Legible Before Secure: How Digital Payments Can Formalize Poverty

When the Tax Collector Arrives Before the Consumer Protection Does

In July 2018, Uganda’s government looked at 27 million mobile money accounts and saw a tax base. A 1% levy on all mobile money transactions took effect on July 1. Within a month, transaction values had fallen 24%. McLeod Russel, a major tea processor, stopped paying its estate workers by mobile money and went back to flying cash by helicopter to six estates because the digital channel had become more expensive than the physical one. The government backed down within weeks, cutting the levy to 0.5%, but the signal had already been sent: a system built on public subsidy and development-sector backing could be turned into a revenue instrument the moment it reached scale.

Tanzania tried the same thing in 2021. A levy stacked on top of existing VAT and excise duty pushed mobile money taxes to 60% of total transfer costs, according to GSMA’s analysis. Person-to-person transactions fell 38% in three months, from 30 million to 18 million per month. Cash came back. Human rights organizations went to court. The government eventually abolished the levy in 2022. Ghana’s 2022 e-levy — 1.5% on digital transactions — caused 47% of users to reduce their mobile money usage, by the Imani Center for Policy and Education’s reckoning, before the rate was cut to 1%.

These aren’t cautionary tales from the fringe of digital payments policy. They are the main story. When digital payment systems reach scale, they change who can see poor people’s economic lives, and they create tools that other actors — governments, platforms, lenders, employers — can use long before the poor themselves have built the rights and buffers to make that visibility safe.

James Scott made this argument in “Seeing Like a State” (1998), drawing on land registries, collectivized agriculture, and urban planning. His central point: states prefer legible subjects. Informal life, precisely because it is hard to see and measure, gives poor people a margin of autonomy that formal systems tend to erode. Digital payments are the most powerful legibility technology deployed in developing countries since the land registry. The question Scott’s framework forces on the table is whether legibility produces inclusion or control — and for whom, and in what sequence.

The Gains Are Real. So Is the Trap.

The scale of what has happened is genuinely impressive. The World Bank’s Global Findex shows the share of adults in developing economies making or receiving digital payments rose from 35% in 2014 to 57% in 2021. By then, 71% of adults had a formal account. This is infrastructure, not pilot.

The poverty case is real. Tavneet Suri and William Jack’s 2016 paper in *Science* estimated that access to M-PESA lifted roughly 194,000 Kenyan households out of extreme poverty, with women showing the most concentrated gains through movement into business and away from subsistence farming. The mechanism was shock absorption: when illness, drought, or a bad harvest hit, mobile money allowed faster and wider transfers from support networks than cash had permitted. That matters enormously in places where informal insurance is often the only insurance.

Brazil’s Pix shows what public digital rails can do at scale. Launched in November 2020, the Banco Central do Brasil’s instant payment system processed 64 billion transactions in 2024 — a 53% year-on-year increase — worth approximately R$26.4 trillion, roughly twice Brazil’s GDP. By end-2024, 76% of Brazilians were using Pix, more than debit cards (69%) or cash (69%). The Fundação Getulio Vargas estimated that small businesses using Pix saw average revenue increases of 15-20% because they could accept digital payments without card terminal costs. Pix succeeded because it was designed as a public utility: instant, always on, cheap, and simple.

India’s UPI reached tens of billions of monthly transactions by early 2026. The World Bank’s own evidence briefs on digitized government-to-person (G2P) payments find consistent patterns: lower leakage, lower operating costs, reduced waiting time for recipients, and better predictability of benefit flows. In fragile settings — where a delayed payment can mean a distress sale of assets rather than a managed consumption shortfall — predictability has a real poverty value.

So the case for digital payments is solid. The problem is what happens when the state, having invested in digital rails, starts treating them as instruments for something other than inclusion.

India: Precision in Delivery, Gaps in Allocation

Start with India, because India is the most instructive large-scale example.

Aadhaar-based biometric authentication was introduced into India’s Public Distribution System — the subsidized grain program that serves hundreds of millions — to reduce leakage and improve delivery. Reetika Khera and Jean Drèze’s research, published across multiple studies in the *Economic and Political Weekly* and an NBER working paper by Muralidharan, Niehaus, and Sukhtankar, tells a consistent story: the reform produced “pain without gain.” Leakage reduction was limited, partly because identity fraud was not the main source of leakage in the first place. Meanwhile, authentication failures created new exclusion — biometric mismatches for manual workers whose fingerprints are worn, network failures in rural areas, errors in database seeding that made legitimate beneficiaries invisible to the system.

The consequences were not abstract. Khera and Amod Moharil’s 2024 survey of Aadhaar correction centres in Delhi found that 40% of those experiencing authentication errors had to make multiple visits to get corrections made. Drèze and colleagues documented over two dozen starvation deaths between 2017 and 2019 in Jharkhand directly linked to Aadhaar-induced exclusion from welfare — cards cancelled for non-linking, authentication failing for months without fallback. An 82-year-old woman named Sukra Oram had enrolled in Aadhaar but could no longer authenticate biometrically because her hands had changed with age. She died without retrieving her benefits.

The World Bank’s own analytical framing makes the key distinction with unusual clarity: Aadhaar is proof of identity. It is not proof of eligibility or priority. DBT is a secure payment pipeline. Neither answers the harder question of who should receive support. A digital state can get dramatically better at paying the person it has selected while still selecting the wrong person, missing vulnerable people, or creating exclusion through system failures that fall disproportionately on those least able to navigate bureaucratic redress.

This is the core distinction the legibility framing forces into view: precision in delivery is not the same as justice in allocation.

East Africa: The Informal Economy Gets Discovered

Back to East Africa, because this is where the next phase of digital payment policy will be most consequential.

The mobile money tax episodes in Uganda, Tanzania, and Ghana illustrate one specific sequencing failure. Governments invested in financial inclusion through mobile money partly because it was good policy and partly because donors and development finance institutions encouraged it. Once the systems were embedded in daily life — once enough poor people depended on them for wages, remittances, and payment — those same systems became attractive as revenue instruments. The 2018 Uganda levy was designed to help meet a revenue target of 16.2 trillion Ugandan shillings. Tanzania’s 2021 levy was explicitly justified as a way to tax the informal economy now that it had become visible.

That justification is not irrational. Informal economies do need to contribute to public finances. But the sequencing mattered enormously. In Uganda, lower-income users who migrated away from mobile money because of the levy moved to agent banking, where no comparable taxes applied — a system requiring physical access that poorer, more rural users had less of, as UNCDF’s impact study documented. The burden fell disproportionately on the poor. In Tanzania, rural users who reverted to cash lost access to savings tools, loans, and government transfers that came through the mobile money ecosystem. The levy taxed not just a transaction but a set of capabilities that had been built on top of it.

Cash, in these cases, was the fallback. But it’s worth being precise about what that means, and honest about what cash actually does. Cash is private. It works offline. It is hard to audit at scale. For a street vendor or day laborer, that opacity can be a shield against predatory relatives, arbitrary officials, or rigid welfare rules. These are real benefits. Cash is also how abusive partners control household finances, how local officials extort benefit recipients at the last mile, and how petty corruption compounds poverty. The opacity that protects also enables. Digital payments replace those specific dynamics — with different ones. The issue is whether the new dynamics serve poor people better, and under what conditions.

The payment mechanism literature adds texture here. IPA’s 2024 evidence review on cash transfer payment channels finds that digital payments can cut transaction costs and reduce time burdens for recipients. But it also identifies what drives the difference between systems that work and those that don’t: account usability, payment predictability, low transaction costs, local payment infrastructure quality, and the support ecosystem around the payment system. The payment rail itself is only part of the story. A polished digital interface sitting on weak last-mile infrastructure still produces exclusion. The infrastructure visible to a World Bank project team and the infrastructure visible to a rural beneficiary with an unreliable phone and a distant agent are often two different things.

What Success Exposes

Brazil shows what the mature version of this challenge looks like. Pix has become central enough to daily life that fraud is now a first-order policy problem. Chargeback rates in Brazil run at roughly 3.5% — compared to 0.47% in the US — and financial losses from Pix-related fraud reached $700 million in 2023, per the analysis firm PayByrd. Scale is a triumph and a vulnerability simultaneously. A system that handles 76% of Brazil’s adults and processes more daily transactions than Visa and Mastercard combined is also an attack surface, a concentration risk, and a source of market power for whoever sits between users and the rails.

These are the governance questions that follow success. The IMF’s 2025 paper on digital payment innovations in Sub-Saharan Africa — by Ali, Balima, Espinosa-Vega, and Lacaille (IMF Working Paper WP/25/27) — argues the region should encourage mobile money and fast payment systems while strengthening infrastructure, interoperability, competition, and security. That’s right. The region’s challenge is no longer proof of concept. It’s how to govern systems that have become basic economic infrastructure for the poor.

What would governing them better actually require?

Uganda and Tanzania’s mobile money tax reversals point to one concrete lesson: don’t use digital rails as a first-pass revenue instrument before the system has built user trust and before formal consumer protections are in place. Tax design matters. Exemptions for small transactions, thresholds below which no levy applies, and explicit protection for G2P payment flows would have dramatically changed who bore the burden of those African levies.

India’s experience points to a different lesson: build grievance redress before building enforcement. The Aadhaar CAG report in 2022 found grievance redress mechanisms remained inadequate a decade after the system launched. The sequencing was backwards: biometric authentication was mandatory before exception handling was functional. A simpler rule would have required working fallback channels — OTP authentication, alternative IDs, functional correction centres — before making biometric authentication a precondition for food rations.

A third lesson follows from both: identity systems should not silently become eligibility systems. Aadhaar was designed as proof of identity. Its practical function in welfare became proof of eligibility, with errors in the identity layer cascading directly into exclusion from food and pension entitlements. The distinction Drèze insisted on — identity, eligibility, and priority are three separate questions requiring three separate policy answers — should be a design principle, not an academic observation.

None of this is about slowing digital payment adoption. Countries that have built functioning public digital rails — Brazil’s Pix, India’s UPI, the mobile money ecosystems of East and West Africa — have created something genuinely valuable. The question is what political and institutional commitments need to accompany that infrastructure. Scott’s “Seeing Like a State” ends not with an argument against legibility but with an argument for what he calls metis: the practical, local, experiential knowledge that gets destroyed when top-down systems optimise for their own administrative convenience. Digital payment systems optimised for state capacity and platform efficiency at the expense of that local knowledge — of the seasonal income volatility of a market trader, the documentation gaps of a migrant worker, the biometric unreliability of an elderly manual laborer — will produce the outcomes India and Uganda and Tanzania documented.

The hard part was never making the poor legible to formal systems. That turns out to be technically achievable. The hard part is making the formal systems worthy of that legibility.

-----

Sources: Global Findex Database 2021 (World Bank); Suri & Jack, “The Long-Run Poverty and Gender Impacts of Mobile Money,” Science, 2016; Banco Central do Brasil, Pix Statistics (bcb.gov.br/estabilidadefinanceira/pix); GSMA, Tanzania Mobile Money Levy Impact Analysis, 2022-23; UNCDF, “The Impact of Mobile Money Taxation in Uganda,” 2021; Muralidharan, Niehaus & Sukhtankar, “Identity Verification Standards in Welfare Programs,” NBER Working Paper 26744, 2020; Drèze, Khalid, Khera & Somanchi, “Pain without Gain? Aadhaar and Food Security in Jharkhand,” Economic and Political Weekly, 2017; Khera & Moharil, 2024 Delhi Aadhaar correction centre survey; IPA, “Payment Mechanisms for Cash Transfers,” Evidence Review, 2024; Ali, Balima, Espinosa-Vega & Lacaille, “Digital Payment Innovations in Sub-Saharan Africa,” IMF Working Paper WP/25/27, 2025; James C. Scott, “Seeing Like a State,” Yale University Press, 1998.