Introduction

0.1What this study asks

On Monday 6 March 1933 every bank in the United States was closed by proclamation. The proclamation said for how long: "from Monday, the Sixth day of March, to Thursday, the Ninth day of March, Nineteen Hundred and Thirty-three, both dates inclusive". The archive's summary of that week has three dates in it, the closing order, the Act passed three days later and the first reopenings four days after that, and it notes that not every bank came back. Few emergency acts in money and credit are written down so completely. Most records of such acts say what was done. Fewer say for how long, on what condition the act would end, or who carried its cost while it lasted.

Every study in the series puts the same question to its archive: what must be decided now, and what would make the decision change? Here the archive is one of money and credit in emergencies. Its own question, under that one, is narrower. Of an authority's emergency act on money, credit or a market, it asks three things of the record: whether it calls the act temporary, whether it names when or on what condition the act would stop, and whether it names the people who paid for it in the meantime.

The question is about records, not about acts. It does not ask whether a suspension was wise, whether a rescue worked, or whether a debtor should have been relieved. It asks what a later archive has kept that would let a reader judge an emergency act by the promise it made about its own end.

0.2The series and its question

The series is called The Time of Questions and Answers. Its studies read archives of decisions kept by the studios of the RISCON family, and every one of them asks the charter's question in its own domain: what must be decided now, and what would make that decision change? Every study codes its cases on the same axes, so that a final volume can set the archives beside one another: who decides and with what authority; on what grounds and with what missing; by when, and who waits and pays meanwhile; whether the decision can be undone; what would make it change, and whether that could be observed; and when it was judged.

The first study of the series read Skeptron's records of epidemics and public health. That study was about knowledge at the moment of acting: what a public body knew when it closed a gate or counted the dead. This second study reads the same archive's records of money and credit. The two archives are made in the same way, by the same studio, in the same form, and the two studies were made with the same tools in the same order. That makes the second study, among other things, a comparison: where the answers differ, the difference lies in the domain or in the records, not in the method.

In money and credit the series' question has an especially plain form. Many of the acts read here describe themselves as holding operations, meant to last until something else happens. What would change such an act is, first of all, what would end it. A record that names the end lets a later reader ask whether the act kept its promise. A record that does not leaves the reader with an act and no measure for it.

0.3Sub-questions

The question breaks into five, each registered as a hypothesis before any case was coded for it (Chapter 2 gives them in full, Chapter 12 judges them).

1. Is the act's end named in terms a later reader could check? (F1) 2. Where an act is called temporary and its fate is told, did it stay longer than promised? (F2) 3. Are the day an act was decided and the day it took effect, or was settled, given as two dates? (F3) 4. Is a headline sum told apart from the sum that could actually be used? (F4) 5. Are the people who lost by the act, or by its delay, named? (F5)

A sixth is a test: in how many records do all four of the things a later judge would need stand together?

0.4The argument in brief

The ledger's answer is clear on three of the five hypotheses and unclear on one.

Most records name the act and not its end. Among the cases with an act, 30 support the first hypothesis and 6 go against it on the lenient reading, a share of 83%, with 8 left unclear and 40 to which it does not apply; on the strict reading the share is 88%. The pattern holds in every group of the archive that has acts: in the crisis history and in the credit history alike.

Most records do not name who paid. The fifth hypothesis is supported by 35 cases against 3 on the lenient reading (92%), and by 31 against 1 on the strict (97%). Depositors, note holders, debtors, taxpayers and shareholders appear in the actors' own texts more often than in the archive's summaries: the joint statement of 12 March 2023, for instance, names shareholders and certain unsecured debtholders as unprotected and a special assessment on banks as the way losses to the deposit insurance fund would be recovered. The records seldom carry such sentences over.

Two predictions about the shape of the records failed. This study expected the archive's crisis and credit histories, whose titles are built on gaps (between an order and a reopening, between a charter and an opening, between a notice and an effective date), to keep two dated moments apart in most cases. They do not: the third hypothesis goes against, 13 to 28 on the lenient reading (32%). The dates are often in the title and the studio's recurring sentences, not in the case's own text. The study also expected the cases with a figure to keep the announced sum apart from what could be drawn. They seldom do: the fourth hypothesis goes against, 7 to 31 (18%); only in the credit history is it even unclear.

The second hypothesis, that temporary acts outlast their term, cannot be judged on these records. It applies to only a handful of cases, because records that call an act temporary and also say what became of it are rare; that rarity is itself part of the answer.

The test passes in one case of the 42 with an act: the bank holiday of 1933. The actors' own texts, where they survive, often hold what the records leave out, and the study reads them for that: the Privy Council's minute of 26 February 1797, the Bullion Report's terms for resuming cash payments, the Gold Standard (Amendment) Act of 1931, President Nixon's address of 15 August 1971, the Federal Reserve Board's release of 14 September 2008, the joint statement of March 2023 and the Swiss emergency ordinance of 16 March 2023 with its amendment of 19 March.

0.5The temporary act

The study's central object deserves a name: the temporary act. It is an act that suspends an obligation, holds a limit or extends a facility, and that says, or implies, that it will end. Its record owes a later reader three things. The first is its term: a date, a period, or a condition. The second is its bearer: whoever carries its cost while it stands. The third is its outcome: whether it ended as said, ended early, or outlasted its term.

The actors' texts give the term in several forms, and the forms differ in what a later reader can check. The plainest is a period of time: the Swiss Federal Council gave its emergency ordinance a life of six months, and the Board's exception of 14 September 2008 "expires on January 30, 2009, unless extended by the Board". An event is harder to observe but still names something: the Bullion Report of 1810 noted that, as the law then stood, "the Bank would be compelled to pay in cash at the end of six months after the ratification of a definitive Treaty of Peace". Hardest to check is a term left to another body: the minute of 1797 left the end to Parliament, and the Act of 1931 suspended the gold standard "Unless and until His Majesty by Proclamation otherwise directs". And some temporary acts name no term at all; President Nixon's address directed the Secretary of the Treasury "to suspend temporarily the convertibility of the dollar into gold or other reserve assets", but did not say until when.

These forms are what the fourth literature reviewed in Chapter 1 argues about: a rule with an escape clause, a ratchet that does not go back, an emergency power that defines its own limit. The records let the study see which form an act's own words took, and how often a later archive keeps any form at all.

0.6Terms

A few terms recur and are explained where they first appear in the chapters; they are gathered here for the reader's convenience.

• A suspension of payment or of convertibility stops a holder's right to exchange a note or a deposit for gold or cash, without cancelling the note or the deposit.
• A bank holiday closes banks by order for a stated time.
• A lender of last resort is a central bank or other authority that lends to banks or markets when no one else will, usually against collateral; Bagehot's rule, discussed in Chapter 4, says how such lending should be done in a panic.
• A peg fixes the value of a currency against gold or another currency; a regime is the set of rules under which a currency is managed.
• A stand-by arrangement is a line of credit from the International Monetary Fund, approved in a total and released in tranches tied to reviews and conditions.
• A receiver takes over a failed bank to protect its insured depositors and dispose of its assets.
• Emergency liquidity assistance is lending by a national central bank to its own banks against collateral, under a ceiling, outside the ordinary operations of the central bank.
• A moratorium defers the repayment of principal or interest for a stated time.

0.7Records rather than events

Skeptron's cases are not the events. Each is a reading, made in 2026, of one or more public sources: a central bank's history page, an official history of the IMF, a press release, a translated law collection, a chronicle compiled centuries after the events it records. Each case gives a scenario of a few sentences and five questions with answers: when the source dates the event, a concrete fact, the boundary of what can be judged, a calendar check, and the condition on which a later comparison would hold or be withdrawn.

Two things follow. A good deal of every summary is the studio's own method, repeated in case after case with the names changed. This study calls those recurring sentences the frame, explains them once in Chapter 2, and does not count them as findings. And where the actor's own text survives, it is a better witness to what the act promised than the archive's summary. The study therefore reads the source texts first wherever it has them, and the record after.

0.8The archive and the choice of cases

This study took every public-record case in four of its domains: the crisis history (35 cases), the credit history (29), and the public-record cases of the markets (14) and money (6) domains. Synthetic cases, written by the studio to test reasoning, were left out, as the series requires. The ledger thus holds 84 cases, in five groups: the crisis history (16), the credit history (29), markets (4), money (6), and 29 windows of observed prices for three crypto-assets, taken out of the crisis history and the markets domain because they record ninety days of prices up to one close and hold no actor and no act.

Of the 84, the study reads 24 closely, in the order of their dates. They were chosen because each shows an emergency act or a founding made in an emergency, because together they cover nearly four thousand years, and because each has partners along the same axis: three suspensions of convertibility (1797, 1931, 1971) and a bank holiday (1933); four rescues or refusals (1907, 1998, 2008, 2023); four reliefs of debtors or loans of grain (Hammurabi, Solon, 194, 1069). Two market records without an act, the South Sea year and Black Tuesday, were included to show what a record of a market's movement holds when no authority acts in it.

The choice is not a sample. It does not let the study say how often emergency acts in history named their end; it lets the study say how often this archive's records of them do, and what the actors' texts add where the study has them. Cases left out of the close reading (among them the Plaza agreement, the baht in 1997, Indonesia, Mexico in 1994, Argentina's convertibility law, Enron, Barings, the flash crash, the Libor fines and the withdrawal of India's old notes) are all in the ledger and in its counts.

0.9The cases at a glance

Skeptron's record of Lehman holds two entities, keeping the holding company's bankruptcy of 15 September 2008 apart from its dealer subsidiary's use of the Federal Reserve's existing credit facilities.

0.10Four literatures and where this study stands

Four literatures bear on the question, and Chapter 1 reviews them. The first, on the lender of last resort, runs from Thornton and Bagehot to the modern accounts of Humphrey, Bordo, Goodhart, Tucker and Mehrling: it asks what a central bank owes to the market in a panic and on what terms, and it has always cared about announcement, since Bagehot's rule asks that the lending be made known in advance. The second, on manias, panics and the history of crises, runs from Galbraith and Kindleberger to Minsky, Friedman and Schwartz, Reinhart and Rogoff, Diamond and Dybvig and Gorton: it asks why crises recur and how runs begin, and it treats the emergency act mostly as a response. The third, on debt, relief and the international lender, runs from the ancient law collections edited by Roth and Westbrook and the arguments of Graeber and Hudson to Tomz on reputation and the histories and critiques of the IMF by Boughton, Blustein, Radelet and Sachs and Feldstein: it asks who is relieved, who lends and on what conditions. The fourth, on rules, discretion and the emergency that stays, runs from Kydland and Prescott and Bordo and Kydland to Eichengreen, Higgs, Gross and Ní Aoláin and Posner and Vermeule: it asks whether an emergency departure from a rule is part of the rule or a break with it, and whether emergency powers recede.

All four care about the end of an emergency act. The first asks for the terms of lending to be known; the third asks when relief ends and what conditions release the next tranche; the fourth makes the end of a suspension the test of a rule's credibility and the persistence of emergency powers its central worry. None of them reads, case after case on the same fields, what an archive keeps of the end. That is this study's narrower place.

0.11The actors' texts

The study reads the public texts of the actors wherever it could have them on terms that allow them to be reproduced: works in the public domain, the works of the United States government, and the official enactments of the Swiss Confederation, which Swiss law leaves free to use. They include proclamations, Acts and addresses from London and Washington, testimony to Congress, a central bank's release, a deposit insurer's release and a joint statement, two Swiss ordinances, the ancient laws and chronicles, and the classic texts of Thornton and Bagehot; Chapter 2 lists them by chapter.

Some of these were had only as transcriptions of scanned pages that no one has checked against the print; the study gives their sense in its own words and does not quote them. Where a text is in German, Greek, Latin or classical Chinese, the English is this study's and says so. For the Fund, the European Central Bank, Britain in 1992, Mexico and Korea there was no such text, and those cases rest on the summaries and the notes.

0.12Reading the numbers

No count is given without what it is a count of. A hypothesis applies to some cases and not to others: the first, third and fifth to the cases with an act, the fourth to the cases with a figure, the second only to the few cases that call an act temporary and say what became of it. Each verdict is given with its four numbers, the cases that support it, those that go against it, those left unclear because the two readings differed or the field was not settled, and those to which it does not apply, together with the share that supports it among those that decided. The strict reading counts only what both readings found stated in the record; the lenient reading counts also what they found implied. A share that sits near the line, or rests on few cases, is said to do so.

0.13The method in brief

Every case is coded on the series' axes and on eight fields of this study's own: whether the case shows an act, its kind, whether it is presented as temporary, whether the condition of its end is named and whether that condition could be observed, what became of it, whether two dated moments are kept apart, whether a figure is given and whether the announced figure is kept apart from what was available, and who bore its cost. Each field was read twice by the same model, separately, with a value, a basis (stated in the record or implied by it) and a reason; where the two readings differ, the field is left unsettled. Three fifths of the deciding cases or more make a hypothesis supported, two fifths or fewer make it go against, and anything between leaves it unclear. The strict reading counts only what both readings found stated in the record; the lenient reading also counts what they found implied.

All of this, with seven predictions, was fixed in writing before the coding began. The machine marked as frame every sentence most of whose five-word runs recur in 5% or more of the cases of its series, and the readers were told not to read the frame as the case's own text.

The ledger is not the study's evidence about history. It is a count of what the archive's records hold. The study's evidence about the acts themselves is the records and the actors' texts, read in the chapters; the ledger tells the reader how typical what a chapter shows is.

0.14The readers and how far they agree

The two readings by one model agree on this study's eight fields in 81% to 98% of the cases, field by field (Chapter 2 gives each). A sample of 30 cases was read again by models of two other families, on the same instructions. The two readings of the first model and each further reader gave the same value in 89% and 95% of the cells both settled, and the two further readers agreed with each other in 88%. Where every family that read a case had to agree for a cell to count, the verdicts kept their direction: the first and fifth hypotheses supported, the third and fourth against, the second unclear. No hypothesis changed direction with the reader, as one did in the first study.

0.15What this volume brings to the series

The first study found, in the same archive's records of epidemics, that a public act on disease usually came before its cause or effect was settled, that the records named a gap in time between an event and its notice or count in somewhat more than half of the acts, and that they seldom named who bore the cost of an act still pending. This study finds the last of these again, more strongly: the archive's records of money and credit name the bearer of an emergency act's cost even less often than its records of epidemics named the bearer of an act on disease. It finds the second reversed: the records of money and credit keep two dated moments apart less often than the records of epidemics kept the time of an event and its record apart. And it adds a question the first study could not ask, because acts on disease seldom say how long they will last: the question of the end. Chapter 13 sets the two studies side by side on the first study's own published numbers.

0.16Three contributions

The study makes three contributions, each narrower than the literatures it draws on.

First, it counts, on one set of fields, what a present-day archive preserves of emergency acts from Babylon to 2023, and finds the act preserved and its term, its losers and its fate mostly lost.

Second, it reads the actors' own texts beside the archive's summaries for the forms in which an emergency act named its end (a date, a period, an event, another body's decision, nothing), and shows that the archive's summaries keep the form less often than the texts state it.

Third, it adds a second domain to the series on the same archive and the same method, so that the difference between the two studies can be read as a difference in the records of two kinds of decision.

It does not add a new history of any crisis, a model of panics, or a claim about how often emergency acts succeed.

0.17What the evidence cannot show

The records are the archive's summaries, written in 2026 from the sources the studio chose; a summary that does not name a bearer does not show that the act's own texts were silent. The source texts the study could obtain are uneven: the actors' words survive and are free to reproduce for some cases and not for others (the IMF's and the European Central Bank's statements, the British statements of 1992 and the Mexican and Korean governments' documents were not had), and some of those obtained are transcriptions not checked against the printed page, which the study reads for their sense and does not quote. One model coded the ledger twice; other families read only a sample of thirty cases. The cases were chosen, not drawn, and the counts describe this archive, not the history of emergencies. Counts about acts leave out the twenty-nine windows of prices. The study was written by AI from the records, the source texts and the reading notes, and its readings of the scholarly works go no further than those notes allow.

0.18The plan of the study

Chapter 1 reviews four literatures: on the lender of last resort and the central bank, on manias and panics, on debt, relief and the international lender, and on rules, discretion and the emergency that stays. Chapter 2 describes the archive, the ledger, the frame, the study's own fields and hypotheses, the source texts, and two cases coded field by field.

Part II reads the emergency act by its kind. Chapter 3 reads suspensions of payment and convertibility (1797, 1931, 1933, 1971); Chapter 4 reads lending and rescue in a panic (1907, 1998, 2008, 2023); Chapter 5 reads the relief of debtors and the lending of grain (Hammurabi, Solon, 194, 1069); Chapter 6 asks who pays while an act lasts (Diocletian's edict, the South Sea year, Mexico in 1982, Greece in 2015).

Part III compares across the cases. Chapter 7 reads headline sums and what could be drawn; Chapter 8 reads three foundings made in or after emergencies (1694, 1913, 1944); Chapter 9 reads pegs, rates and regimes abandoned (1931, 1971, 1979, 1992); Chapter 10 reads market records with no act, the crash of 1929, the South Sea year and the price windows.

Part IV draws the threads together. Part IV asks which endings could have been checked, gives the verdicts and the failed predictions, and sets the results beside the literature and beside the first study; the conclusion closes.

0.19How the study was made

The study was made in one way from beginning to end, and it says so here once. Its design (the question, the cases, the fields, the hypotheses, the plan of the chapters, the list of works with a reading note for each, and the source texts) was set by an editorial session before any chapter was written. The ledger was coded by machine as described above. The chapters were drafted by a model from the records, the source texts, the reading notes and the ledger's numbers; checked against the records and by machine; read by models of other families, whose findings were judged against the text; and corrected. This introduction, the preface and the conclusion were written by the editorial session from the plan, the ledger and the chapters. Everything the study says about the past rests on the records, the source texts and the reading notes, and nothing else.