Governments worldwide are trying to move retail payments from cash to instant digital rails. The policy debate often focuses on launch: central-bank fast-payment infrastructure, retail-facing central bank digital currencies, mandates, and fee subsidies. This reflects the limits of older technologies: cards have expanded, but card-based systems often remain more attractive to higher-income users and have not fully displaced cash (Alvarez and Argente 2024, Han and Wang 2021). Mobile money promises to extend financial services to people underserved by banks (Muellbauer and Aron 2019).

The harder question comes after launch. A platform can be technically elegant, publicly supported and still fail to become the way people pay (Alvarez et al. 2023a). Cash is hard to beat: it is immediate, simple, widely accepted, and helps users track spending. Peer-to-peer digital payments can preserve the central advantages of cash – instant settlement and close transaction control – while adding convenience, security, and digital records. But they must be easy for poorer households to use, valuable enough for banks and merchants to support, trusted enough for users to try, and widespread enough that each new user can transact with many others. Because many adults remain outside formal finance, overcoming these barriers is central to financial inclusion. Leveraging granular microdata on individual transactions and user characteristics, we study the drivers of digital payment adoption in Brazil, Costa Rica, and Mexico (Argente et al. 2025).

Barriers to adoption

Figure 1 shows why the same technology can face different entry costs. Around 2017, Brazil and Costa Rica had much higher bank-account ownership than Mexico, which also lagged in card ownership, debit-card use, prior digital-payment experience, mobile-phone ownership, and internet access. These gaps matter because Brazil’s Pix, Costa Rica’s Sinpe Móvil, and Mexico’s CoDi require a bank account and a mobile phone. Pix and CoDi depend on internet-connected apps, while Sinpe Móvil can also be used through SMS.

Connectivity is a second layer of the same problem. Internet quality varies substantially across municipalities within these countries. Instant payments do not require high bandwidth for every transaction, but poor connectivity makes routine actions – opening the app, loading a QR code, confirming a payment, checking that money arrived – less reliable. These small frictions matter when cash is the alternative.

Figure 1 The starting barriers were not the same

Notes: Financial-inclusion and technology-access indicators before the main adoption episodes, with Brazil, Costa Rica, and Mexico highlighted. 
Source: Argente et al. (2025), based on World Bank WDI and Global Findex data.

Adoption is not the same as enrolment

A useful first test is whether people actually transact, not simply whether they can. We therefore collect adoption data from each platform over time. Figure 2 measures adoption as active participation in a transfer. Pix reached about 60% of Brazil’s adult population within a year and moved close to universal adult adoption by 2024. Sinpe Móvil grew more slowly but reached around 80% adoption by 2024. CoDi remained at only about 2–3% active adoption by the end of 2024, despite more validated accounts.

This distinction is central for policy. Enabling accounts, registering users, or downloading apps does not create a payment network; active use does. In successful cases, adoption followed the familiar S-curve of a network technology: slow take-up, rapid acceleration once useful to others, and saturation.

Figure 2 Three adoption paths: Pix, Sinpe Móvil, and CoDi 

Notes: Adoption is measured as active participation in at least one transfer, scaled by the population aged 15 and above. 
Source: Argente et al. (2025), based on central-bank administrative data and population data.

Key pattern: A rapid low-income gradient

A decisive feature of Pix and Sinpe Móvil is not only scale. They spread beyond affluent early adopters. Figure 3 shows that, in both Brazil and Costa Rica, the average income in municipalities where users were adopting the platforms fell over time, as adoption moved from richer toward poorer places.

Usage patterns tell a complementary story. Figure 3 compares payments per capita with local GDP per capita early and late in diffusion, demonstrating that the income gradient in usage becomes much flatter. By 2024, lower-GDP municipalities were using Pix and Sinpe Móvil far more intensively than at launch; in Costa Rica, the usage gradient was almost flat. The key takeaway is that broad cash substitution is associated with rapid low-income catch-up. This highlights an inclusion challenge: usage among adopters can equalise even when adoption shares remain higher in richer places, so reducing entry costs still matters.

Figure 3 Income and adoption dynamics

Notes: Regions correspond with municipalities. GDP per capita per municipality corresponds with year 2021. Panels (a1) and (a2) depict the GDP of the municipalities where someone had adopted, defined as having made at least one transaction, over time. Panels (b1) and (b2) compare payments per capita and regional GDP for the earliest and latest year in which data is available. Panels (c1) and (c2) also make a comparison across years, but for the share of adopters per region against regional GDP. 
Source: Argente et al. (2025).

The marginal user changes as the network grows

Microdata from Costa Rica shows inclusive diffusion at the individual level. Early Sinpe Móvil adopters were young, urban, disproportionately male, and more likely to be high-skilled. As adoption expanded, new adopters increasingly resembled the broader population.

Figure 4 shows this transition. The average age of new adopters rose from around 35 to above 40. The share of women rose from below 40% and converged to around 50%. The urban share fell, and the share of low-skilled workers among new adopters rose to about 60%. These patterns show the platform escaping the early-adopter segment.

Figure 4 Sinpe Móvil moved beyond the early-adopter group

Notes: Characteristics of marginal Sinpe Móvil adopters over time. 
Source: Argente et al. (2025), based on Costa Rican administrative microdata.

The mechanism here is network value: a platform becomes more useful when more relatives, friends, co-workers, and businesses can use it (Alvarez et al. 2023b). Figure 5 shows Sinpe Móvil moving from a fragmented 2016 network with many isolated users to a large core in 2020 and one dense network by 2024. Once dense, joining is no longer mainly a bet on the future; it is instead immediately useful.

Figure 5 Network density increased as adoption rose

Notes: Network structure of Sinpe Móvil peer-to-peer transactions in 2016, 2020, and 2024. Values in parentheses show the percentage of adults who had adopted Sinpe Móvil by the end of each year. 
Source: Argente et al. (2025).

Why CoDi did not tip

Mexico’s CoDi shows how adoption can stall before network effects take hold. CoDi faced a harder initial environment: fewer banked adults, lower mobile-phone ownership and digital-payment experience, and more dependence on reliable connectivity for QR-based use. Awareness stayed limited: around one-third of adults in 2021 and less than 40% in 2024. CoDi also had a weaker supply-side value proposition: voluntary participation, tighter restrictions for fintech firms and non-bank providers, and limited bank fee incentives to integrate aggressively. The result was a thin network on both sides of the market.

We also compare CoDi with DiMo, a newer Mexican phone-number transfer option, and find that awareness remains higher among richer households for both systems, but the income gap in actual use is much larger for CoDi. The broader design lesson is that reducing transaction steps can matter most for lower-income users.

Consumer adoption is not enough for merchant adoption

Person-to-person success does not automatically produce merchant adoption. Businesses face different frictions: accounting, reconciliation, transaction limits, fraud management, and payment confirmation. Figure 6 shows firms lagging behind individuals in Costa Rica; while in Brazil, we see a gradual rise in the person-to-business share of Pix transactions as the individual network expanded. This does not make person-to-person adoption irrelevant for firms; however, it is not sufficient. To become a retail-payment substitute for cash, merchant tools must meet the needs of businesses, not just consumers sending money to friends.

Figure 6 Firm adoption lagged behind consumer adoption

a) Sinpe Móvil: Individuals versus firms

b) Pix: Person-to-business share 

Notes: Panel A shows Sinpe Móvil adoption by individuals and firms in Costa Rica. Panel B shows the person-to-business share of Pix transactions in Brazil.
Source: Argente et al. (2025).

When adoption is broad, cash starts to move

Our final question is whether these systems substitute for cash. Figure 7 compares actual currency in circulation with synthetic-control counterfactuals for Brazil, Costa Rica, and Mexico. In Brazil and Costa Rica, actual cash in circulation diverges downward from the synthetic comparison after Pix and Sinpe Móvil expand. In Mexico, after CoDi, there is no comparable divergence.

Evidence from Costa Rica points the same way locally: regions with higher Sinpe Móvil adoption have lower withdrawals from local cash-custody facilities. Survey data is consistent with substitution: adopters reported making payments without cash for about 15.6 days on average, compared with 6.4 days for non-adopters. The finding is not that cash disappears altogether. It is that broad digital-payment adoption changes the need to hold and withdraw cash.

Figure 7 Currency in circulation fell below the synthetic benchmark in the successful cases

Notes: Actual and synthetic log currency in circulation. The vertical line marks the launch date of each platform. 
Source: Argente et al. (2025), using the synthetic-control approach of Abadie and Gardeazabal (2003).

The same inclusive networks can also deepen formal finance. In Brazil and Costa Rica, we show an acceleration in depositors at commercial banks after Pix and Sinpe Móvil launched. Microdata from Costa Rica shows that first-time Sinpe Móvil users were more likely to start using other digital transfer services shortly afterwards. The key is making formal financial services more useful to people who previously relied on cash.

Policy lessons

Our evidence points to five design priorities for instant-payment systems, retail central bank digital currencies, and other next-generation payment infrastructure:

1. Measure active use, not registration.
2. Design for low-income users from the start by reducing complementary-technology requirements and making the interface simple enough to compete with cash.
3. Ensure interoperability and supply-side participation at launch. Brazil used a stronger mandate, Costa Rica benefited from large public financial institutions’ early participation, and Mexico relied more on voluntary and restrictive participation, slowing network formation.
4. Treat awareness and trust as infrastructure. Central-bank backing helps, but users must know the system exists, understand how to use it, and believe payments will be safe and reversible when appropriate. Security and fraud-resolution rules should be visible before launch, not hidden in technical documents.
5. Build merchant-facing tools separately from consumer tools. Firms need payment confirmation, accounting integration, and reconciliation systems.

The broader lesson is that the success of instant payments is distributional. A platform concentrated among richer, urban, digitally savvy users will not replace cash economy-wide. Pix and Sinpe Móvil worked because adoption moved quickly towards poorer users and the network became dense. CoDi shows that without low-income catch-up, the promise of digital payments can remain unrealised.

References

Abadie, A and J Gardeazabal (2003), “The Economic Costs of Conflict: A Case Study of the Basque Country”, American Economic Review 93(1): 113-132.

Alvarez, F and D Argente (2024), “Consumer Surplus of Alternative Payment Methods”, Review of Economic Studies, rdae112.

Alvarez, F, D Argente, and D Van Patten (2023a), “Are Cryptocurrencies Currencies? Bitcoin as Legal Tender in El Salvador”, Science, 382(6677), eadd2844.

Alvarez, F E, D Argente, F Lippi, E Méndez and D Van Patten (2023b), “Strategic Complementarities in a Dynamic Model of Technology Adoption: P2P Digital Payments”, NBER Working Paper No. 31280.

Argente, D, P G Alvarez, E Méndez and D Van Patten (2025), “Drivers of Digital Payment Adoption: Lessons from Brazil, Costa Rica, and Mexico”, NBER Working Paper No. 34280. 

Han, P and Z Wang (2021), “Technology Adoption and Leapfrogging: Racing for Mobile Payments”, SSRN working paper.

Muellbauer, J and J Aron (2019), “The economics of mobile money: Harnessing the transformative power of technology to benefit the global poor”, VoxEU.org, 7 May.