The price of raising dollars through the foreign-exchange market depends partly on which dealer handles the trade and which side of that dealer’s book the customer occupies, according to a new Bank of England staff study of 27.3 million outright forwards.
That is an execution-quality finding, not evidence that dealers have charged abusive prices. The study finds meaningful price dispersion after placing trades into narrow comparison groups, but it cannot fully separate market power from relationship services, collateral terms, bargaining and customer choice.
The distinction matters for corporate treasurers, asset managers and insurers that use forwards to obtain or hedge dollars. An aggregate market curve may be a useful reference, but it is not necessarily the price available to a particular customer.
From 178.5 million reports to 27.3 million trades
The Bank staff working paper, by Marco Grotteria and Alex Kontoghiorghes, uses confidential UK transaction reports from January 2021 to October 2025. The authors began with 178,473,017 UK EMIR reports for FX forwards.
They restricted the data to six dollar pairs involving the Australian dollar, Canadian dollar, Swiss franc, euro, sterling and yen. That left 46.7 million reports. Quote-orientation and price filters reduced the population to 45.7 million, requiring at least one observed G15 dealer parent cut it to 38.4 million, and excluding trades between entities in the same banking group produced the final 27,291,457 arm’s-length forwards.
The cleaning is material. Raw reporting data can include amendments, cancellations, valuation updates and daily state reports for one contract. The researchers reconstructed those lifecycles and retained one record for each economic transaction. Fourteen G15 parent groups appeared across the six pairs and 58 months.
For the central comparison, trades had to share a currency pair, a half-hour execution window and a ten-day maturity bucket. These controls remove the common forward curve and much of the market-wide movement around execution. The resulting dealer “wedges” are relative price estimates around that common component, not standalone quotes or a measure of the total cost of synthetic dollar funding.
Dispersion survives the first controls
The monthly dealer wedges had standard deviations of 4.4 annualised basis points when a dealer bought foreign currency forward and 3.9 when it sold foreign currency. The distance between the 10th and 90th percentiles was roughly seven basis points.
The dispersion was not simply a permanent league table of cheap and expensive dealers. After the authors removed both common month effects and fixed dealer effects, standard deviations of 3.4 and 3.1 basis points remained. Weighting the results by notional did not materially change the pattern, which argues against tiny dealer-month samples driving the averages.
The second finding is directional. The paper defines a dealer selling foreign currency as buying dollars at maturity. On that side, the average estimated wedge was 2.15 annualised basis points, compared with minus 0.38 basis points when the dealer bought foreign currency. The difference was 2.52 basis points.
It became much larger at the front end. For maturities up to one month, the difference between the two sides was 11.82 annualised basis points. It fell to 4.10 basis points between one and three months and 2.07 basis points beyond three months.
Annualisation is crucial. An 11.82 basis-point annualised difference on a 30-day contract is roughly 0.97 basis points over that term. On $100 million of notional, that is about $9,700 before allowing for the precise contract convention and currency conversion. It is an illustration of scale, not an estimate of a customer’s all-in saving from switching dealer.
Relationships explain some of the level, not the asymmetry
The authors tested whether different dealers merely serve different customers. They added fixed effects for each client-dealer relationship, then allowed those effects to vary by currency pair. They also controlled for trade size, electronic versus bilateral execution and whether the pair was among the three most liquid in the sample.
The directional gap stayed between 2.23 and 2.66 basis points across those specifications. By contrast, the sum of the two directional wedges, which the paper uses to test for different relative dollar-funding conditions across dealers, fell from 1.78 basis points to as little as 0.20 and was no longer statistically distinguishable from zero.
That split disciplines the interpretation. Observable transaction characteristics and persistent client relationships account for much of the average level. They do not remove the difference between the two sides of dealers’ books. The result is consistent with a mixture of intermediation costs and residual customer demand, rather than a clean estimate of dealer markup.
Nor does the dataset show whether a customer requested competing quotes, could move its collateral arrangement or valued balance-sheet access and operational reliability enough to accept a different price. Those are central to any claim about pricing power. A relationship can create switching friction and still provide a service that a half-hour price comparison does not capture.
The practical test is a better execution record
For an FX execution lead, the paper suggests that transaction-cost analysis should segment trades by dealer, direction and maturity, rather than compare every fill with one market midpoint. The useful internal record would preserve the request-for-quote set, executable alternatives, timestamp, collateral terms, size, venue and reason for dealer selection.
For treasury and market-risk teams, the maturity conversion should be explicit. Annualised basis points make contracts of different tenors comparable, but budgets and realised funding costs are paid over the contract term. A large annualised short-dated gap can be economically relevant when notionals are large or positions roll frequently without being the same as an annual charge on the full amount.
The Bank for International Settlements estimated that non-banks outside the United States had $26 trillion of off-balance-sheet dollar obligations in FX swaps, forwards and currency swaps at mid-2022, while non-US banks had $39 trillion. That establishes why short-term synthetic dollar funding matters. It does not justify applying the paper’s price gaps to those global stocks, which cover different instruments, entities, jurisdictions and dates.
This is Staff Working Paper No. 1,207, not an official finding of the Bank or its policy committees. Its sharper contribution is narrower: comparable dollar forwards did not clear at one observable transaction price, and the largest difference appeared where maturities were shortest. The next question is whether customers with auditable access to several executable quotes face the same pattern.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.