The Bank That Closed at 3:30 and Left the Dollars Missing: Herstatt’s Hidden Payment Risk
A payment can be final while the deal is not
On 26 June 1974, a German bank was closed in the middle of an ordinary working day. Bankhaus Herstatt’s counterparties had already delivered Deutsche marks in Frankfurt to complete foreign-exchange transactions. They expected to receive US dollars through New York later that day. Instead, German authorities ordered the bank into liquidation at about 3:30 p.m. Frankfurt time, and its New York correspondent suspended outgoing dollar payments. Some firms had performed their side of a trade but were left waiting for the other side.
This was not the familiar risk that a currency’s price might change before a trader could sell it. The agreed exchange rate could have been perfectly fair. The danger was in the mechanics of delivery: two payments that appeared to be one transaction were actually moving through different banking systems, at different times. The gap between them created an opportunity for a counterparty failure to become a loss of principal.
The episode gave the financial world an enduring phrase, ‘Herstatt risk’. It also helps explain why modern payment infrastructure is not a boring technical detail beneath the real economy. The moment a business sends money across currencies, the question is not only how much it will receive. It is whether the arrangement makes the two sides of the exchange depend on one another.
The clock that split a currency trade in two
A foreign-exchange transaction usually has at least two obligations. One party delivers currency A; the other delivers currency B. An agreement can specify the amount and settlement date, but that does not mean the payments travel together. Historically, correspondent banks and national payment systems processed the two currencies separately. Their business hours did not necessarily overlap. An instruction might become final in one country before the receiving payment could be made in another.
Herstatt was active in foreign-exchange markets during a volatile period after the collapse of the Bretton Woods fixed-exchange-rate system. A Basel Committee study of bank failures describes how the bank’s foreign-exchange losses and inadequate controls contributed to its collapse. That is a separate question from the problem its counterparties faced when payments stopped. A bank can fail because of bad positions, while its failure creates settlement losses for people whose trades were otherwise sound.
The Bank for International Settlements’ June 2026 analysis reconstructs the decisive sequence: the German liquidation order on 26 June 1974, the bank’s 3:30 p.m. Frankfurt closure, and the suspension of dollar payments in New York. The practical asymmetry is the entire story. The Deutsche mark leg had gone out. The dollar leg had not come back. Even a trader who had correctly predicted the exchange rate could be caught.
A worked example: the missing half of a ₹2.125 crore exchange
Consider a hypothetical Indian importer or treasury desk exchanging rupees for dollars through a bilateral arrangement. This is an illustrative calculation, not an account of a real Herstatt transaction. Assume the firm agrees to buy US$250,000 at a contractual rate of ₹85 per dollar. Its rupee obligation is 250,000 × 85 = ₹21,250,000, or ₹2.125 crore. The expected receipt is US$250,000. The agreed rate determines the two amounts; it does not by itself guarantee simultaneous delivery.
Suppose the firm pays the ₹21.25 million irrevocably at 10:00 and the dollar leg is scheduled for later. The intermediary responsible for the dollars then fails before making payment. At that instant, the firm has a gross principal exposure equal to the amount already delivered. The eventual economic loss is not necessarily the full ₹21.25 million: insolvency recoveries, guarantees, collateral and the legal structure may change the outcome. But the firm is exposed to much more than a small exchange-rate movement.
Now separate that problem from market risk. If the firm has to arrange a replacement purchase at ₹88 per dollar, buying the same US$250,000 costs ₹22,000,000. The difference from the original contract is ₹750,000, or 3.53% of the original rupee amount. That is a replacement-cost illustration. It should not simply be added to an assumed total principal loss without analysing recovery and whether the firm is counting the same cash flow twice. Settlement risk, replacement cost and liquidity pressure are related, but they are not synonyms.
The numbers demonstrate why a cheap quoted exchange rate can be an incomplete bargain. Saving a few basis points on the price is not necessarily attractive if the settlement process exposes the full principal to an unreliable counterparty. For a business with tight working capital, even a temporary delay in receiving the dollars can disrupt a supplier payment long before a final legal loss is known.
The damage spread through confidence
The immediate Herstatt failure mattered because other banks could not know which counterparties might be unable to complete payments. According to BIS histories, institutions began withholding or delaying outgoing transfers until they could confirm that incoming funds had arrived. A defensive action that looks sensible for one bank can become dangerous when many banks do it together. The network experiences a shortage of trust and, potentially, of liquidity.
This is a financial version of a traffic jam caused by every driver waiting for proof that the next junction is clear. No single bank has to intend a systemic crisis. Each institution’s attempt to avoid becoming the next unpaid creditor can prevent other institutions from receiving the funds they need. The result is payment gridlock. Herstatt helped central bankers recognize that the settlement machinery itself could transmit distress.
The aftermath contributed to international cooperation. The Basel Committee on Banking Supervision traces its creation at the end of 1974 to serious disturbances in currency and banking markets, notably Herstatt’s failure. The committee did not spring from this one episode alone; the period included other major banking failures and wider financial changes. Nevertheless, Herstatt became a defining example of why cross-border supervision and payment-system design needed coordinated attention.
The elegant rule: both currencies move, or neither does
The central engineering response is payment versus payment, usually abbreviated PvP. Under a true PvP mechanism, the final transfer of one currency occurs if and only if the final transfer of the other currency occurs. If the dollars cannot be delivered, the linked rupee or euro payment does not settle irrevocably as an isolated half of the trade. The design attacks the specific gap that trapped Herstatt’s counterparties.
The distinction from merely scheduling payments for the same hour is important. Two independent payments can be timed to coincide and still fail separately. PvP links their finality. A familiar analogy is a transaction in which neither party releases its item until the exchange can complete, but real PvP systems achieve that condition through formal rules, accounts and operational infrastructure rather than trust in an informal promise.
The BIS notes that CLSSettlement, launched in 2002, provides PvP for eligible foreign-exchange trades across 18 major currencies. PvP addresses the principal risk of paying one currency without receiving the other. It does not eliminate every hazard: a party may still face replacement costs if a trade fails, liquidity needs, operational outages, legal complexities or market losses. No payment mechanism turns a bad economic decision into a good one.
India’s less obvious part in the story
It would be inaccurate to suggest that an Indian rupee transaction must use CLS to be protected. The Indian rupee is not one of CLSSettlement’s 18 eligible currencies, but India has its own relevant infrastructure. The Clearing Corporation of India Limited (CCIL) states that it began settlement of interbank USD–INR transactions in November 2002 and moved the segment to payment-versus-payment settlement in April 2015. CCIL’s published procedure explains that the rupee leg settles through members’ accounts with the Reserve Bank of India, while the dollar leg uses CCIL’s settlement-bank account.
The detail is important because a global list of CLS currencies is not a complete map of the world’s PvP protection. The BIS’s June 2026 review explicitly identifies CCIL in India as one of the other PvP arrangements for currency pairs not eligible for CLS. That is a concrete example of local financial infrastructure solving a global problem in a jurisdiction-specific way.
The protection also has a boundary. CCIL’s described arrangements apply to eligible transactions between participating institutions under the relevant rules. An individual sending a small international remittance, a freelancer waiting for an overseas invoice or an exporter receiving money through a commercial bank is not automatically using the same interbank settlement mechanism. Their actual risk depends on the service provider, the contracts, the payment chain and when each obligation becomes final.
The 2026 data: progress, but not disappearance
For years, public discussions repeated the alarming finding that large volumes of foreign-exchange trades remained exposed to settlement risk. The BIS estimated in its 2022 analysis that about US$2.2 trillion of deliverable FX turnover on a typical day in April 2022 was settled without adequate risk mitigation, roughly 31% of the relevant turnover under that study’s classification. The scale was a reminder that financial technology had not solved every payment gap.
A June 2026 BIS study using the 2025 Triennial Survey introduced a more granular measurement of actual settlement methods. It reported that approximately 36% of average daily FX settlement volume in April 2025 used PvP, which eliminates FX principal settlement risk. A further 54% used arrangements such as intragroup settlement, netting or controlled timing that mitigate but do not necessarily eliminate it. About 10%, or US$1.4 trillion of daily gross settlement, remained on a gross bilateral basis fully exposed to settlement risk.
Those figures cannot simply be read as a fall from 31% to 10% under an identical measurement. The BIS explicitly explains that the 2025 framework categorizes settlement methods differently and measures gross settlement obligations. Netting, intragroup payments and controlled settlement are distinguished from fully unprotected gross bilateral flows. The correct lesson is that much risk has been reduced, but significant exposures remain and the quality of the mitigation varies.
How to read the 2025 distribution: PvP 36% means the two currency legs are linked for finality; mitigated methods 54% means risk is reduced or managed but not uniformly eliminated; gross bilateral 10% means the full principal may be exposed. These categories describe settlement methods, not probabilities of failure. A 10% share does not mean a 10% chance that a payment will be lost.
Why netting helps without performing magic
Another common risk-reduction technique is netting. Suppose two institutions have several offsetting obligations with the same value date and compatible legal terms. Instead of sending every gross payment separately, they may calculate a smaller net amount. If one party owes US$100,000 and is owed US$60,000 in compatible obligations, the net amount payable may be US$40,000. That reduces the cash that has to move by US$60,000, or 60% of the original US$100,000 outgoing obligation.
But netting is not identical to PvP. The remaining US$40,000 still has to be settled, and the protection depends on enforceable agreements, currency legs, counterparty structure and finality. In a default, the legal enforceability of netting can be as important as the arithmetic. The BIS warns against assuming that every netted trade is fully protected simply because the amount being transferred is smaller.
This illustrates a broader financial principle: reducing exposure is not the same as removing the failure mode. A business might accept residual risk when it is small and well controlled. It should not confuse that judgment with a guarantee that payment will arrive.
What a treasury team should ask before the trade
For a company buying or selling foreign currency, the useful questions go beyond the exchange-rate quote. Which legal entity is the counterparty? Who actually receives and sends each currency? Are both legs linked through PvP, or does one become final first? What are the settlement date, cutoff times and time zones? Is there a third-party bank or correspondent in the chain? What happens if a payment is delayed, rejected or reversed?
A prudent treasury process also establishes exposure limits, verifies payment instructions independently, and keeps a liquidity buffer for a delayed receipt. Larger institutions may use documented counterparty limits, approved settlement arrangements, collateral, netting agreements and reconciliation controls. Small businesses may not control the interbank infrastructure, but they can ask their provider about timing, finality, fees, error handling and available safeguards.
It is useful to separate four risks on a one-page worksheet: market risk (the rate moves), principal settlement risk (one currency is delivered and the other is not), replacement-cost risk (a failed trade must be replaced at a worse price), and liquidity risk (cash arrives too late to meet an obligation). A product description that says ‘secure international payments’ does not answer all four questions.
The limit of the Herstatt analogy
Not every cross-border payment resembles a 1974 interbank FX deal. Retail remittances, card transactions, trade finance and correspondent transfers involve different contracts, protections and finality rules. A payment that is delayed by compliance screening is not necessarily a settlement default. A bank’s published exchange-rate spread is not evidence that its FX settlement is unsafe. The point of Herstatt is to identify a specific structural vulnerability, not to treat all international transfers as suspect.
The history is also a warning against dramatic but imprecise statistics. The BIS’s 2026 survey gives a better classification of settlement methods than earlier headline comparisons. It does not imply that US$1.4 trillion is lost each day. That is an amount exposed to a failure mode, not an estimate of actual daily losses. Confusing exposure with loss would turn a serious risk into sensational misinformation.
For Indian readers, the CCIL example is encouraging without being a blanket guarantee. A carefully designed local PvP system can remove a particular principal risk for eligible participants, but users still need to understand their own payment arrangements. The right question is never simply whether a currency appears on a famous international list. It is how the specific transaction settles.
The transaction was finished on paper, not in the banking system
Herstatt’s collapse revealed that an exchange-rate agreement and a completed exchange are two different things. The parties could agree on the price, record the obligation and even pay one side in full while the other side remained vulnerable. That is a powerful lesson for every financial system built from separate institutions: the handoff can be the most dangerous part.
The solution that emerged is both technical and ethical. Technical, because payment rules and settlement systems can link the finality of two currencies. Ethical, because a firm offering a cross-border transaction should not treat the customer’s exposure as invisible merely because it sits in back-office operations. The bank that closed at 3:30 did not invent counterparty risk. It made a hidden version of it impossible to ignore.
Sources, calculations and photograph
Primary sources: BIS Quarterly Review (June 2026), 2025 FX settlement survey and Herstatt reconstruction; BIS Quarterly Review (December 2022), earlier FX settlement risk methodology; Basel Committee on Banking Supervision, official history; Basel Committee, Bank Failures in Mature Economies (2004), Herstatt case; CCIL, USD–INR clearing and PvP settlement procedure; CCIL, USD–INR segment history and April 2015 PvP introduction; Basel supervisory guidance on FX settlement risk. All worked rupee figures above are hypothetical, not historical exchange rates; arithmetic is 250,000 × 85 = 21,250,000 and 250,000 × (88 − 85) = 750,000.
Featured photograph: the former Herstatt Bank buildings in Cologne, photographed by Georg Schelbert on 8 September 2012, Wikimedia Commons, CC BY-SA 3.0. It is a later architectural photograph of the bank’s former premises, not an image of the June 1974 closure.