Chapter 2. Literature: A Guarantee or a Loan, and Who Writes the Rule

2.1What stops a run

The theoretical anchor of the first debate is Diamond and Dybvig's model of a bank that offers deposits withdrawable on demand while holding illiquid assets. The bank provides something valuable, a kind of insurance of liquidity to its depositors, but it can settle in either of two states: one in which only those who need cash withdraw, and one in which all withdraw because each expects the others to. A run can thus happen without any fundamental weakness in the bank. In their argument, either deposit insurance or a lender of last resort can remove the bad state, though both raise problems of incentives (Diamond and Dybvig 1983). Either instrument, in the model, can do that work.

Gorton takes the run out of the bank and into short-term debt of every kind. In his account, crises are runs on such debt, from bank deposits in the nineteenth century to repurchase agreements in 2007 and 2008, which happen when holders lose confidence that the debt can be redeemed at face value; they recur wherever such debt exists, and regulation and government guarantees have been responses to them (Gorton 2012).

Metrick's article on 2023 applies the question to a single bank. He describes a bank with a concentrated base of uninsured depositors and large unrealised losses on securities that faced a very fast run, and an official response in two parts: a systemic risk exception to protect all depositors and a new lending facility for banks. He discusses regulatory and supervisory failures and the questions the episode raises for deposit insurance and for run risk (Metrick 2024). His account thus contains both instruments at once, the guarantee to depositors and the loan to banks.

On the other side stands the older doctrine of the lender of last resort. Bagehot's Lombard Street of 1873 is its classic statement; its rule, as usually summarised, is to lend freely in a panic, at a high rate of interest, on good collateral, and to announce beforehand that this will be done (Bagehot 1873). Humphrey's survey of the doctrine's history presents it as the central bank's duty to protect the money stock and the banking system by lending to solvent institutions on good collateral at a penalty rate, rather than to rescue insolvent individual firms (Humphrey 1989). Charles Kindleberger and Robert Aliber, in their survey of crises across centuries, hold that a lender of last resort is needed to stop a panic from spreading, and that the timing and wording of its announcement matter (Kindleberger and Aliber 2005).

In Diamond and Dybvig's model the two positions are alternatives, not rivals. The difference that matters here lies elsewhere. A guarantee is a promise to a creditor about an outcome, and can be announced without a sum, because its size is set by the claims it covers. A loan is a transaction with a debtor on terms, and in the classical doctrine its size is set by what the borrower can pledge.

2.2Who writes the rule, and when

The second debate is about legitimacy and timing. Tucker, a former central banker, argues that independent agencies such as central banks need principles of legitimacy if they are to hold great power in a democracy: a clear mandate, limits on what they may do, transparency and accountability to elected bodies. He applies this to emergency lending and holds that rescues of institutions are political choices that need rules set in advance and a framework agreed with the government (Tucker 2018).

Posner and Vermeule argue the contrary about where the real checks lie. In modern American government, in their account, the executive is in practice little constrained by legislation and the courts; political and informal checks, such as public opinion, parties and the media, do the main work; in emergencies the executive acts first and the other branches defer, and this is acceptable (Posner and Vermeule 2010). Their book is controversial, and its subject is American.

Gross and Ní Aoláin offer a map rather than a side. Their study of law in times of crisis sets out models of the legal response to emergency: that ordinary law applies fully, that emergencies lie outside the law, and that emergency powers are accommodated within it (Gross and Ní Aoláin 2006). Their cases are chiefly terrorism and threats to national security, but the models bear on decisions that suspend ordinary legal rules in a crisis.

Two further works bear on the question of time. Finn Kydland and Edward Prescott show that a plan which is best today may not remain best when the future arrives, because once private agents have acted on it the policy maker has an incentive to depart from it; they argue for rules in place of discretion (Kydland and Prescott 1977). Robert Higgs argues that American government grew in jumps during national crises and did not return fully to its earlier size afterwards, a pattern often called the ratchet effect (Higgs 1987). The first bears on whether an emergency promise binds later; the second on whether an emergency measure ends.

2.3The debates side by side

The authors do not all address one another. What joins them here is that each makes a claim that a text written on the day can be read against: not whether a run was stopped or a rescue was legitimate, but what was promised, how much, by whom and until when.