Chapter 1 — When a Shock Finds a Balance Sheet
Prologue. The Obligation That Arrives on Monday
The public date of a crisis is usually the date on which a price falls, a bank closes or a government announces emergency powers. An obligation has another calendar. Wages are due at the end of the week. A customer expects a transfer in the morning. A fuel supplier requires assurance before loading a truck. The collision between those calendars is where an economic disturbance becomes a crisis for people who did not make the original bet.
In March 2023, the failure of Silicon Valley Bank made this distinction unusually plain. The bank's concentration of uninsured deposits, interest-rate exposure and rapid withdrawals belong to the story of its failure. They do not exhaust the story of the next business day. A payroll firm, a startup treasurer and a regulator faced different questions. The first needed an operational payment rail; the second needed to know whether its deposit would be available; the third had to decide which claims and functions public authority could protect. The authorities' March 12 announcement said depositors would have access to all their money beginning Monday. The same announcement said shareholders and certain unsecured debtholders would not be protected.[P-1] A function could be restored without preserving every prior claim.
That distinction is the book's starting point. A shock is an event or change: a rate rise, a freeze, a sudden loss of demand, a default. It becomes a crisis when obligations built for one set of conditions can no longer be performed on time, and when ordinary adjustment cannot bridge the gap. The contracts are linked. A lender's collateral is another person's asset; a household's payment is a bank's expected cash flow; an electricity generator depends on gas, water, communications and crews. A failure in one place can make prudent behavior elsewhere look like panic, because waiting to see what happens may mean missing a deadline that cannot be recovered.
Neither the language of contagion nor the language of confidence is enough. Contagion can suggest a disease that moves without choices. Confidence can suggest that a better speech would repair a missing asset. Both can be useful shorthand, but the account needs the contracts underneath: who can demand payment, which collateral is actually deliverable, who has legal authority to suspend a claim, and what loss appears when the books are marked honestly. This book follows those questions through banks, sovereign balance sheets, trade finance, households and a cold-weather energy failure. The cases are different enough to keep the concept of crisis from becoming a synonym for a bank run.
An economic crisis also has several endings. A market may stop falling. A bank may reopen under a new owner. Workers may still lose shifts, borrowers may wait years for a workout, and a regulator may discover that the records needed to explain the episode were never kept in a usable form. A rescue can preserve a service while making a political choice about whose loss is delayed. A liquidation can enforce a credible hierarchy while interrupting a service that cannot be replaced by Monday. Those judgments should be made with their operational and distributional costs visible.
The chapters do not offer a universal crisis rule. The evidence is drawn from specified primary reports and public decisions, with their dates and limits recorded in notes. When a scenario is constructed to test a mechanism, it is labeled as such. The aim is to understand where the next obligation lands, which person can act before it comes due, and how the eventual recovery allocates the claim that could not be paid.
Notes and Sources
[P-1] US Department of the Treasury, Board of Governors of the Federal Reserve System and Federal Deposit Insurance Corporation, joint statement, March 12, 2023, https://www.federalreserve.gov/newsevents/pressreleases/monetary20230312b.htm. Original public statement read for its Monday-access and claimant-treatment language; it does not establish subsequent payment completion for every depositor.
Chapter 1 — When a Shock Finds a Balance Sheet
The shock does not arrive at a blank page. It reaches firms and households that already owe money, hold assets they may or may not be able to sell, and have made plans around receipts that have yet to arrive. Two organizations can experience the same fall in sales or the same rise in interest rates and face very different choices the following morning. One has cash, a line of credit it can actually draw, and customers who will pay within the month. The other has a warehouse full of goods, a loan due on Friday, and a lender asking for more collateral. Calling both organizations distressed conceals the decision that matters: which obligation must be met before the asset can earn or be sold for a sensible price?
Consider a constructed example. A regional equipment supplier has financed a fleet of specialist machines with loans that mature sooner than its service contracts. Its customers pay reliably, though only after their own projects reach agreed milestones. When interest rates rise, the quoted value of the fleet falls. The supplier can still maintain the machines and serve customers. Yet a loan renewal is due before its next large receipt. If the lender refuses to renew, a sale may have to take place in a week. A buyer who knows the deadline can offer less than the fleet would be worth to an operator with time to use it. The sale then weakens the supplier's accounts further, making another lender more cautious. None of this proves that the fleet was worthless or that the supplier should have been kept alive at any cost. It shows why the order of payments, valuations and decisions belongs in the story of a crisis.
The same sequence can run the other way. Suppose the supplier's service contracts have become unprofitable because a cheaper and better machine has entered the market. A lender who grants more time may merely finance another period of losses. The assets' old valuation cannot settle that question. The analyst needs to know what the machines can now earn, how quickly they could be adapted, and whether an alternative owner could use them. Distinguishing a financing interruption from a failing business is difficult precisely because the forced-sale price can worsen the business while the prospect of further losses can make patient finance unreasonable.
The rate change and the run
The Basel Committee's October 2023 report on the banking turmoil provides a particularly sharp historical example, though its subject is banking rather than equipment. It describes how Silicon Valley Bank Financial Group grew rapidly during a period of venture capital activity and low rates, taking in largely uninsured deposits and investing heavily in longer-maturity securities. As rates rose in 2022, deposit withdrawals and unrealized losses grew together. The institution had an exposure to both the price of its assets and the timing of its liabilities.[1-1]
On March 8, 2023, the group announced a balance-sheet restructuring. According to the committee's account, it had sold $21 billion of available-for-sale securities, expected an after-tax loss of $1.8 billion, planned additional term borrowing and sought to raise $2.25 billion of capital. The following day, withdrawals accelerated. The committee reports more than $40 billion of deposits leaving on March 9, while management expected further outflows on March 10. The actual and expected amounts should not be collapsed into a single observed figure.[1-2]
The announcement was an attempt to improve the balance sheet. It also revealed the cost of doing so at the moment when depositors were deciding whether to wait. A depositor with a large uninsured balance did not need to calculate the bank's final recovery value to prefer an immediate transfer. If many similarly placed depositors made that choice, the bank's ability to wait for a capital raise diminished. The composition of its customers mattered: the report describes a concentrated network of venture investors and technology firms whose funding concerns could move together. The loss on securities was one part of the sequence; the alignment of withdrawal incentives turned a balance-sheet problem into an operating deadline.
This is a historical explanation offered by an official review, not a proof that every bank with long assets and uninsured deposits will fail. The committee also describes other distressed banks with different histories, and its comparison of withdrawal speeds warns against reading a common threshold from unlike institutions. A bank can hold long assets and manage their risks; a depositor base can be diversified; a capital raise can succeed. The relevant test is the joint arrangement: what must be paid, how fast claims can leave, what assets can be sold or pledged, and which parties can act before the next deadline.
The discount window illustrates the importance of operations. The committee says SVB had not tested its ability to borrow there in 2022 and lacked suitable collateral and operational arrangements when time became scarce. It adds a qualification: better readiness would probably not have prevented failure, though it might have made resolution more orderly.[1-3] A theoretical source of liquidity is not the same thing as cash available on the required day. Documents must be in place, assets eligible, staff able to execute, and the receiving institution ready to process the transaction. In an acute crisis these details stop being back-office matters.
What the balance sheet cannot tell us alone
A balance sheet records claims at a date. The crisis asks how those claims behave over time. A bond can be held until maturity by one owner and sold tomorrow by another. A deposit may have no fixed maturity in the accounts, yet its holder can demand payment at once. A line of credit may look like available finance until its conditions are tested. The distinction between an accounting value, a market quote and proceeds from a large urgent sale can become decisive within days.
The balance sheet also does not explain why counterparties change their minds. They may learn new facts about underlying losses. They may see another participant withdraw and conclude that waiting is becoming more dangerous even if their estimate of ultimate value has not changed. They may need cash for their own commitments. A business customer whose payroll is due on Monday faces a different practical problem from an investor able to wait for receivership recoveries. Calling both reactions panic would erase the reason each acted.
The Federal Reserve's April 2023 self-review makes the joint exposure unusually plain. At the end of 2022, the group's securities were 55 percent of assets, against 25 percent for the review's defined large-bank peer group. Uninsured deposits were 94 percent of deposits, against 41 percent for that same peer definition. Its reported common-equity tier 1 ratio was nevertheless 12 percent, above the peer group's 10 percent. These figures are not a contest between a true risk number and a false one. They answer different questions. A regulatory capital ratio relates a specified measure of capital to risk-weighted assets; the deposit share says something about who could demand funds and what part of their balances was outside ordinary insurance coverage. The securities share says something about the asset side, though its maturity, hedging and sale conditions require further inquiry. An analyst who chose the strongest-looking figure and stopped there would miss the combination.[1-4]
Even a more complete dashboard would not make the sequence automatic. The Federal Reserve review describes 31 open supervisory findings at the end of 2022. A finding records a weakness and a process for addressing it; it is not itself a repaired control. The same review says that supervisors recognized some risks without acting with enough urgency. Its preface is candid about the inquiry's boundaries: this was a self-assessment prepared quickly with hindsight, and it examined the bank's evolution and oversight through March 8 rather than the later closure and federal response.[1-5] Those limits matter when asking what could have been done before the run. They prevent the retrospective report from becoming an imaginary real-time instruction manual for every person involved.
The more exact problem was how the bank measured exposure. In the Federal Reserve's account, managers gave weight to a measure of net interest income over roughly a year. Rising rates could improve that measure under their assumptions about how slowly deposit costs would rise. A separate measure, economic value of equity, estimated the present value of future asset and liability cash flows under rate changes. It was sensitive to a longer horizon and to assumptions about how customers would behave. Neither metric is a photograph of what depositors will do. Yet choosing the shorter measure for the board's risk appetite meant that a different warning could remain outside the full board's regular view, even as the bank took decisions about duration and hedging.[1-6]
The difference is not merely a technical argument between two models. A manager who expects deposits to remain cheap can see rising rates as a chance to earn more on assets. If depositors begin spending down cash or moving it to higher-yielding alternatives, that assumed advantage can narrow. At the same time, a fixed-rate security bought when yields were low loses market value as yields rise. Holding it to maturity may preserve the scheduled payments, provided the institution can finance the interval. A depositor's demand for cash can make that proviso unavailable. The short-horizon earnings story and the longer-horizon valuation story meet at the cash desk.
The Federal Reserve review says the firm's interest-rate policy had limited scenario and sensitivity requirements, and that the full board was not shown years of economic-value-of-equity limit breaches. It also records that, by November 2022, supervisors described the firm's interest-rate simulations as unreliable because net interest income and margin had fallen while the model predicted increases.[1-7] A forecast error of that kind deserves more than a correction to the next spreadsheet. It calls into question the assumptions behind positions already taken and the authority that allowed them to persist. A board cannot oversee a response to a breached limit it has not seen. A supervisor who has seen fragments of the problem must decide when fragmented warnings amount to a reason to act.
There is still a limit to the inference. We can say the review found serious measurement, communication and response failures. We cannot calculate from it the deposit withdrawals that would have occurred under a different board packet, nor assume that a better model alone would have supplied capital by March. A serious account should allow the criticism to stand without converting it into a precise counterfactual the evidence cannot support. The failure involved a business model, asset choices, funding composition, management, supervision and a rapid run. Making one ratio or one meeting carry the entire cause would simplify the history at the point where the mechanism matters most.
An exposure that never becomes a crisis
The constructed equipment supplier can show the other side of the mechanism. Suppose its borrowing is staggered across several dates, its service customers have different payment cycles, and it keeps one tested source of working capital whose terms do not depend on the current resale quote for the machines. A rate rise still reduces the value of long-lived assets. It may still force the firm to cancel a project with a poor expected return. It need not require a hurried sale of equipment used in profitable contracts. The organization has arranged its obligations so that it can make an informed choice before a deadline removes that choice.
That flexibility has a price. More cash held in reserve means less money for equipment or staff. Longer-term borrowing can cost more. A committed credit line may carry fees and covenants. A lender will want information, collateral or limits on distributions. The point is not that every firm ought to buy the maximum possible time. It is that a financial saving measured during ordinary months may be the sale of an option the firm later needs. Management should be able to say whose payment will come due first under strain and what can actually be done before it does.
A firm can have ample cash and still be on a path it cannot sustain. The money may allow management to postpone a painful choice about an obsolete service or a product that consumes more resources than it earns. A reserve then protects the organization while its useful activity shrinks. Calling the firm liquid does not settle whether the time purchased will be used to change the business, transfer assets, renegotiate claims or continue the old losses. Liquidity buys time; the business still has to make that time valuable.
The first task in reconstructing a crisis is therefore modest and exacting. List the obligations in the order they could come due. Identify which assets could produce cash on that timetable, not merely which were valuable under ordinary conditions. Record what each decision maker knew at the time and what authority they had to change the schedule. The first loss may occur before the headline failure. It may be the sale at a discount, the cancellation of a viable order, or the departure of an employee who cannot wait for the next payroll. Those losses can be consequences of the original shock, or of the arrangement that made time unavailable.
The constructed supplier might survive by renegotiating one maturity and serving its customers. It might instead need to sell the fleet to a better operator. A bank may obtain liquidity while it raises capital, or its claims may need to be transferred through resolution. No single answer follows from the word crisis. The response has to give a sound activity time to continue while making the owners and creditors of a failed arrangement bear losses according to defensible rules. That judgment becomes harder when the price of the asset is formed in a market where other holders also need to sell by Friday.
The calendar hidden in a line of credit
A balance sheet drawn at midnight can imply more flexibility than the treasury team will have at nine in the morning. To see why, return to the constructed equipment supplier. Its revolving credit agreement permits a draw only if specified representations remain true and no event of default is continuing. The next customer milestone payment is due in three weeks. The lender's compliance certificate is due in two days. If the supplier has already breached a covenant tied to asset values, the apparent unused credit line may not be available when payroll arrives. The relevant cash buffer is not the sum of its bank balance and the headline size of the facility. It is the cash and committed borrowing it can actually obtain after the contract's conditions are tested.
The lender has a calendar too. A credit officer may favor a waiver, but the waiver can require an internal committee, an updated appraisal and consent from another lender. A person with authority to approve the commercial relationship may lack authority to alter the security package. The appraisal may use a sale price from an inactive market. Neither a favorable business forecast nor a sympathetic banker makes the documents executable before the deadline. The firm could seek bridge financing, but a new lender will ask which creditor has first claim on the machines and whether the service contracts can be assigned. The legal right to take collateral is valuable partly because it determines who can move quickly.
This is a constructed scenario, not an account of SVB or of any named company. Its purpose is to locate the frictions that disappear when an analysis says only that a solvent borrower 'could refinance.' That phrase skips the certificate, appraisal, intercreditor consent, collateral search and transfer of funds. Each can be routine in ordinary months and decisive in a week of uncertainty. If the supplier's customers also depend on the same lender, their delayed payments can make the refinance harder just as the supplier needs it most.
A supplier may also have contracts that permit a temporary reduction in output without a penalty, customers who prepay for maintenance or assets that another operator can use immediately. Such features can turn a costly interest-rate move into a manageable adjustment. They have their own price: prepaid customers expect a discount; flexible contracts may win fewer bids; readily transferable machinery may have a lower return than specialized equipment. Resilience is an arrangement of trade-offs, not a free reserve waiting outside the market.
Four values for the same asset
Suppose the supplier's fleet has four defensible numbers attached to it. The accounts carry a depreciated book value. An appraiser estimates what a patient buyer would pay for the fleet as a going concern. A dealer quotes an individual machine, assuming it can inspect, transport and resell it. An auctioneer estimates what the whole fleet might bring if the sale must close before the lender's Friday deadline. These numbers need not agree. The book value follows accounting rules; the going-concern estimate depends on contracts and use; the dealer's quote embeds its costs and risk; the auction price reflects the buyer's knowledge that time is scarce.
Which one is relevant depends on the decision. A lender testing collateral coverage may be required by its agreement to use a specified appraisal method, even if management believes the equipment would earn more in place. A potential acquirer may care about service contracts that cannot be transferred without customer consent. A regulator examining a bank that financed similar assets will ask how much cash could be raised under stress, which may differ from the contractual payments due over years. The same asset can support a viable enterprise and still be poor emergency collateral because it cannot be delivered, valued and sold within the necessary window.
A buyer's low bid may combine a real change in use value with a premium for uncertainty and speed. Calling every forced-sale discount permanent impairment goes too far; so does calling every urgent-sale loss temporary when the asset's earning power has fallen. Without evidence about transactions, contracts and alternative users, the analyst cannot apportion the components precisely. A serious judgment can still be made, but it should be framed as a range of feasible recoveries rather than a single supposedly true value.
The bank example raises a related question about securities held at a cost-based accounting value. A fixed-rate bond may repay its face value at maturity if the issuer pays as promised. Its market value can nevertheless fall when rates rise. If the bank must sell it to meet withdrawals, the market value becomes a realized loss. If it can hold the bond and finance deposits at a sustainable cost, the loss may emerge differently through lower future earnings. Neither view makes the funding problem disappear. The crucial phrase is *if it can hold*. A balance sheet can show contractual asset cash flows; it cannot compel depositors to wait for them.
Exposure, trigger and feedback
A rate rise was a trigger for some balance-sheet losses, but the exposure had been assembled earlier through securities purchases, deposit growth, hedging choices and assumptions about customer behavior. The Federal Reserve review's peer comparisons are useful precisely because they identify unusual concentrations without treating a ratio as destiny.[1-4] To move from exposure to failure requires another sequence: rates change, asset values and customer incentives respond, management chooses how to fund withdrawals, information reaches depositors, withdrawals accelerate, and available collateral or sales proceeds fail to meet the cash demand. There are forks at each step. A successful capital raise, slower outflow or earlier liquidity preparation could alter the path. The record does not tell us how large each alteration would have been.
Feedback makes the sequence difficult to reverse. Selling securities reveals a loss and may raise cash, but the loss can also weaken the case for a capital raise. Raising capital can reassure investors if credible buyers appear; announcing the need for capital can alarm depositors when buyers are not yet committed. A deposit transfer protects an individual customer from delay, but many such transfers make the institution's funding problem worse. Calling the result self-fulfilling is only partly right: the institution also had real interest-rate exposure and a funding structure that made the withdrawal option consequential. A feedback loop can amplify an underlying weakness without inventing it.
The same loop can arise outside banks. If a warehouse operator fears that an equipment supplier will fail, it may demand payment before releasing parts. The supplier then has less cash to complete a profitable contract; the customer sees the delay and withholds its milestone payment. The first concern was based on a genuine refinancing problem. The changed terms of trade enlarge it. A firm may break this loop by pledging receivables or arranging a standstill, but doing so requires reliable records and parties willing to accept a common timetable. Where those records are weak, even a workable business can become hard to finance.
What a decision maker needs before Friday
The minimum practical file is more demanding than a list of assets and liabilities. It needs dated cash receipts and payments, the legal triggers in borrowing agreements, the amount and location of eligible collateral, actual settlement times, and a map of decisions requiring external consent. The institution should distinguish a funding line that is committed from one that can be withdrawn; collateral that is owned from collateral already pledged; a forecast receipt from an invoice the customer has accepted. These distinctions are not a guarantee against loss. They tell management how much time it truly has to recognize one.
The file also needs a question about the business after the deadline. Suppose a waiver buys the supplier six months. What is expected to change in six months: customer demand, costs, asset use or the owners' willingness to inject capital? If the answer is simply that the lender may be more patient, the waiver may transfer the loss rather than reduce it. If a signed service contract will begin paying and the current fleet can perform it, the same six months may preserve real value. A crisis response should purchase time for a decision that can improve the position, not merely move the date of recognition.
The hard part is that information is unevenly held. Management knows more about pending contracts than the lender; the lender knows more about its own constraints than management; customers may know whether their projects will proceed but have reasons not to reveal it. The terms each party offers can therefore be evidence as well as a bargaining move. A lender's insistence on more collateral may signal its own funding pressure, the borrower's deteriorating prospects or a policy applied to all customers. A supplier reading that demand as proof of its individual failure could cut a viable operation too quickly. An analyst reading it as proof of predatory behavior could miss a genuine change in risk.
The chapter's opening question—what is due before the asset can earn—should be asked with this uncertainty intact. It leads naturally to the next problem. Even when a firm knows what it owns, the market price on the screen may not be a price at which it can dispose of enough of it by Friday.
The option hidden in an ordinary contract
Many liabilities look stable until the holder has a reason to exercise a withdrawal or termination right. A deposit payable on demand has no fixed maturity for the customer, even if a bank's historical data show that similar accounts usually remain for years. A supplier may enjoy thirty-day payment terms until its counterparty invokes a clause requiring cash on delivery. A lender may tolerate a covenant breach until it calls the loan or refuses renewal. The contractual option existed before the shock; the shock changes the likelihood that many holders use it together.
This is why an average historical holding period can be a dangerous substitute for a legal maturity. Suppose a bank's commercial customers usually leave large balances in operating accounts. The bank invests part of that apparently persistent funding in long-duration securities. The customers then face uncertainty about the bank and can transfer money within hours. The economic maturity of the funding has shortened at precisely the moment when selling the assets is costly. The bank has not necessarily misrecorded the accounts. It has mistaken behavior under ordinary conditions for a binding commitment under stress.
The constructed supplier faces the same option from the other side. Its customer has the right to cancel a future order with a modest fee, while the supplier has already paid for machinery that only this contract can use. The customer's option is a contingent liability for the supplier, although the fee may be the only explicit amount in the legal agreement. A lender looking only at receivables already booked may miss the production capacity stranded if the order disappears. If a waiver is being considered, the practical question is whether the customer can commit to future volumes or whether the supplier must find a different use for the asset.
A contract whose holders cannot leave quickly, or whose exit rights are costly, makes funding more durable. Long-term deposits with genuine penalties, binding purchase commitments and financing with staggered maturities can give an institution time to adjust. They may cost more in normal years. That cost purchases a different crisis option: the ability to wait for assets to pay rather than sell them all on one day. The price of stability is visible before a shock; the value of stability becomes visible when the deadline arrives.
Notes and Sources
[1-1] Basel Committee on Banking Supervision, *Report on the 2023 banking turmoil* (October 2023), pp. 5–7, https://www.bis.org/publications/report-2023-banking-turmoil.pdf. This is the committee's historical stocktake, relying in part on national reviews; it is not a controlled estimate of the effect of any one exposure.
[1-2] Basel Committee, *Report on the 2023 banking turmoil*, pp. 6–8. The report distinguishes observed March 9 withdrawals from management's expectation of March 10 outflows.
[1-3] Basel Committee, *Report on the 2023 banking turmoil*, p. 7. Its judgment is that improved operational readiness was unlikely to have prevented failure but could have helped make resolution more orderly.
[1-4] Board of Governors of the Federal Reserve System, *Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank*, “Evolution of Silicon Valley Bank,” Table 1 (2022:Q4), https://www.federalreserve.gov/publications/2023-April-SVB-Evolution-of-Silicon-Valley-Bank.htm. Peers are the specified weighted average of large US holding companies outside the LISCC portfolio; these are not all banks.
[1-5] Board of Governors, same review, “Federal Reserve Supervision,” overview, https://www.federalreserve.gov/publications/2023-April-SVB-Federal-Reserve-Supervision.htm, and “Preface,” https://www.federalreserve.gov/publications/2023-April-SVB-Preface.htm. The preface sets the March 8 cutoff, compressed timetable and hindsight limitation.
[1-6] Board of Governors, same review, “Supervision of SVBFG by Critical Risk Areas,” interest-rate risk background and management sections, https://www.federalreserve.gov/publications/2023-April-SVB-Supervision-of-SVBFG-by-Critical-Risk-Areas.htm. The two measures and deposit assumptions are described there; this paragraph interprets their different decision horizons.
[1-7] Board of Governors, same section, interest-rate risk modeling and Table 9. The review reports the limits, information flow and November 2022 finding; it does not estimate the outcome under an alternative board or model.