BIS Competing Rails Paper: Bank, Fintech or Stablecoin Costs

The BIS competing rails paper is BIS Working Paper 1385, "Competing rails for crossborder payments: banks, fintechs, and stablecoins", by Wenxin Du, Catherine Huang and David Scharfstein, published October 19, 2026. It compares the all-in cost of bank wires, fintech money transfer operators and stablecoin transfers, and concludes that stablecoins are not a panacea for cross-border payment frictions.

How the BIS measured all-in cost

The authors use one formula for all three rails: all-in cost equals the FX markup against a Bloomberg benchmark rate plus explicit fees, both expressed in basis points of the amount sent. A basis point is 0.01%, so 100 basis points is 1%. The markup is the gap between the benchmark and the rate the customer received, and fixed fees become basis points by dividing by the payment size.

The rails are banks (correspondent banking), fintech money transfer operators such as Wise and Remitly, and stablecoin transfers. Fintech prices were collected every 30 minutes from the public websites of three providers between January 23 and May 3, 2026, across about 100 corridors at $200, $500, $10,000 and $50,000. Stablecoin costs come from exchange prices, fee schedules and blockchain fee data.

Three details decide how far the numbers travel. The fintech, B2B-platform and stablecoin figures are US-outbound, while the bank sample is a handful of the authors' own personal wires between US and non-US banks, in both directions. The bank figures also leave out wire fees. And the paper's stablecoin sandwich is the retail version, where a sender converts local money into a dollar stablecoin, moves it on a blockchain and converts it back; the fiat legs at both ends are where Lumx sits, as the product section below shows.

This is a BIS Working Paper: research circulated for discussion, not a peer-reviewed policy position of the BIS. Its policy-relevant conclusion is that stablecoins may compete for retail-sized payments while displacing rails for large institutional transfers stays limited, as the liquidity section below shows.

What did the BIS find for each of the three rails?

The BIS findings and the figure each one rests on (Working Paper 1385, sections 1, 5 and 6)

The authors explain the bank result by architecture and pricing power: wholesale spreads for EUR, JPY and GBP against the dollar are 6 to 9 basis points, so a thin market cannot explain 200 to 400. Fintech operators net opposite flows and pay out of local balances they already hold. Stablecoins, in the authors' words, reallocate intermediation across institutions rather than remove it: centralized exchanges play a role analogous to correspondent banks.

Where the averages break for Brazil, Mexico and Colombia

The headline numbers sit on dollar-euro and dollar-sterling. The paper's tables also carry rows for BRL, MXN and COP, where the rails rank differently. All figures below are US-outbound.

BRL, MXN and COP across the three kinds of evidence in the paper (basis points of the amount sent)

Sources are Tables 3, 4 and 5. A negative stablecoin figure means the stablecoin rate beat the Bloomberg benchmark for a payment leaving the United States. The B2B column is FX markup only: one fintech platform executing US wires for a community bank and an educational institution in 2024 and 2025, with a fixed $8 wire fee left out.

Three readings follow. The fintech column falls fast with size, the fixed-fee effect the authors describe. The B2B platform column is the only business-flavored evidence in the paper, and for Tier 3 and Tier 4 currencies it averages above 300 basis points. BRL and COP sit above the retail operators even at $200; MXN does not. And COP is the outlier on the stablecoin side: a positive 59.2 basis points, against a market where the paper's data shows about $1 million a day of trading against dollar stablecoins.

The samples are small and the standard deviations wide (45.9 basis points for BRL and 81.9 for COP in the stablecoin column). The rows show that rails rank differently by currency. They are not a price list, and the corridor explainer covers how a real route is built.

Why stablecoin FX looks cheaper in some markets

For emerging-market currencies the paper reports stablecoin FX costs that are often negative, meaning a better rate than the benchmark for a sender leaving the United States. The authors call this a stablecoin access premium: where residents struggle to buy dollars, they pay extra for dollar stablecoins, and the sender collects that premium.

The authors say plainly that this advantage arises from capital control distortions, not from technological efficiency. The IMF discussant notes that Table 5 already shows the premium compressing. For a Brazil, Mexico or Colombia payout, one good day says little about the next quarter. The USDC to BRL, USDC to MXN and USDC to COP pages show the live conversion.

How deep is stablecoin FX liquidity?

Shallow, and uneven by currency. Beyond the euro, the paper's text names only five currencies (KRW, TRY, BRL, GBP and THB) averaging more than $10 million a day against dollar stablecoins, and its tables add few others.

Average daily trading against USDT and USDC on centralized exchanges, 2025 to 2026, $ millions (Table 5)

The text also gives a longer-run euro figure, about $186 million a day on centralized exchanges over 2021 to 2025, against 488.6 in 2025 to 2026 in the table. In a small pair, one large payment can move the price, which is why displacing traditional rails for high-value institutional transfers is currently limited.

For a business paying invoices, this sets how far you can size up before the price moves. Ask a provider where the conversion is executed and for the largest ticket quoted at a flat price.

Speed, comparability and the cost the rail does not set

On speed, the paper cites SWIFT's 2024 data that about 90% of payments over the network reach the destination bank within one hour. Most delay sits in the last mile, between the beneficiary bank and the final credit, driven by compliance checks and batch processing. Stablecoin transfers settle on a blockchain in minutes, but end-to-end speed depends on the fiat on-ramp and off-ramp at both ends. The paper's example is Circle's expedited fee for same-day USDC redemption into dollars.

On Lumx, the fiat leg is the part with a published clock. Pix (Brazil's instant payment system) and SPEI (Mexico's interbank transfer system) payouts are instant, 24/7/365, with no cut-off. Faster Payments in GBP is instant, up to £1M per payment. SWIFT in USD has a 3:00 PM ET cut-off and takes 1 to 5 business days, subject to intermediary bank hours.

Wise's unit economics for a GBP to EUR transfer, as the paper reports them, are a marginal cost of £0.24 plus 0.06% of the value. Of the £0.24, £0.11 is customer support and £0.06 is compliance. The payment rail itself, covering bank fees, treasury and FX, is £0.02, under 10% of the fixed cost. The authors note that stablecoin cost estimates tend to leave out exactly those two.

The paper is published with a discussion by Tobias Adrian of the IMF, dated June 2026, which raises a fair objection: the three rails are not measured the same way. Bank figures exclude wire fees, fintech figures are all-in, and the stablecoin components sit in separate tables without a single total. His proposal is to compare the same cost concept across matched corridors, one period and one size.

He also cites forthcoming IMF work covering 270 corridors. For a $200 stablecoin transfer it reports an average cost of about 2.2%, ranging from roughly -2% to 8%, and in about one third of corridors the cheapest traditional option beat the cheapest stablecoin option. An average hides both very good routes and very bad ones.

How to use the paper before choosing a rail

<!-- OPERATOR PARAGRAPH --> I would not use this paper to pick a provider, and I would not dismiss it either. I recommend turning its formula into a test you can run in an afternoon: ask each candidate for the amount that lands at your usual payment size, with every fee inside it, quoted in the same minute, and compare that one number. I would put the blockchain fee last on the checklist, as the paper does, and add a line for who handles compliance and support, since the paper's own data puts most of the fixed cost there.

That test also answers the comparability objection: if every quote is the amount in the recipient's currency at your size, the markup versus fee split stops mattering. The pricing explainer breaks that number into spread, rail fee and network cost.

When the BIS numbers are the wrong yardstick

The paper is the wrong guide in four cases. If you send from Brazil, Mexico or Colombia, its data runs the other way: the fintech, B2B and stablecoin tables are US-outbound, the bank wires run in both directions between US and non-US banks, and the paper does not study payouts from Latin America. If you pay invoices of $50,000 or more, its fintech data flattens past $10,000 and stablecoin pricing is least tested. If you need a rate fixed for hours or days, none of the three rails in the paper is a forward contract. And if you send recurring business payouts, a retail cost ranking fits you less than it fits a one-off personal remittance.

What a Lumx quote shows against the BIS cost components

Lumx is stablecoin payments infrastructure for businesses that move money between Latin America and the rest of the world: one API to collect, hold, convert, and pay out in BRL, MXN, COP, USD, EUR, and GBP or in USDC and USDT, over local rails such as PIX, SPEI, PSE, Bre-B, ACH, FEDWIRE, SEPA, and Faster Payments, with SWIFT and on-behalf-of payments and collections (POBO and COBO) in USD, EUR, and GBP, plus named virtual accounts, custodial wallets, and KYB/KYC built in.

The BIS test is the one number the recipient sees, and a Lumx locked quote gives you its closest pre-trade version. A request for a locked rate gives back a base exchange rate, a final exchange rate and a final target amount, plus an expiry. The lock runs 30 seconds, 1 minute or 5 minutes; the 30-second lock has no fee and the longer ones carry a fee for volatility risk. A new quote always comes with a new exchange rate ID. If your platform adds its own margin, a partner fee in basis points, a flat amount or both, it is included in the locked quote.

On rails, the payouts the paper's tables point to run as follows. Pix and SPEI payouts settle instantly, around the clock, and COP pays out over Bre-B (Colombia's instant payment system) or ACH Colombia. Bre-B has a regulatory cap of 1,000 UVB per transaction, about COP 12.1 million, and the rail is fixed when the destination is registered, so an invoice above the cap is refused on a Bre-B destination, not split or rerouted. For those invoices, register an ACH Colombia destination, which has no per-payment limit. USD can also leave by SWIFT, with UETR and MT103 confirmations and a settlement window of 1 to 5 business days. On the SWIFT route the docs warn that intermediary banks can deduct fees outside Lumx's control, so what arrives can still change, which is the correspondent-banking cost the BIS describes.

On liquidity, conversion uses Lumx's own liquidity and institutional counterparties only, among them BTG Pactual. Hedging instruments for longer horizons come from your bank or FX broker. The coverage page lists pairs and rails by currency, and the off-ramp explainer covers the conversion step.

The next step is a test. Take your usual ticket, request a locked quote for it once you are onboarded, and read the final target amount against what your recipient needs to receive. The exchange rates docs show the request.

Methodology and sources

Checked on November 1, 2026. Every figure about the paper comes from the full text of BIS Working Paper 1385, read from the PDF that day, including Tables 3, 4 and 5, the Wise unit-economics data and the IMF discussion. The SWIFT one-hour figure and the IMF corridor study are citations inside the paper that we did not read directly. Lumx facts come from docs.lumx.io pages read that day: exchange rates, partner fees, rail cut-off times and coverage.

For the regional picture, see Latin America stablecoin payments versus traditional. For international USD wires, see stablecoins vs SWIFT.

Verified on November 1, 2026. Operational context, not legal, tax, or investment advice.

Cover photo: Haberdoedas on Unsplash.