The question we were asked
A contact from the quantitative investment world put the following question to us:
“On the subject of trade finance, a question came to me from my quant world: in your view, are the risk and return models already very mature? In the quant space, there is an ongoing search for better data, factors and models to consistently deliver strong performance and remain competitive. How does that compare in this asset class?”
The comparison with the quant world is the right way to test the asset class. It separates two very different sources of return: returns from a better signal-driven investment selection and returns from better management and control of a real transaction.
Two different sources of return
In quantitative strategy, the edge often comes from identifying better data, factors or signals before the market prices them in. Once a signal becomes widely known, it can decay. In trade finance, the unit of risk is different. We are not underwriting a price movement on a screen; we are underwriting a transaction chain: purchase order, shipment, invoice, receivable, buyer obligation, insurance and collection.
The better question is not only “what is the probability of default?” but also “what exactly are we underwriting, and is the path back to cash visible, verifiable and legally enforceable?”
Our short answer is that the basic risk framework in trade finance is mature, but real differentiation does not come from a generic default model alone. It comes from transaction selection, documentation, disciplined credit insurance, legal structure, operational control, and portfolio construction.
Put simply, we believe trade finance is attractive not because risk disappears, but because many of the key risks can be identified, structured and controlled before capital is deployed.
The low-default record is real — but this is partly a result of selection
The ICC Trade Register is a useful starting point, as it has consistently shown that trade, supply chain, and export finance are low risk overall. Its dataset is large — the 2024 edition, compiled from 22 contributing banks and financial institutions, is described by the ICC as representing close to a quarter of global trade-finance transactions. That is important evidence. But it should be read carefully, and this is the part most summaries skip.
The data is contributed by roughly two dozen predominantly large banks and is weighted towards established counterparties and controlled trade products. The register’s own detail shows where this matters: Export-finance transactions in the sample are guaranteed by export credit agencies to an average of around 94% of their value — so the headline loss rates partly reflect government guarantees, not raw credit performance.
On a transaction-weighted basis, defaults on import letters of credit and import/export loans have been rising, with the increase attributed to mid-size corporate obligors. In other words: the asset class looks well-behaved in aggregate because the largest, best-collateralised flows dominate the data — though the risk rises exactly where the smaller, less-standard business sits.
This is better described as selection, bankbook, or survivorship bias, and a quant will recognise it as the latter, where the data reflects the transactions that banks choose to originate, retain, and report.
The practical conclusion for a private trade-finance manager is blunt: we should not simply borrow the register’s low-default profile. We must create our own, transaction by transaction, through selection, documentation, legal structure, insurance discipline, and portfolio control.
Why the opportunity exists
Trade finance is not a niche corner of finance. The WTO estimates that 80–90% of world trade relies on trade finance, mostly short-term credit, insurance, or guarantees. At the same time, the ADB’s 2025 Global Trade Finance Gap Survey puts the global trade-finance gap at US$2.5 trillion — unchanged from 2023 and equivalent to roughly 10% of world trade.
That combination matters: the market is large and essential, yet access to capital remains constrained. The opportunity therefore does not arise from taking blind risk; it arises from financing transactions for which the route back to cash can be verified.
Two invoices that look identical to a model
A simple example illustrates the difference. Imagine two invoices, each for EUR 1 million, each due in 90 days, each issued to apparently strong buyers, and each covered by credit insurance. In a model, they look almost identical. In practice, they can be very different.
In the first, the goods have not yet been delivered, the buyer has not accepted them, the insurance assignment is unclear, and a payment dispute could block any claim. In the second, delivery is confirmed, the buyer’s obligation is accepted, the receivable is properly assigned, the policy terms are reviewed, and the repayment route is clear.
The model sees two similar invoices. The underwriter sees two very different risks. That is why, in trade finance, the edge is not only in estimating default probability — it is in proving that the cash-flow path is real, legal, correctly insured and enforceable before capital is deployed.
A word on the term “alpha”
I use the word “alpha” more carefully. In the strict investment sense, alpha is the excess return after proper risk adjustment. As the CFA Institute describes it, alpha is the risk-adjusted excess return remaining after accounting for market risk exposure.
What trade finance delivers is better described as a structural premium: compensation for illiquidity, complexity, documentation, legal structuring, insurance basis risk, and access to transactions that many banks or larger lenders cannot process efficiently.
This is not a weaker claim. It may be a more durable one, where the structure is genuinely hard to replicate. A public-market signal can be copied and arbitraged away. A strong trade-finance structure is far harder to replicate because it depends on access, legal control, verified documents, insurance, and collection mechanisms.
These are not standardised numbers on a screen — they are practical controls that either exist or not. Using the label too loosely risks conflating a structural premium with pure excess return.
In the same spirit of precision, we would avoid sweeping claims that trade-finance funds have “never had a negative month.” Well-known failures in the sector, including Greensill, show why the asset class earns more trust by naming its failure modes than by denying them.
Credit insurance is valuable — but it is not a guarantee by label
Credit insurance is central to how we de-risk a transaction, though it should never be treated as an automatic backstop. Export-credit insurance provides conditional cover against buyer non-payment, and claims may be declined if policy requirements are not met.
In practice, cover applies only if the receivable, policy terms, assignment, documentation, notification process and claim conditions all align. Confirming that the insurance is assignable and that its conditions are met is a precondition of financing for us, not an afterthought.
Where quantitative thinking genuinely adds value
None of this makes our business model-free. The honest statement is that models are necessary but not sufficient. Quantitative work earns its keep at the portfolio level, particularly in addressing three problems:
- First, hidden concentration: individual transactions may appear uncorrelated, yet exposure can cluster within the same buyer, insurer, originator, sector, country, commodity or shipping corridor.
- Second, fraud detection across repeated transactions.
- Third, pricing insurance basis risk — the gap between the risk one believes is insured and the risk the policy pays out under stress, a distinction that becomes particularly important when an insurer reduces or cancels a limit.
The most dangerous mistake, though, is to mistake the low historical default rate for a model. The data is broad, but for the private and more bespoke segment, it can still be selective, incomplete and procyclical. A model trained on it essentially learns that little ever goes wrong — until an idiosyncratic event arrives and dominates the picture.
For any structure that aims to deliver a stable return, that tail is the thing to respect, not to assume away.
In summary
Yes, the risk and return models in trade finance are mature. But they describe a world where aggregate data makes it appear calmer than the segment we finance.
The opportunity is not that the asset class has historically shown low defaults; it lies in carefully selected, well-documented, short-duration transactions where the repayment source is visible, legal and insurance protections are tested before capital is deployed, and the portfolio is built to avoid hidden concentration.
The return is not mainly generated by a better forecast. It is generated by disciplined execution in a market where many risks are avoidable if the structure is right — and costly if it is not.
And if a quant wants to take it further, the relevant question for us is: “How do I price correlation and insurance basis risk in a portfolio where all individual deals claim to be uncorrelated?” That is where we believe the most useful investor discussion begins, and we are grateful for your feedback.
Notes on sources
The ICC Trade Register coverage figure (close to a quarter of global trade-finance transactions, 22 contributing institutions) reflects the 2024 edition; the ADB trade-finance gap of US$2.5 trillion is from ADB’s 2025 Global Trade Finance Gap Survey. WTO and export-credit-insurance references are general and should be confirmed against the current primary publications before external use, as figures vary by edition.
