, re-establishing banking relationships is likely to be a lengthy and costly process (Greenbaum and Thakor 1995). Moreover, since the health of financial institutions and the economy are closely linked, banks are more likely to experience difficulties in periods of broad economic slowdown and may, as a result, amplify cycles in the real economy (Suarez and Sussman 1997).

2.2 The Costs of Lender-of-last-resort Policy

A lender-of-last-resort facility provides a form of insurance for both banks and their depositors. By reducing the costs of risk-taking, a lender of last resort can reduce the incentive for banks to manage their own liquidity and for depositors to monitor banks' activities. This opens the way for banks and depositors to adopt moral-hazard driven behaviour. This can result in a more fragile banking system, as banks take on more risk and depositors cease to demand risk premiums from more risky banks, as they would if banks were fully exposed to the consequences of their risk-taking. Thus, while a lender of last resort may reduce the frequency of bank runs, it may create an environment where the runs that do occur are more severe.

While authors such as Schwartz (1986) acknowledge the market will not always deliver the first-best outcome, they argue that the resulting costs are less than those arising from inappropriate public support of financial institutions. The loss of information-intensive business relationships brought about by banking illiquidity may be no more costly than the failure of other types of firms (Kaufman 1994).

Proponents of free banking, such as Selgin and White (1987) and Dowd (1992) make the case against a lender of last resort more strongly, and deny the need for any government-sponsored lender of last resort. They argue that the fundamental cause of bank panics is the legal restrictions on the banking system, particularly the government's monopoly in the issue of currency. If banks were free to issue their own notes, secondary markets in bank notes would provide adequate information to note holders about the condition of each bank. In a system free of restrictions, the market would produce a stable financial system through the development of institutions, such as clearing house associations, mitigating banks' information asymmetries.

2.3 Issues in the Design of Lender-of-last-resort Policy

One of the first to set out the ways in which lender-of-last-resort policy should be conducted to minimise the moral hazard costs of last-resort support was Bagehot (1962). Implicit within Bagehot's discussion of lender-of-last-resort policy was the notion that liquid funds should be provided to the financial market rather than directly to individual financial institutions. Schwartz (1986) and Goodfriend and King (1988) concur, arguing that the central bank does not have superior, or more timely, information than the market as a whole. Therefore, liquidity support should be provided to the market, with the market then allocating the liquidity. Seen in this light, the provision of liquidity to individual institutions involves discretionary government assistance that is likely to distort the competitive landscape. While private market solutions may be able to handle idiosyncratic failures, only the central bank can address an economy-wide increase in demand for liquidity occasioned by a macroeconomic shock (Bordo 1990). However, it can also be argued that responding to difficulties at an individual institution by providing liquidity to the market is not true last-resort lending as it can only succeed if market participants are prepared to continue to extend finance to the troubled institution.[4]

Bagehot advocated that, in operating a lender-of-last-resort facility, a central bank should:

  • offer liquidity support ‘freely and vigorously’, but at a penalty rate;
  • accommodate anyone with good collateral; and
  • make clear in advance the central bank's readiness to lend freely.

In recommending that the lender of last resort offer support freely, Bagehot advocated that it should lend to any sound borrower – non-banks as well as banks. Subsequent theoretical analysis, such as that outlined in Section 2.1 above, argues that provision of last-resort support should be confined to those financial institutions that are of systemic importance and that a special case can be mounted for the provision of support to banks.

Charging a penalty rate strengthens the incentives facing financial institutions, encouraging them to make their own arrangements to insure against liquidity shortfalls. Penalties must be well targeted. Setting penalties too low may result in overuse of last-resort lending. Charging a penalty rate that is too high may impair a borrowing institution, which is likely to already be in a fragile position. Moreover, penalty-rate financing that is too costly may create an incentive for institutions to ‘gamble for resurrection’ by adopting high-risk strategies in an attempt to trade out of any difficulties (Levonian 1991).

Bagehot's second caveat was that the lending should be based on good collateral. This means that last-resort support should be available to illiquid, but not insolvent, banks. Such an approach aims to separate the risk of system-wide fire sales of assets prompted by liquidity shocks from idiosyncratic bank failures due to inefficiency or mismanagement. In this way, Bagehot was seeking to distinguish between instances where fundamentally sound banks were brought down by market failure and closures of poorly managed banks, where the full force of market discipline should be applied.

In the midst of a crisis, distinguishing between illiquid institutions that are solvent and institutions that are insolvent may be extremely difficult. In many cases, liquidity support is sought at the very time that the market itself is having difficulty assessing the solvency of an institution (Goodhart 1985). Bagehot recognised it is difficult, particularly in times of financial distress, to accurately value financial assets. In a period of asset deflation and disruption of price relativities which often surrounds banking difficulties, it is difficult to establish whether the fall in the price of a particular financial instrument is due to short-term illiquidity or inherent weakness. Bagehot argued, therefore, that collateral should be valued at pre-panic prices. This, however, presupposes that pre-panic valuations are sound and not distorted by asset-price bubbles (see Kent and Lowe (1997)), which are often seen in the run-up to banking crises.

If systemic consequences are taken into account, it is possible that public action may be required to ensure the smooth exit of a clearly insolvent institution. It also may be the case that providing support to an insolvent institution may be less costly than allowing it to fail (Guttentag and Herring 1983).

The third element in Bagehot's recommendations was that central banks should make clear in advance their readiness to lend unsparingly. During a bank run, anything that increases depositors' fears that their deposits will not be repaid is likely to exacerbate a run. Bagehot argued that the lender of last resort should allay depositors' apprehension by publicly declaring that liquidity would be available. However, maintaining some uncertainty about the conditions under which lender-of-last-resort policy may be activated (‘constructive ambiguity’) may induce financial institutions to be more cautious than they would be if they were certain of being bailed out (Enoch, Khamis and Stella 1997).

Constructive ambiguity, however, gives central banks the scope to overuse their lender-of-last-resort powers by appealing to the systemic consequences of bank illiquidity. Central banks face an asymmetric pay-off when conducting lender-of-last-resort policy. On the one hand, there is always some chance that failure to provide liquidity support may lead to cascading financial distress and macroeconomic costs – an outcome which would attract heavy criticism. On the other hand, much less criticism might be expected if the central bank provided (inappropriate) liquidity support to an institution since such support could always be justified on the basis that the central bank's assessment of the (unobservable) probability of financial instability was high. It is argued, therefore, that central banks will systematically overestimate the threat to the macroeconomy when an illiquid institution cannot meet its payments.

Another difficulty with the systemic-risk approach is that it risks fostering the presumption that certain institutions are ‘too big to fail’. This creates a competitive distortion favouring larger banks over smaller banks.

Kindleberger (1978) proposed that the provision of lender-of-last-resort support should be uncertain and, in addition, responsibility for the provision of last-resort support should be divided among a small number of institutions. The division of responsibility is needed to ensure credible ambiguity. Spreading responsibility disperses public pressure for assistance, reducing the likelihood that individual public officials will fall sway to the demand of lobbyists seeking to be bailed out.

An alternative approach to overcome the asymmetric pay-off faced by central banks is the application of a rules-based approach. Mishkin (1998) argues that, in adopting a systemic-risk focus, monetary authorities should adopt a rule that the first large institution should be allowed to fail in any period of financial stress. Authorities should then stand prepared to extend the safety net to the rest of the financial system if they perceive there is a serious systemic risk. While this approach goes some way towards preventing overuse of the lender-of-last-resort facility, it may not fully address the trade-off between short-term financial disturbances and longer-run moral hazard costs. For example, it may lead a troubled institution to do whatever it can to forestall disclosure of its own difficulties; ultimately, this could lead to deeper and more widespread problems.

Rather than seeking to reduce public sector intervention, one approach to minimising moral hazard costs is to recognise that last-resort support yields a second-best result that warrants additional regulation (Goodhart 1985). Prudential regulation, such as requirements on the proportion of assets that banks must hold in liquid form and capital adequacy rules, specifically target banks' proclivity for risk-taking.

Although there were times during the nineteenth century when neither governments nor banking industry groups were prepared to act as a lender of last resort, the case for a lender of last resort has been widely accepted for most of Australia's history. This history is sketched out in the remainder of the paper. For reference, a brief chronological summary is presented in Appendix A.


There are a number of comprehensive surveys of the theoretical literature concerning lender-of-last-resort policy, including Freixas et al (1999); Moore (1999) and Bordo (1990). [3]

Another way of ensuring greater market involvement in lender-of-last-resort arrangements is through public sponsorship of concerted private sector lending. This has the advantage of ‘bailing in’ the private sector, increasing the incentives for banks to actively monitor their bank counterparts. However, the coordination problems that affect depositors can also affect the interbank market. The central bank, by virtue of its role at the centre of the payments system, has an advantage in organising private sector liquidity support. Coordination problems may be resolved by the central bank bringing banks together – once the banks are able to reassure one another they will not run, interbank lending may resume (Freixas et al 1999). [4]