Banca d’Italia published a July 2026 mystery-shopping paper that sent 200 USDC across ten corridors and found total remittance costs from 0.30% to nearly 9% of the amount transferred. The on-chain leg averaged about 0.4%. Fiat funding, exchange spreads, and cash-out fees did most of the damage.
The paper, Are Stablecoins Efficient for Remittances?, is Markets, Infrastructures, Payment Systems No. 86, by Alberto Di Iorio, Enrica Di Stefano, Michele Mascioli, and Giorgio Trebeschi. Independent write-ups from CoinDesk and crypto.news matched the same core numbers after the study circulated at the end of July.
How the experiment worked
Researchers acted as ordinary retail customers. They funded exchanges, bought USDC, moved it on a public chain, sold it, and withdrew local currency. That is the full “stablecoin sandwich”: fiat in, on-chain transfer, fiat out.
Design choices from the paper:
- Instrument: USDC only (Circle’s dollar stablecoin), chosen for regulated access in the EU and Japan under local crypto rules.
- Ticket size: 200 USDC, aligned with the World Bank Remittance Prices Worldwide (RPW) retail benchmark.
- Hub: Italy, with bidirectional corridors to Argentina, Brazil, South Africa, the UAE, and Japan.
- Dates: live transfers on 24 and 26 March 2026.
- Chain (baseline): Ethereum.
- Venues: Binance and Kraken on the Italian side; Ripio (Argentina), Foxbit (Brazil), BitOasis (UAE), and Valr (South Africa) on the spokes.
Each transfer was split into five phases: funding, purchase, on-chain transfer, sale, and withdrawal. That breakdown is what lets the paper show where money and time go, not only the headline total.
The cost spread, corridor by corridor
Across eight fully comparable corridors (Japan is treated separately), total costs ran from 0.30% (Italy → Argentina) to 8.96% (Argentina → Italy). Other published corridor totals include Italy → Brazil at 2.70%, Brazil → Italy at 2.21%, Italy → South Africa at 4.58%, South Africa → Italy at 5.44%, and both UAE routes near 7–9%.
The on-chain transfer was almost always the cheap middle. It averaged roughly 0.4% and hit 0.01% on Brazil → Italy. Funding cards, trading spreads, and withdrawal fees dominated.
Argentina needs a caution label. The rock-bottom Italy → Argentina print used the official BCRA retail FX rate. Parallel market rates (including the rate implicit in USDC/ARS trades) still sit above that official print. The paper says the apparent bargain partly reflects that FX gap, not pure blockchain efficiency. The reverse Argentina → Italy run was expensive in large part because buying USDC with pesos was costly under the same measurement frame.
Against World Bank RPW country averages, USDC undercut outbound averages for Brazil (2.21% vs 9.96%), South Africa, and Italy’s outbound average. The UAE was the clear exception: traditional RPW averages looked cheaper than the card-heavy USDC path the researchers could complete.
Versus Wise simulations on the same bilateral pairs and amount, USDC was cheaper on three comparable routes (including Brazil → Italy at 2.21% vs Wise’s roughly 4.7–4.9%) and more expensive on four (including Italy → Brazil at 2.70% vs Wise’s 2.20%). The authors stress different sample dates and limited observations: this is an indicative bench, not a universal ranking.
Speed is a payments-infrastructure story
On-chain settlement finished in under 15 minutes on seven of eight main corridors. End-to-end time told a different story.
Where domestic instant rails funded and cashed out the exchange:
- Italy (TIPS)
- Brazil (Pix)
- Argentina (Transferencias 3.0)
…full bank-to-bank-style settlement landed in under 20 minutes. South Africa, relying on ordinary bank transfers at the edges, took one to two business days. The blockchain was not the bottleneck; the local banking last mile was.
Japan could not be compared cleanly. Retail access to dollar stablecoins and outbound transfer rules forced an unhosted-wallet intermediate step and transfer fragmentation. Nominal costs looked low (about 1.3–1.6% on the routes tested), but the workflow was not ordinary retail remittance UX.
What that means for dollar rails people already use
Latin America is not a footnote in this paper. Argentina and Brazil are two of the five destination legs, and the authors cite literature putting Latin America among the regions with the highest cross-border stablecoin flows relative to GDP. For freelancers, families, and SMEs who already move dollars as USDC or USDT, the practical read is blunt:
- Cheap chain ≠ cheap remittance. The fee you see on a blockchain explorer is usually the small slice.
- Corridor and venue matter more than the ticker. The same 200 USDC can land under 3% or near 9% depending on exchange, funding method (bank transfer vs card), and which way the money is flowing.
- Instant local rails are the real speed unlock. Pix-side Brazil and Transferencias 3.0 Argentina made the full path competitive on time; slow bank edges erased the on-chain advantage.
- Keeping the stablecoin (no final off-ramp) can change the economics, especially where people prefer dollar balances over local cash. That “toast” model is outside the full sandwich the bank measured for remittance-style end-to-end cost.
None of this is a verdict that stablecoins “don’t work” for cross-border money. It is evidence that they do not automatically beat money-transfer operators once you pay to get in and out of crypto. The paper’s own conclusion: no systematic cost advantage; performance is corridor-specific; on/off-ramp frictions and domestic payment quality are first-order.
Takeaway
Banca d’Italia put 200-USDC remittances through real exchanges and real corridors and got a messy, useful map: costs from about 0.3% to 9%, on-chain fees mostly tiny, and speed gated by Pix-class local rails rather than block confirmation. For readers who already send or receive dollar stablecoins between Europe and Latin America, treat exchange spreads, funding method, and cash-out fees as the product decision, not the gas line.
This is research on payment cost structure, not a recommendation to use or avoid any particular coin, corridor, or app.
