The Moment Payment Becomes Conditional
Foreign-exchange settlement risk becomes clear when we follow the two payments that complete a trade rather than looking only at the exchange rate. The focus here is the condition that turns those payments from two separate promises into one protected exchange: payment-versus-payment, or PvP. Understanding that mechanism lets us trace why principal exposure disappears in the protected path, while replacement-cost and liquidity risks remain around it.
The danger lives between two final payments
Imagine that Party A sells currency A and buys currency B from Party B. The trade is agreed, but the economic exchange is not complete until both currencies have been delivered. If A’s payment becomes final before B’s payment, A has performed its side while waiting for the other side to perform. That waiting interval is the source of settlement risk.
“Final” matters here. A final payment is not merely an instruction sent into a banking system. It is a transfer that can no longer be cancelled or reclaimed through the ordinary payment process. Once currency A has been delivered with finality, Party A may not be able to reverse that delivery simply because currency B does not arrive. If Party B defaults in the interval, A’s exposure is measured against the full value already paid.
The sequence can be written as four events:
- The counterparties agree an exchange of two currencies.
- One currency is delivered and becomes final.
- The other currency is delayed, interrupted or not delivered because the counterparty fails.
- The paying party must absorb the loss or find replacement liquidity while still missing the currency it expected.
The exchange rate is not the problem in this sequence. The problem is that the two legs have different finality. The currencies may travel through different domestic payment systems, correspondents and time zones. Even if the underlying trade is straightforward, the infrastructure that completes each leg may not be synchronized.
Consider a hypothetical trade in which Party A sends $100 million and is due to receive the equivalent amount in another currency. If the dollar payment becomes final and Party B defaults before delivering the other currency, Party A has not merely missed a scheduled receipt. It has already surrendered $100 million of principal. It may also need to find the missing currency elsewhere, perhaps when market conditions have changed. The example is not a forecast of a likely loss; it isolates why the risk measure follows the amount delivered rather than the size of an administrative delay.
In ordinary conditions, counterparties may make these payments successfully for years. That history can make a bilateral arrangement feel safe, but it does not change the sequence that would apply on the day one participant fails. A settlement design is most valuable when it changes the payment condition before the failure occurs, rather than relying on participants to decide correctly after one side has already paid.
Bankhaus Herstatt made this timing problem concrete on 26 June 1974. German authorities ordered the bank into liquidation and closed it at 3:30 p.m. Frankfurt time. Earlier in the day, counterparties had already delivered Deutsche marks to settle dollar–mark trades. After the closure, Herstatt’s New York correspondent suspended outgoing dollar payments. The counterparties had therefore paid the currency they sold but did not receive the currency they purchased.
The important detail is not simply that a bank failed. Banks fail in many financial arrangements, and other mechanisms are designed to distribute or manage those failures. Herstatt exposed a failure of sequencing: a payment that was final in one location became irrevocable before the corresponding payment in another location was made. The first side could lose the principal, not merely the expected interest or the cost of correcting a message.
The consequences spread beyond the counterparties directly involved. Market participants began delaying payments until they had confirmation that their counterparties had paid. That defensive response made the network less liquid precisely when it needed to keep moving. Cross-border payments and FX trading slowed sharply, threatening a wider loss of confidence. The episode subsided only after major central banks signalled their readiness to provide sufficient liquidity.
This is why settlement risk has a financial-stability dimension. A single failure does not automatically become a global crisis. The relevant danger is the feedback loop: one unpaid leg makes another participant reluctant to pay, that delay leaves a third participant short of expected funds, and uncertainty encourages more withholding. The mechanism can turn a bilateral default into payment gridlock.
The distinction between a promise and a final transfer also explains why ordinary counterparty confidence is not a complete substitute for infrastructure. A participant can believe that a trading partner is sound and still be exposed to the order in which payment systems operate. Trust may support a market in normal times; it cannot make a completed payment conditional after the fact. The system-level question is whether the settlement process itself limits the amount that any one participant must place at risk while waiting.
PvP changes the condition, not just the probability
PvP addresses the problem by changing when a final payment is allowed to occur. A PvP system ensures that the final payment of one currency occurs if and only if the final payment of the other currency occurs. Neither side is asked to rely on the other side’s promise after surrendering its own principal. The two payment legs are made conditional parts of a single settlement event.
That condition removes the central exposure in the Herstatt sequence. If Party A’s currency cannot be released unless Party B’s currency is also being released, Party A does not complete only half of the exchange. If Party B fails before the condition is met, the protected settlement does not complete in the ordinary way; Party A may face delay or another consequence, but the original currency payment is not left final and unpaired. PvP therefore eliminates FX settlement risk for the settlement leg.
The word “conditional” is doing the important work. PvP does not wait until after an unpaired payment to repair the damage. It places the condition before final release. In a bilateral arrangement, Party A may have to decide whether to send currency A based on Party B’s reputation, the timing of confirmations and the cost of being cautious. In a PvP arrangement, the payment rule makes that decision part of the shared process. The protection is structural rather than a matter of one trader making a more prudent choice at the last moment.
The diagram also shows what PvP does not require. It does not require a central counterparty to become the buyer and seller for every participant. A central counterparty, or CCP, interposes itself between participants and assumes their obligations. PvP performs a different function: it synchronizes the final payment of two currencies. Some systems may combine the functions, but the payment-versus-payment condition should not be confused with the obligation-assumption function of a CCP.
The difference becomes clearest when a participant fails. A PvP system’s first protection is to stop the corresponding payment from becoming final; it does not, by that fact alone, promise to complete the failed party’s obligation or to find a replacement trade. A CCP has a different responsibility because it has inserted itself into the contractual chain. Treating the two functions as interchangeable would make the protection sound broader than it is. The question for a settlement system is first whether the two currency legs are released together; the question for a CCP is how obligations are managed when a participant cannot perform.
That distinction matters for judging what has actually been protected. PvP removes the principal exposure created by paying one currency without receiving the other. It does not guarantee that a participant will never lose money on the trade. If the original transaction has to be replaced after a failure or cancellation, the current market price may be worse. That is replacement-cost risk: the cost of recreating the position is separate from whether the original settlement legs were exchanged safely.
PvP also does not guarantee that an expected currency will be available at the exact moment a firm needs it. If the protected process does not complete, the participant may still face a liquidity problem. A bank may have scheduled another payment, funded a customer position or planned a securities transaction around the expected receipt. Preventing an unpaired final payment avoids principal loss, but it can leave a timing gap in the institution’s cash position.
This is a useful test for claims about settlement safety. Ask first whether the claim concerns the principal exchanged in the settlement leg, or the broader consequences of a delayed or cancelled trade. A system can succeed at preventing an irreversible one-sided transfer while still requiring the participant to manage a funding shortfall. Conversely, a participant may replace a failed trade at a cost even though no currency was released unpaired. These outcomes are related in practice, but they are not the same measurement.
The difference can be expressed simply:
- Settlement risk: one currency is delivered finally and the other is not. PvP removes this exposure for the protected settlement leg.
- Replacement-cost risk: the intended trade must be recreated at a worse price. PvP alone does not remove it.
- Liquidity risk: a currency expected at a particular time does not arrive, even if no unpaired payment is made. PvP alone does not remove it.
This separation prevents a common overstatement. “PvP eliminates FX settlement risk” does not mean “PvP makes the trade risk-free.” It means that the mechanism changes the condition for final delivery and prevents the specific principal-risk sequence that made Herstatt so damaging.
The conditional rule can also help the market coordinate during stress. In a decentralised market, each participant normally has to decide whether to release an outgoing payment while waiting for an incoming one. Under PvP, the system provides a shared answer: the outgoing leg will not be processed unless the incoming leg is also ready for final payment. That reduces the incentive to withhold payments defensively and can help prevent a chain of precautionary delays from becoming gridlock.
This coordination benefit is easy to miss if PvP is described only as a technical switch. In a network of obligations, one firm’s decision to wait can leave another firm short of the currency it needs to make its own payment. That second firm may then wait as well, even if neither firm wants to stop trading. PvP does not make the counterparties solvent, and it does not guarantee that every instruction will settle on schedule. It does, however, give the connected payment legs a common rule for when finality is allowed. In that narrow but important sense, it replaces a chain of bilateral judgments with a shared settlement condition.
PvP systems can often enable netting of payments as well. If several obligations can be offset before the final exchange, the participants may need to move less liquidity through the settlement process. But this benefit should not blur two separate ideas. Conditional payment protects the relationship between the two currency legs; netting reduces the amount that must move. A smaller payment can still need a settlement method, and a netted amount is not automatically protected by PvP.
For example, if two counterparties owe each other several payments in the same currency on the same value date, calculating a single net amount can reduce the cash that must cross their accounts. That reduces the size of the exposure that remains to be settled. It does not, by itself, ensure that the net amount will be exchanged for the other currency at the same moment. The risk question therefore has two parts: how much must move, and what condition governs the final movement. PvP answers the second question; netting primarily changes the first.
CLSSettlement is the largest and best-known PvP system. As of June 2026, it settles payments involving 18 major currencies. Its importance illustrates both the power and the boundary of a shared system. Where the relevant trade, currencies and counterparties can enter the system, the conditional rule changes the principal-risk calculation. Where they cannot, a bank may need to rely on a weaker method or accept bilateral exposure.
That boundary is the reason PvP is a benchmark rather than a complete answer to the FX market’s settlement problem. The strongest mechanism can eliminate the targeted risk for the transactions it covers, but it cannot protect a trade that the participant, currency pair, product or operating process cannot bring into the conditional path. The next task is to compare the methods used outside PvP and see what each one preserves, reduces or leaves exposed.