The Chain You Never See Moving Your Money
You wire money from a bank in Port Moresby to a supplier in Rotterdam. The transfer clears. What you never see is the compliance officer in New York who reviewed a risk flag on your originating bank, decided its jurisdiction scored poorly on a de-risking checklist, and quietly began unwinding the relationship that made your wire possible. You won't know until the next one fails and nobody can tell you exactly why.
That invisible middle layer is correspondent banking: the system by which smaller or less-connected banks access payment infrastructure they cannot build themselves. It is the plumbing of international commerce, and its internal architecture is the reason some small economies wake up one morning to find they have lost practical access to US dollar transactions. Not through any sanction. Not through any legal prohibition. Through a commercial decision made thousands of miles away by someone who never heard of them.
A Chain Is Only as Stable as Its Weakest Link's Appetite for Risk
Correspondent banking works through tiered relationships. A major money-center bank, one of the half-dozen institutions that operate true global dollar clearing, sits at the top. Below it are regional correspondents: mid-sized banks in Frankfurt, Singapore, or Johannesburg that maintain accounts at the top tier and, in turn, offer correspondent services to smaller national banks. Those smaller banks serve the actual customers: the importer in Suva, the remittance sender in Honiara, the cocoa exporter in São Tomé.
The chain can run three or four tiers deep. That depth is not incidental. It is the structural fact that determines everything about fragility.
Here is how it plays out. A Pacific island bank, call it Meridian National, holds a correspondent account with a mid-sized Australian bank, which in turn clears dollars through a New York correspondent. Meridian National has roughly 40,000 customers, a clean regulatory record, and transaction volumes that are modest by any global standard. The Australian mid-tier bank runs its annual profitability review. Meridian National generates approximately USD 180,000 in fees, but requires a compliance team to monitor transactions from a jurisdiction flagged as elevated-risk under the Australian bank's internal framework. The numbers don't work, and the relationship is terminated with ninety days' notice.
Meridian National now has no dollar correspondent. It can approach the New York bank directly, but the New York bank's minimum onboarding threshold is calibrated to institutions with balance sheets several times Meridian's size. The door is not technically closed. The economics make it effectively so. Meridian's customers lose dollar access, and the island's import trade, priced overwhelmingly in dollars, becomes significantly more complicated and expensive to finance.
Nothing illegal happened. Nobody was sanctioned. The structure of the chain did it.
The Compliance Arithmetic That Drives Disconnection
To understand why disconnection spreads rather than stabilizing, you have to understand the incentive structure at the correspondent banks doing the cutting.
After the 2012 HSBC settlement and subsequent regulatory actions, global banks faced a new compliance reality: fines for facilitating money laundering or sanctions violations could run into the billions. The rational response was not to improve monitoring of each relationship individually. It was to exit entire categories of relationships where the cost of getting it wrong exceeded the revenue of getting it right. Jurisdictions with weaker anti-money-laundering frameworks, smaller transaction volumes, or complex local ownership structures became candidates for de-risking. The World Bank documented a measurable decline in correspondent relationships in the Caribbean and Pacific, with some national banking systems losing half their active correspondent relationships within five years. Half.
The mechanism is essentially actuarial, and it is worth being direct about what that means: a spotless transaction history is largely irrelevant. A bank maintaining 200 correspondent relationships does not have the bandwidth to perform deep due diligence on each one continuously, so it builds a risk-scoring model. A Financial Action Task Force grey-list designation scores poorly. Banks in those jurisdictions get flagged. When the annual review comes, flagged relationships with sub-threshold revenue get cut. The individual bank being cut may have never processed a suspicious transaction in its life. It doesn't matter: the category triggered the score, not the record.
This is the arithmetic that small economies cannot easily escape. Their banks are small by definition. Their transaction volumes will rarely clear profitability thresholds designed for institutions ten times their size. And their jurisdictions often have regulatory capacity that lags FATF standards, not because of bad faith, but because building a sophisticated financial intelligence unit requires resources that are simply scarce in a country of 100,000 people.
Why Depth in the Chain Is Not a Safety Net
A common intuition holds that a longer chain, with more intermediate correspondents, should provide redundancy: if one link breaks, another exists. The opposite is closer to the truth, and the confusion here has cost policymakers years.
Every additional tier creates an additional point of discretionary commercial review. It also creates what practitioners call nested correspondent banking, where the top-tier institution may not even know which ultimate customers are transacting through the chain. A New York bank clears for a German bank that clears for a Kenyan bank that holds accounts for Somali money-service businesses. The New York bank has no direct relationship with the Somali end and limited visibility into those transactions. That opacity, not any actual wrongdoing, is itself a compliance risk that drives de-risking at the top. The chain looks, from a distance, like a web of redundancy. It functions, under pressure, more like a row of dominoes.
Depth makes the system look more connected on paper while making it more fragile in practice. The more tiers between a small economy's bank and the dollar clearing system, the more commercial reviews that relationship must survive each year, and the more exposure it has to any one reviewer deciding the numbers don't justify the risk.
One structural feature amplifies all of this: correspondent banking is not a regulated utility with access obligations. It is a set of private commercial contracts, terminable at will. The entire edifice of global dollar access for smaller economies rests on relationships that exist only as long as they remain profitable to the party holding the stronger commercial position.
The Geography of Who Gets Cut
Look at which economies have experienced the sharpest correspondent banking attrition and the pattern is not random. Small island states in the Pacific and Caribbean, economies in sub-Saharan Africa dependent on remittance flows, and post-conflict states attempting to rebuild financial systems all appear disproportionately. What they share: high perceived jurisdictional risk relative to transaction volume, limited ability to lobby for better risk scores in foreign compliance frameworks, and banking sectors too small to justify the fixed costs of direct relationships with top-tier institutions.
Ask yourself: what exactly is a small, heavily dollarized economy supposed to do when the correspondent chain snaps? There is no local currency to fall back on when you have surrendered monetary sovereignty. Dollar access is not a convenience in those economies. It is the operating system.
When correspondent relationships thin out, the first effect is typically a spike in remittance costs, as fewer formal channels compete for the business and informal channels fill the gap. Informal channels are, predictably, harder to monitor, which in turn validates the compliance concerns that drove the de-risking in the first place. The loop closes on itself. The jurisdictions that most need the system's confidence end up generating the evidence that justifies cutting them.
What Actually Breaks the Loop
Three approaches have shown partial traction, none of them complete.
Regional correspondent hubs attempt to pool the compliance infrastructure of several small economies behind a single, better-capitalized institution that can meet top-tier onboarding requirements. The Eastern Caribbean Central Bank has experimented with variations on this model. It reduces the number of individual relationships that need to survive annual reviews, but it concentrates risk in a way that should give anyone pause: if the hub bank loses its own correspondent relationship, the entire region goes down together. Trading distributed fragility for concentrated fragility is not obviously progress.
Standardized know-your-customer utilities, where banks share due diligence data on correspondent relationships through a common platform, reduce the per-relationship compliance cost for top-tier institutions. Lower compliance costs shift the revenue threshold for keeping a relationship profitable, which in theory makes small-economy banks viable again. The SWIFT KYC Registry is the most prominent example. In practice, adoption has been uneven and coverage of smaller jurisdictions remains incomplete.
The third approach is regulatory: the Basel Committee and the Financial Stability Board have both issued guidance urging banks not to engage in wholesale de-risking, arguing that blanket termination of relationship categories is itself a failure of risk management, not an expression of it. Guidance without enforcement teeth, though, changes commercial behavior slowly, partially, and unevenly. That is not an opinion. That is the documented record.
The structural problem persists because it is not, at its root, a compliance problem or a technology problem. It is a scale-mismatch problem inside a private system with no access obligations, and those are the hardest problems to fix because fixing them requires someone with power to accept costs they are not currently required to bear. Until the commercial arithmetic changes, whether through pooled infrastructure, binding regulation, or some mechanism not yet tried, the chain will keep finding its weakest links. It will keep letting them go quietly, and the economies on the other end will keep absorbing the cost of a system that was never designed with them in mind.