This chapter sets the ledger's counts beside the four literatures that Chapter 1 mapped, and beside the first study of this series. The counts come from "The ledger's numbers"; Chapter 12 gives the verdicts on F1 to F5 and on the series' hypotheses in full, together with the predictions registered before coding. The readings of the works come from their reading notes, and where this study draws a conclusion from them it is marked as this study's reading. Where a debate is treated in full elsewhere in this study, the topic map sends the reader there; what follows is only what the records add to each debate, and what they cannot reach.
13.1Beside Bagehot
Lombard Street puts the duty in its own words: "we must keep a great store of ready money always available, and advance out of it very freely in periods of panic, and in times of incipient alarm" (chapter II). The book's explanation of the rate is that "Very large loans at very high rates are the best remedy for the worst malady of the money market when a foreign drain is added to a domestic drain" (chapter II).
For a later reader to judge whether that rule was kept, a record of a rescue would have to hold who asked, on what security, at what rate, until when, and who was refused. Chapter 4 states that requirement and then reads five rescue-era records against it. What this chapter adds is the ledger's count of what such records hold when read on the study's own questions.
Those records are dated acts. The 1907 record holds a private clearing house that kept its resources for its members and refused the Knickerbocker Trust Company, a non-member, whose solvency could not be settled in hours; the trust then suspended business on 22 October. The 1998 record holds private firms agreeing to put in about 3.5 billion dollars, the old holders diluted and control passed to a committee, with testimony that no Federal Reserve money was offered. The March 2023 records hold a bank closed and the Federal Deposit Insurance Corporation named receiver — a receiver being the official appointed to wind up a failed institution — and, in the other, a bank takeover supported by two liquidity lines, whose limits the record sets apart from the amount available, one with preferential rank in bankruptcy and one with a federal guarantee. Emergency liquidity assistance, in the Greek case read in Chapter 6, is lending by a central bank to banks shut out of the market, within a ceiling set by the bank's own governing body.
The records usually name what was done and seldom name what would end it or who bore its cost. Among the twenty-four cases read closely, fourteen name no condition on which an act would end or be revised, four name one that could be observed, and seventeen name no bearer.
What they do not hold is the substance of Bagehot's rule. The rescue records give no rate, no security and no term. The elements the classical rule turns on — a high rate, good security, an announced intention, a term — are the elements these records most often fail to carry.
The questions that ask what would end an act are the questions the records leave empty. The Credit Suisse record's two lines and the Greek ceiling case show support beyond ordinary bank discounting, but no record here gives repo terms, rates or durations. The 1907 record shows the gap a private clearing house could not close for an institution outside its membership.
What the records cannot say is whether any rescue worked, whether any refusal was right, or how often a panic spread. The ledger holds dated records of acts; it holds no outcome field for any rescue, and this study does not supply one.
13.2Beside Kindleberger and Minsky
The mania tradition asks what a financial crisis looks like over time: how a shock becomes a boom, how the boom becomes a panic, and how a lender's intervention fits the sequence. This study asks a documentary question of a different kind: does the record of one dated act hold that act's term, the condition of its end, and the name of who bore its cost while it stood. What this section adds to the mania debate is the answer the crash and market records give to that second question, and how far that answer can reach the first.
Charles P. Kindleberger and Robert Z. Aliber's Manias, Panics and Crashes: A History of Financial Crises (2005), in its fifth edition, is a narrative synthesis of many cases, not a formal test. Its reading note says the book is built on Minsky's model of displacement, boom, euphoria, distress and revulsion; that Kindleberger argues crises recur in recognisable patterns rather than as random shocks; that he holds an international or domestic lender of last resort is needed to stop a panic from spreading; and that the timing and wording of its announcement matter. Minsky's Stabilizing an Unstable Economy (1986) supplies the mechanism: a long calm encourages more fragile finance — hedge, speculative and Ponzi finance — so that a small shock can set off a crisis, with a big government and a central bank acting as stabilisers. Chapter 10 reads both against the study's crash and market records in full.
Such a narrative requires a sequence with dates and behaviour at each stage; these records hold a dated act or a dated run of numbers. A record of two dated days can be placed at the distress stage; it holds neither the displacement nor the boom the narrative needs, and it holds no actor whose timing a narrative could read.
The South Sea record of 1720 is read apart the same way. Both readings find an act in it but differ on its kind, and the record itself keeps the company's charter, its right to take over part of the national debt, its trading privileges and its share price apart from one another. It presents no term, names no condition on which the year's arrangements would end, gives no figure that could be drawn, and names no group who bore the cost while the year ran; it is one of the four close-read cases whose kind of act the two readings could not settle. Across the ledger the silence is wider: none of the eighty-four records names a condition on which its own judgment would be revisited, though eighty-two carry a date or a period by which something must be done, and none names a decider in its own text. The mania tradition asks what follows a boom; these records ask what an act was written to do, and on that question most of them say nothing.
Skeptron's price windows record observed prices up to one closing price; they hold no decision and no actor, and this study reads them only as a group. The ledger gives them their own place beside the crisis and credit records because they hold nothing the others do: twenty-nine cases of ninety days of observation for Bitcoin, Ether or Litecoin, in twenty-eight of which the record shows no act and in the remaining one the reading is unsettled. On this section's question the group leaves everything undecided: none of its cases bears on the prediction that most cases with an act name no observable condition of ending. On the prediction that a record sets an announced figure apart from what was available to be drawn, sixteen of the group go against it and thirteen are undecided, none supporting; that count describes what the group contains, not the honesty of any figure. A window cannot hold a stage of a mania: no euphoria, no distress, no announcement, no lender.
This study's findings sit beside the mania tradition at the point Kindleberger's note makes, because an announcement's timing and wording matter to his account, and the announcement is a moment a narrative needs in its own right. On the whole ledger the prediction that most acts name two different dated moments among their announcement, their taking effect and the settlement of their amount goes against: twelve cases name two such dates, twenty-three do not, twenty-six are unclear and twenty-three do not apply, and a count that also reads what is implied gives thirteen in favour and twenty-eight against. Most acts in this ledger arrive as a single dated statement. That is a finding about what a record holds at the date it is written; it cannot say whether an announcement's wording helped or hurt, and this study does not read it that way.
Minsky's mechanism is a claim about what precedes a crisis: calm making finance fragile until a shock finds a system built on short-term claims. These records are dated at the act, and so cannot show what the calm was or what it produced; where they carry a figure it is a limit, a ceiling or a loan, not a state of confidence. Kindleberger's narrative across many episodes and Minsky's hypothesis about the sequence before a shock are both accounts of behaviour over time, and a count of records measures only what each record contains. On the stages of a mania, these records contain very little.
13.3Beside Reinhart and Rogoff
Chapter 6 reads the debate about who pays in a debt crisis, including Reinhart and Rogoff's catalogue, against this study's records of Mexico in 1982 and Greece in 2015. What this section adds is narrower and documentary: it sets a count of episodes beside a count of records, says what each of the two counts can hold, and reports the study's own counts in plain words against that comparison.
Reinhart and Rogoff's book is a catalogue: its object is the episode, counted and placed in a sequence.
A count of crises and a count of records measure different objects. A count of crises measures how often events occurred, across countries and centuries, and with what aftermath: defaults, restructurings, long downturns. A count of records measures, for one dated act, whether the record holds the act's term, a condition on which it would end, and a named bearer of the cost. One counts events; the other counts what a written record contains about an act. Neither can be turned into the other, and this study's argument stays on the second object, saying so where the two meet.
The two records named in the opening sentence show the difference from the inside. The Greek record is set at 28 June 2015 and concerns the stoppage of an increase in emergency liquidity assistance — lending by a central bank to banks shut out of the market — under an existing ceiling; the record gives no figure for that ceiling, names no condition on which the ceiling or its stoppage would end or be revised, and names no one who bore the cost while it stood. A catalogue entry for either country could record the episode and its aftermath; it could not tell a later reader whether the actor wrote a term.
The counts across the whole ledger answer this study's own question in the same documentary terms. Among the twenty-four cases the study reads closely, fourteen name no condition of ending. Reinhart and Rogoff's catalogue covers centuries of financial crises across many countries — sovereign defaults, banking crises, currency crashes and inflation — and sets each episode in a sequence of defaults, restructurings and recoveries; this study's twenty-four close cases ask what each record holds at the moment of its act. These are counts of what records hold, not judgments about the episodes.
What Reinhart and Rogoff's catalogue supplies that these records cannot is the aftermath — what happened to defaults, restructurings and recoveries after the events, recorded across many countries. What the records supply that a catalogue cannot is the wording of an act at its moment: whether the actor wrote a term, a condition of ending, or a named bearer. A catalogue cannot show what any single record holds; a count of records cannot show how often crises recur. The first study of this series, Meridian's Before the Count: Public Decisions on Epidemics and What Their Records Allow, c. 1600 BC – AD 2020 (2026), made the same comparison in a different domain from eighty-nine records of its own; its stricter counts left all five of its hypotheses unclear, and it reports that four of its seven registered predictions failed. Its counts are of its own cases and questions, not of this study's.
One further place the two literatures touch is Reinhart and Rogoff's "this time is different", which is about belief at the top of a boom. This study reads the absence of any condition of ending as a documentary fact in the act's own words, not as evidence of belief. What the counts cannot settle is whether any deferral, ceiling or rescue worked, how often a panic spread, or what any authority's record of motives held; the ledger holds no field for outcomes, and this study supplies none.
13.4Beside the ratchet
Emergency acts in these eighty-four records almost never state what would end them, and that is the fact Robert Higgs's argument about government in crisis most needs and the records least supply. This section adds a documentary point the argument does not turn on: how few of these records hold the words that would let a later reader see whether an emergency act outlived its emergency.
To test the ratchet, a reader needs three dated moments: an act taken in an emergency, the moment the emergency ended, and a record of whether the act continued after that moment. The ratchet's claim is about what happens after the emergency; Higgs's evidence, by the note, is episodes from the Civil War to the 1970s read for the size and powers of government. A record written at the act supplies the first moment by definition. It supplies the second only if someone wrote down when the emergency was declared over, and the third only if the record says what the act became.
The counts answer in plain terms. Of the eighty-four cases, twenty-eight record no condition on which an act would end or be revised, and thirty-eight say nothing at all about what became of it; one case records that the act ended as stated, one that it was reversed or ended early, three are left unresolved, and forty cases carry no act to report on. The hypothesis closest to Higgs's claim — that where a record presents an act as temporary and says what became of it, the act usually lasted longer than first said — is not settled here: in five cases the answer is left unresolved and none of those five supports or refuses it, and under the more generous count one case goes against and three are unresolved, because the condition of entry — a temporary act that also says what became of it — is rarely met. The number of records that hold both an end and a continuation is one of forty-two.
Records that seldom name an end cannot test the ratchet on the claim itself: they cannot show that an act outlasted the emergency that produced it, because they rarely record when the emergency ended or what the act became. What they can show is narrower and documentary, and it matters because both sides of the ratchet debate assume that emergency acts come with a stated limit. Higgs's argument is about institutions' size after crises; the question put to a record here is what was written into it at the time. The two meet only where a record holds an end and a continuation together, and that number is one of forty-two.
Where the actor's own text survives, the comparison can be turned around. Skeptron's records are modern archive summaries of public sources; where an actor's own words survive they are read here as evidence of what the actor wrote, not as the record's reading of it. The 1998 case carries an actor's account of what the arrangement was not, in Alan Greenspan's testimony of 1 October 1998: "This agreement was not a government bailout, in that Federal Reserve funds were neither provided nor ever even suggested." In the records read here that naming is what is missing.
The cases here are heterogeneous in their legal form: a proclamation with a stated term and a statute that approved actions after the fact in 1933, a multilateral body's board approving a lending programme in 1997, a receivership in 2023, a private recapitalisation in 1998 arranged without public money. A count of records cannot decide which model a case belongs to; Chapters 3 and 4 read the cases. What the count can say beside them is this: across these eighty-four records the act's own words usually name what was done and seldom name what would end it, and where an end is named — in a proclamation, a statutory exception, an ordinance with a term — the naming sits in the actor's document rather than in the summary an archive later keeps of it. Whether any of these acts outlasted its emergency is a question about institutions' size over time, and these records do not answer it.
13.5Beside the rule with an escape clause
Chapter 3 treats the rule with an escape clause and the four suspensions; Chapter 9 treats Kydland and Prescott's rules against discretion and the four departures from rules.
The four suspensions in this study's cases are the Bank Restriction of 1797, Britain leaving gold on 21 September 1931, the bank holiday of 1933, and the closing of the gold window on 15 August 1971. Their records are not alike on this axis. The 1797 record holds that the Bank of England's obligation to exchange its notes for gold was suspended; Chapter 3 reads the terms of resumption in the Bullion Report of 1810. The 1931 record is accompanied by the Act's own words: "Unless and until His Majesty by Proclamation otherwise directs", a suspension that has no term, with only the Treasury's power over the exchanges given a term of six months. The 1933 record holds a proclamation declaring the holiday from Monday 6 March to Thursday 9 March, a statute of 9 March approving and confirming what had been done, and member banks in Federal Reserve cities licensed to reopen from 13 March. The 1971 record holds Richard Nixon's address of 15 August, which describes the suspension of convertibility as temporary: "suspend temporarily the convertibility of the dollar into gold or other reserve assets" (the address of 15 August 1971).
This study reads those four records against the escape clause, and the reading is its own. On the clause's terms, a suspension works as a rule's escape only if the actor's own words promise a return — a term, a condition or a date. Among the four, the ledger's close reads give the 1797 case as stating a term with an observable exit and an outcome of ended as stated, and the 1933 case as calling the act temporary without a term, with an observable exit. The 1931 and 1971 cases differ on how the act is presented and name no exit. The asymmetry is documentary: two records carry a term or a mechanism of return, two carry none in their own texts. The records give no reasons for the lengths, and Chapter 3 states that limit plainly.
Where the records bear on the debate is narrow but real. Bordo and Kydland's clause is a claim about what a suspension must mean for the public to trust it; a record can show whether the words of the act gave a term, a condition or a date, and whether the record later says the act ended as stated. That is a test of the clause's documentary requirement, not of its success. Barry Eichengreen's Golden Fetters: The Gold Standard and the Great Depression, 1919–1939 (1992), by its note, argues that the interwar gold standard transmitted and prolonged the depression, that countries that kept to it were forced into deflationary policy, and that recovery began after countries abandoned gold and devalued; Eichengreen and Jeffrey Sachs's Exchange Rates and Economic Recovery in the 1930s (1985), by its note, argues that countries leaving gold earlier and depreciating recovered sooner, while devaluation helped one country at some cost to others. These are claims about consequences, and the ledger holds no field on consequences.
Where the debate cannot be settled from these records is where Kydland and Prescott's argument lives: over time, in expectations and incentives. A record of a suspension is a statement at one date that rarely states both a term and its outcome, so F2 cannot test the argument (see 13.4). That is a finding about records. It cannot decide whether a rule with an escape clause or discretion served any of these countries better.
13.6Beside the clean slates
This study's four oldest records are not one kind of document. Hammurabi's law is a conditional rule from a reign range of 1792–1750 BC, with the individual enactment date undetermined; the record notes that no judgment under the rule is recorded. Solon's release is recorded as later historians tell it, in an archonship of 594 BC, with the ban on loans on the security of the debtor's person, the cancellation of debts and the redistribution of land kept apart. The spring grain loans of 194 are an entry in a chronicle compiled long after the event, giving the order's seasons without a rate of interest and without a result. The green sprouts loans of 1069 are a proposal, not an execution. Chapter 5 reads each in full.
What these records hold that the clean-slate debate can use is the wording of the acts themselves. Hammurabi's law names its own condition and its own term: "If any one owe a debt for a loan, and a storm prostrates the grain, or the harvest fail, or the grain does not grow for lack of water; in that year he need not give his creditor any grain, he washes his debt-tablet in water and pays no rent for this year" (law 48, translated by L. W. King, 1910). A rule that names the storm and that year is a record of a rule, not of a cancellation carried out. For Solon, Aristotle's Athenian Constitution (translated by F. G. Kenyon, 1891) says: "Solon liberated the people once and for all, by prohibiting all loans on the security of the debtor's person: and in addition he made laws by which he cancelled all debts, public and private." The chronicle of 194 gives the loan's seasons — grain lent by household size from the third to the seventh month and taken back in the tenth — and names no rate and no result. The proposal of 1069 gives repayment in halves with the summer and autumn taxes, in kind or in coin, deferred to the next harvest in a disaster, and is a proposal rather than an execution.
The ledger's close reads for these four are telling in what they leave out. On how the act is presented, three of the four leave the question open and one gives no answer. On the condition of its end, one names an observable one, two leave it open and one names none. This study reads that combination as the documentary shape of the clean-slate problem: the ancient rule names a harvest and a year, and the later documents name neither a bearer nor an outcome.
Beside Graeber's and Hudson's arguments, that shape limits what a count can settle. Their claims are about practice across centuries — whether debt cancellation was a recurring institution of rule. Four records, of which one is a conditional rule, one a later history, one an annal entry and one a policy proposal, cannot test recurrence. What they can show is what the texts say, and the ledger's counts say that the records hold conditions and seasons rather than bearers and outcomes. Whether such practices recurred is the debate's own question. Westbrook's collection covers both the legal collections and the practice shown in contracts and records, and this study reads the two of them together when weighing these four records.
13.7What the archive adds and what it cannot
Each holds a dated event, a scenario and five answers, with the public source Skeptron checked; Chapter 2 sets out what a case is, how the ledger's codings work, what the studio's recurring sentences are and why they are not counted as findings.
What a later archive of this kind adds is comparability. Every case is read on the same set of questions, so the eighty-four cases can be counted together in a way that twenty-four narrative chapters cannot be. The two readings are of one model run twice; the lowest agreement, on figures, is 68 of 84, and the agreements are reported in full in 2.2 and 12.5. The Swiss ordinance and its amendment are in German, the statement of 17 October 1979 is read from an uncorrected transcription, and the Privy Council's minute is printed in Cobbett's Parliamentary History in text marked as uncorrected.
What it cannot add is equally clear. A record is Skeptron's reading of a source, not the source itself, and where the original could not be reached the reading rests on the reading notes. One model run twice produces two codings of the same model, not two independent readers. Twenty-nine of the eighty-four cases are price windows that hold no decision and no actor. And the counts describe what records hold; they hold no field on outcomes, no field on rates, and no field on whether any act was right or wrong.
Beside this study stands the first study of the series, Meridian's Before the Count: Public Decisions on Epidemics and What Their Records Allow, c. 1600 BC – AD 2020 (2026). By its own account it was made in the same way from Skeptron's records: twenty-three cases read closely and eighty-nine read through one model run twice, on its own set of questions and its own hypotheses E1 to E5, registered before the counts were made. Its question, as this study reads it, was: when a public body acts on disease — to close a gate, to count the dead, to try a remedy, to notify, to declare — what was established at the hour it acted, and does the record let a later reader keep apart the time of the act, the time of the evidence and the time of the count? Its table records each of its five hypotheses as unclear when counted strictly; counted leniently, it records E1, E2, E3 and E5 as supported and E4 as going against, with shares from 36% to 83%, and it reports that four of its seven registered predictions failed. Its counts are of its own eighty-nine cases on its own questions; they are not counts of this study's cases. Where a verdict word and a share in either book's table do not sit on the same scale, this chapter reports each as the table gives it.
The two studies differ in object more than in method. The epidemic study's records are of acts on disease, and its fields ask whether the record separates the times of act, evidence and count. Where the two touch is on the shape of a record: both found their stricter counts weaker than their more generous ones, and both found that the case's own text often names less than the question needs. This study's F3 and F4, both recorded as going against when counted strictly, stand beside five hypotheses that the epidemic study recorded as unclear when counted strictly.
What neither study can supply is the outcome of any act, the reasons any authority gave, or the identity of any private person who bore a cost. The counts answer the study's own question about records, and the Conclusion sets the answer beside the study's limits and beside the conditions that would change it.