What the Numbers Can—and Cannot—Tell Us
This chapter examines how the trade-finance gap is measured and what those measurements leave hidden. The goal is practical: to read a headline such as “$2.5 trillion” or “41% of SME applications rejected” without confusing a provider estimate with a firm census, or observed letter-of-credit use with access to finance. The numbers are strong enough to establish a persistent problem, but not strong enough to tell us every reason a particular firm did—or did not—obtain bank protection.
The global benchmark is a scale signal, not a census
The Asian Development Bank’s 2025 Global Trade Finance Gap Survey estimates that unmet demand for trade finance remains $2.5 trillion. That is the same dollar estimate reported in the 2023 edition. The unchanged headline is important, but it needs a comparison base. In the earlier edition, the gap represented about 10.6% of average global merchandise trade. In the newer edition, merchandise trade had grown, so the gap represented around 10% of merchandise trade.
The two statements can both be true: the gap’s share of trade can fall while the absolute amount of unmet demand remains $2.5 trillion. A smaller percentage does not make the remaining constraint small. Nor does an unchanged dollar estimate prove that no access improved. It says that growth in the volume of goods traded has not been matched by a comparable disappearance of unmet financing demand.
The denominator is also deliberately limited. ADB continues to use merchandise trade—goods—as the basis for consistency with earlier editions. The brief notes that adding services would produce a larger global trade denominator and a different percentage. “About 10%” should therefore be read as roughly 10% of merchandise trade under this survey’s comparison, not 10% of every cross-border economic transaction.
The numerator is an estimate of unmet demand, not a ledger of rejected applications. ADB describes the figure as the only global benchmark of unmet demand for trade finance and says it is intended to help policymakers, banks, and development institutions understand the scale and distribution of the problem. The estimate is based on survey information from more than 110 trade-finance providers. Those respondents may represent up to one-third of the global market, while the total size of that market is not known.
That design gives the benchmark a specific strength. Providers are close to the supply decision: they see requests, constraints on liquidity and capital, country risks, compliance costs, and the products they can offer. Their responses can reveal whether the banking system perceives demand and supply to be expanding or constrained. But a provider-side benchmark cannot observe every firm that does not approach a bank. It also cannot, by itself, establish that every request would have been profitable, creditworthy, or socially valuable if approved.
The survey’s timing matters as well. The underlying data used to estimate the 2025 gap cover 2023 and 2024, while respondent views were collected in the first half of 2025. Some figures describe observed or reported past activity; others describe expectations in a period of geopolitical and policy uncertainty. ADB reports that more than 73% of respondents expect the supply of trade finance to increase, and its news release says 80% expect demand to increase. The detailed brief gives 87.5% for a more specific question about demand rising as trade diversification accelerates. These percentages should not be merged into one universal forecast: they refer to related but distinct survey questions.
The direction of the expectation is nevertheless clear. Companies are seeking new suppliers, markets, and regional trading relationships, and banks expect those changes to generate financing needs. A new corridor often begins with less accumulated information than an old one. Firms need working capital and risk mitigation while banks are still learning about the counterparties, legal environment, currency, and logistics. A stable aggregate gap at that moment can mean that relative coverage has improved in some places while the next wave of trade is creating new demand elsewhere.
Regional figures make the unevenness visible, but they must remain in their own scope. In the selected emerging markets summarized by ADB, trade finance supported about 25% of trade in parts of West Africa, from 3% in Cambodia and the Lao People’s Democratic Republic to 20% in Viet Nam in the Lower Mekong region, and about 10% in Central America. The comparison cited by ADB is at least 60% in advanced economies. These are findings from particular country studies using their own data and methods, not components that can be added to produce a global rate.
The studies also measure a market’s financing architecture, not just a bank’s willingness to issue a letter of credit. A low coverage share can mean that firms have little access to banks, that supply-chain finance is unavailable, that traditional instruments dominate, or that firms have learned to avoid applications after earlier failures. ADB notes that the regional studies use different base data, even though their broadly similar gap magnitudes help validate the direction of the global concern. Their differences are therefore evidence of uneven conditions, not a reason to manufacture a precise ranking of countries.
Some of the regional work uses a model to estimate what might happen if coverage expanded. The ADB brief reports an estimate that doubling trade-finance coverage would increase trade flows by about 8% on average in the markets studied. That is a modeled counterfactual, not an observed global result. It is useful because it translates access into a possible economic consequence; it does not tell us that every additional unit of finance will generate the same trade response, or that finance alone can overcome weak demand, logistics, or market access.
The same caution applies to prices. The selected studies report an average letter-of-credit cost of around 2%–5% of transaction value, while working-capital loans used as a substitute often averaged about 15% annually. A transaction fee and an annual interest rate are not identical costs, and neither number is a universal price for every market or firm. Their value is diagnostic: formal protection can be expensive in the very markets where access to safer, cheaper alternatives is weakest.
The hidden margin behind observed use
The most important measurement problem appears when the unit of analysis becomes the firm. There are at least five different states that can be hidden behind a simple statistic: a firm may need trade finance and apply; apply and be accepted; apply and be rejected; decide not to apply because it expects rejection or cannot manage the process; or trade without a letter of credit because another payment term or financing source is suitable. These are economically different outcomes.
ADB’s SME figures show why the distinction matters. In the 2025 survey, the reported rejection rate for SME trade-finance requests was 41%, close to the 40% rate for large and mid-cap corporates. It would be tempting to read the near equality as proof that the SME disadvantage has disappeared. ADB does not make that claim. The convergence may reflect dedicated SME strategies, better screening, support from nonbank financial institutions, changes in bank risk appetite, or a reduction in the number of SMEs willing to apply after previous rejections. A single survey cannot determine which explanation dominates.
The missing applicants are not necessarily firms without a real need. Evidence from regional African work summarized by ADB indicates that around 17% of SMEs in Kenya and Tanzania reported a genuine need for trade finance but did not apply. The reasons included high collateral requirements, fear of rejection, and unfamiliarity with the process. This is sometimes called self-rationing: the firm removes itself from the formal application pool before a bank makes a decision.
That behavior changes the interpretation of a rejection rate. If many of the most constrained firms never apply, the observed rejection rate may fall even while exclusion remains high. Conversely, a high rejection rate does not prove that every rejected transaction was viable or that a bank could safely have approved it. The rate describes decisions within the pool of applications. It does not describe the counterfactual pool of firms that would have applied under different prices, information, documentation, or advisory support.
A more informative access measure would follow the firm through the whole decision path. It would record the amount and term requested, whether the firm applied, the price and collateral required, the bank’s decision and stated reason, whether the firm used another payment arrangement, and whether a non-applicant had a genuine financing need. It would also distinguish a new customer from a repeat borrower. Without those links, a rise in approved transactions could simply mean that already-established firms are doing more business, while a fall in rejection could mean that the most discouraged firms have disappeared from the application pool.
Consider two possible future patterns. If applications from young and small traders rise, approval becomes more common, and the total cost of finance falls without a deterioration in repayment performance, the evidence would support broader access. If reported rejection falls while applications from those firms also fall, the same headline would be consistent with stronger self-rationing. Neither scenario can be inferred from a single aggregate rejection rate. The missing information is not a technical footnote; it determines whether an apparent improvement reaches the firms at the center of the access problem.
The same issue appears in the study of letters of credit by Vietnamese firms. The study’s outcome is whether a firm used a letter of credit for a reported trade flow. That is valuable evidence about payment choice, but it is not a direct observation of an application, a rejection, a fee quote, or a firm’s preferred but unavailable term. A firm with low observed use may be excluded, may rely on open-account trade because its relationships are strong, or may use another form of finance. The payment indicator alone cannot distinguish those cases.
This is why a user profile can be misleading. If existing letter-of-credit users are older, larger, and more productive than other traders, one interpretation is that stronger firms use the instrument more. Another is that these characteristics make the instrument easier to obtain, leaving firms with weaker records underrepresented among users. The evidence supports asking which mechanism is operating; it does not support choosing one explanation from the descriptive comparison alone.
The distinction between use and access also protects against a common policy error. If the objective is to close a financing gap, it is not enough to increase the number of letters of credit recorded in a market. More use could reflect safer relationships, larger transactions, or cheaper bank access for firms already able to participate. It could also reflect useful new coverage for previously excluded traders. The statistic becomes meaningful only when paired with information about who applied, who was rejected, who self-rationed, what the product cost, and whether the transaction could proceed under an alternative term.
The available evidence can support three firm conclusions. First, the global problem is large and persistent under ADB’s merchandise-trade benchmark. Second, access is uneven across markets and the expected reorganization of trade may increase demand before information and finance catch up. Third, the visible application and usage data omit an important part of the problem. They cannot, by themselves, tell us whether non-use represents successful private trust, an unaffordable product, a rejected application, or a firm that never entered the process.
Those limits are not a reason to discard the numbers. They tell us how to use them. The ADB survey establishes scale, direction, and system-level constraints. Firm-level data can then investigate how characteristics such as size, age, ownership, and experience shape payment choices. But to connect the two, the analysis needs a model that holds risk, bank cost, transaction scale, and hidden demand together. Without that model, a lower ratio can be mistaken for a solved problem, and a higher rate of use can be mistaken for broader access.