A run on a bank can be met in two ways. The holders of the bank's short-term debt can be told that they will be paid whatever happens, so that none of them has a reason to be first in the queue; or the bank can be lent enough cash to pay all who ask. The first is a guarantee and the second a loan. In practice the two are mixed, and a public text that announces either has to say three things if a reader is to hold the authority to it: what is promised, how much of it, and until when.
This article reads the public texts of a few days in March 2023 in which both instruments were used, on two sides of the Atlantic, and asks what they say on those three points. The research question is narrow: how did the documents of March 2023 set the guarantee and the loan, and how did the sums they announced differ from the sums that could be used? The answer is sought in the wording of four primary texts: the press release of the Federal Deposit Insurance Corporation of 10 March 2023 on the closing of Silicon Valley Bank, the joint statement of the Department of the Treasury, the Federal Reserve and the FDIC of 12 March, the Swiss Federal Council's emergency ordinance of 16 March on additional liquidity assistance loans and federal default guarantees, and its amendment of 19 March. Beside them stand two modern records of the episodes kept by Skeptron, a studio of the RISCON family, and the counts of a ledger built on those records by a study of emergency acts in money and credit, called here the parent study.
The contribution is to set deposit protection and central bank lending side by side on the record. The works read here tend to put their weight on one instrument or the other. March 2023 offers a pair of texts written within days of each other in which one government chose, in public, to protect depositors, and the other chose, in law, to lend. Read together, they show a difference of form that the labels hide. The American texts commit to an outcome for a class of people (depositors will have their money) without a sum; the Swiss texts commit to a procedure for a class of institutions (loans of three kinds, opened in order) without a sum either, and fix the only sum they name, the loss guarantee, as a ceiling with a first loss in front of it.
Two debates are joined. The first is about what stops a run: a guarantee to the holders of short-term claims, as in the work of Douglas Diamond and Philip Dybvig, Gary Gorton and Andrew Metrick on deposit insurance and runs (Diamond and Dybvig 1983; Gorton 2012; Metrick 2024), or a lender of last resort lending on good collateral to solvent institutions, as in Walter Bagehot's rule and Thomas Humphrey's account of the classical doctrine (Bagehot 1873; Humphrey 1989). The second is about rules: whether rescues need rules set in advance and agreed with the government (Tucker 2018), or whether in an emergency the executive acts first and others defer (Posner and Vermeule 2010), with Oren Gross and Fionnuala Ní Aoláin's models of law in a crisis as a map between them (Gross and Ní Aoláin 2006). The American and Swiss texts fall at different points among these positions, and the article shows where, without deciding which is better.
The placement of the secondary works in these debates, and their application to the March 2023 texts, are this article's own. What is said of each work stays within the published reading notes of the parent study, and no secondary work is quoted. The primary texts are quoted verbatim and at length; the Swiss ones are in German, and the English beside them is this study's, made for the parent study and this article, not an official translation.
Section 2 lays out the two debates side by side. Section 3 describes the texts, the records and the ledger. Section 4 reads the American texts of 10 and 12 March, and Section 5 the Swiss texts of 16 and 19 March. Section 6 sets the two together and asks what the evidence can and cannot carry, and Section 7 concludes.