When the Wire Goes Dead
The letter arrives on a Tuesday. Ninety days' notice, three paragraphs of legal boilerplate, and a polite sign-off from a compliance officer whose name means nothing to the civil servant holding the page in a finance ministry somewhere in the eastern Caribbean. No drama, no negotiation. Just the quiet administrative fact that the local commercial bank's account at a major New York correspondent has been closed, and with it the country's ability to clear US dollar transactions. Fuel arrives in dollars. Medicines are invoiced in dollars. The flour that keeps the bakeries running is priced in dollars. Now go explain that to the minister.
This is not a hypothetical scenario dressed up for illustration. It is the operational reality that dozens of small economies have faced, and understanding why it keeps happening requires looking at the internal architecture of correspondent banking itself, not at the familiar headlines about sanctions or money laundering that usually accompany the story.
The Plumbing Nobody Sees
Correspondent banking is the system by which banks in different countries move money for each other. A small bank in Port Moresby or Bridgetown doesn't hold an account directly at the US Federal Reserve. To send or receive dollars, it needs a relationship with a bank that does. That larger bank, typically sitting in New York, London, or Frankfurt, is the correspondent. The smaller bank is the respondent. The respondent holds a nostro account (literally, "our account") at the correspondent, and the correspondent processes dollar payments on the respondent's behalf, charging a fee for the service.
Simple enough in outline. Ruinous in practice. Every transaction the correspondent processes on behalf of its respondent exposes it to regulatory scrutiny. US law, specifically the Bank Secrecy Act and its associated rules, holds the correspondent liable if it processes transactions connected to money laundering, sanctions violations, or terrorist financing. The correspondent cannot fully audit every customer of every respondent bank. It is, in effect, being asked to vouch for the compliance standards of institutions it doesn't directly control, in jurisdictions it may barely know.
The fee structure makes this worse. A mid-sized Caribbean bank might generate a few hundred thousand dollars a year in correspondent fees. A large US correspondent bank might spend several million dollars in compliance staff time monitoring that relationship. The math, when a compliance officer runs it, produces a number that ends the conversation.
How the Risk Calculus Actually Works
Here is where the internal structure of the relationship starts determining outcomes.
A major US correspondent bank maintains what is effectively a portfolio of respondent relationships, each assigned a risk score built from several inputs: the jurisdiction's rating on Financial Action Task Force (FATF) grey and black lists; the respondent bank's own compliance infrastructure; the mix of transaction types flowing through the account; and the volume-to-scrutiny ratio. A bank in a FATF-listed jurisdiction processing a high volume of cash-intensive transactions from politically exposed persons generates a risk score that sits uncomfortably high on the portfolio.
The correspondent's compliance team then does something that looks like portfolio management. It ranks respondent relationships by expected revenue divided by expected compliance cost. Relationships below a certain threshold get flagged for review. Below a lower threshold, they get terminated. The process is not punitive in intent. It is actuarial. The consequences, though, are punitive in effect, and that distinction matters more than most policy discussions acknowledge.
Consider two banks that opened correspondent accounts with the same New York institution in the same year. One is a mid-sized bank in a FATF-grey-listed Pacific island nation, processing remittances and government payments, with a compliance team of two people and no dedicated transaction-monitoring software. The other is a regional bank in a larger economy with a robust compliance department, automated screening against sanctions lists, and a diverse transaction mix that includes trade finance for exportable goods. When the correspondent's annual review runs, the first bank fails the ratio test. The second doesn't. They started in the same place. The internal scoring mechanism sent them in opposite directions.
If the grey-listing were the whole story, two banks in the same jurisdiction would rarely get different outcomes, yet they do, routinely. Jurisdiction matters enormously, but compliance investment at the respondent level shapes the result just as much. That asymmetry is almost never discussed in the policy literature, which tends to treat de-risking as something that happens to countries rather than to specific institutions within them. The omission is not innocent; it lets governments off the hook for a problem they could partly address.
The Nesting Problem That Multiplies the Risk
Correspondent relationships are not always direct. A small bank in a very small economy may not hold a direct relationship with a New York correspondent at all. It may have a relationship with a regional bank, which itself has a correspondent relationship with the New York institution. This is nested correspondent banking, and it is the point at which the risk calculus becomes genuinely toxic.
The New York correspondent, looking at the regional bank's account, sees a blended flow of transactions. Some originate from the regional bank's own customers. Some originate from the smaller, nested respondent banks the regional bank services. The New York correspondent often cannot see through that second layer to identify the ultimate originator of a given transaction. From its compliance perspective, the regional bank is not just one respondent. It is an opaque aggregator of unknown respondents, each carrying their own risk profile.
This is the structural trap. The smaller the economy, the more likely its banks are to be nested rather than directly connected. The more nested they are, the less transparent they appear to the top-tier correspondent. The less transparent they appear, the higher the risk score assigned to the entire chain. The FATF has specifically identified nested relationships as a red flag, which means the architecture that small economies are forced into by their size is itself the feature that makes them look dangerous.
Think of it like a relay race where the last runner is invisible. The baton is moving; you simply cannot see who is carrying it. The correspondent bank, facing a regulator who will not accept "I couldn't see" as a defence, decides the safer move is to drop the baton entirely.
What People Get Wrong About De-Risking
The standard narrative frames correspondent banking withdrawal as a response to bad actors in small economies. Regulators spot money laundering; banks pull out. That framing is not exactly wrong, but it obscures the mechanism in ways that lead to bad policy responses, and the bad policy responses are by now a decades-long pattern.
The decision to exit a respondent relationship is almost never triggered by a specific bad transaction. It is triggered by a compliance team's forward-looking estimate of regulatory exposure. A bank can lose its correspondent account without a single suspicious transaction ever having been flagged, simply because the expected cost of monitoring the relationship indefinitely exceeds the expected revenue. The risk is probabilistic and prospective, not retrospective.
This distinction matters because it means that cleaning up actual bad actors in a small economy doesn't necessarily restore correspondent access. If the compliance math still doesn't work, the account stays closed. Several Caribbean jurisdictions passed significant anti-money-laundering reform packages, got removed from grey lists, and still found major correspondents unwilling to restore terminated relationships. The reason is straightforward: once a correspondent has restructured its portfolio to exclude a relationship, re-onboarding that respondent represents a fresh cost. The incentive to re-enter is low unless the revenue justification is compelling, and for a small economy, it rarely is.
The policy lever that actually moves correspondents is not cleaner regulation in the respondent jurisdiction. It is the removal of personal liability exposure for compliance officers at the correspondent bank. When regulators offer safe harbour provisions or clearer guidance on what counts as adequate due diligence, correspondents re-engage. When they don't, the math stays broken. Observers who continue to prescribe AML reform as the primary remedy are, at this point, ignoring a substantial body of evidence to the contrary.
The Consequence That Compounds
Losing correspondent access doesn't just make international payments slower or more expensive. It creates a structural dependency on whatever access remains, which is almost always both more expensive and less reliable. A bank that loses its direct New York correspondent may be forced to route transactions through a Caribbean regional hub, which itself has limited correspondent access and charges accordingly. A wire transfer that once took two days and cost twenty dollars might now take five days and cost eighty, routed through two intermediaries, each of which takes a compliance cut.
For ordinary people, this shows up in remittances. The IMF and World Bank have both documented cases where de-risking pushed remittance costs in affected corridors above ten percent of the transferred amount, against a global average that development organisations have spent years trying to push below three percent. The person sending two hundred dollars home from London to Samoa is now sending significantly less, not because of exchange rates, but because the plumbing corroded.
For governments, it shows up in the cost of sovereign debt issuance, the ability to settle import contracts on time, and the credibility of the central bank's foreign currency reserves as a usable buffer. A reserve that cannot be deployed quickly because the settlement mechanism is broken is not really a reserve at all. It is a number on a spreadsheet.
The internal structure of correspondent banking, its risk-scoring portfolios, its compliance-cost arithmetic, its opacity to nested relationships, did not produce these outcomes by accident or malice. It produced them by logic. Small economies are expensive to monitor, cheap to service, and easy to drop. Until the cost of dropping them is made to exceed the cost of keeping them, the wire will keep going dead, and the civil servant will keep writing the memo to the minister.