Digital Financial Readiness and the Formal Borrowing Gap

The Inversion: Everyone Borrows, Almost Nobody Borrows Formally

The financial inclusion story of the last decade was told in accounts opened. That story is largely won. The Global Findex Database 2025 — the World Bank's demand-side survey, fielded in 2024 and released in 2025 — records Sub-Saharan African account ownership rising from 34% of adults in 2014 to roughly 58% in 2024, with Kenya (90.1%), Ghana (81.2%) and South Africa (81.1%) at or near universal coverage.

The credit story tells the opposite. Findex measures borrowing behaviour separately from account ownership, and the divergence is stark. In Uganda, 90.2% of adults borrowed money in the past 12 months, but only 10.8% borrowed from a formal bank or similar financial institution. In Nigeria the figures are 78.9% and 6.1%. In Kenya, 81.9% and 12.6%. In Ethiopia, 53.7% and 2.6%.

This is the formal borrowing gap: the difference between credit demand that is unambiguously revealed and credit supply that is formally intermediated. On the fourteen markets examined here it averages roughly 60 percentage points. Credit is being extended at enormous scale — it is simply being extended by family, savings clubs, traders and mobile operators, none of whom report to a bureau, none of whom price risk explicitly, and none of whom build a durable credit history for the borrower.

Digital Readiness Is Not the Same as Digital Credit

A useful way to read these markets is as a two-factor scorecard: digital financial readiness — the infrastructure and behaviour that generates scoreable data — and credit conversion, the extent to which that data is actually turned into formal or quasi-formal lending.

The two do not move together. South Africa has the highest readiness profile on the continent and one of the lowest digital-credit conversion rates. 81.1% of adults hold an account, but only 1.7% borrowed from a mobile money provider — a rounding error against Kenya's 32.4%. South Africa's credit market is bank-and-bureau shaped: it has real bureau infrastructure, so it never developed a telco-led credit rail. The result is a large, digitally visible population that remains bureau-thin at the lower income deciles.

Kenya, Uganda and Ghana show the inverse. Bureau coverage is limited, but mobile money generated a transaction ledger deep enough for M-Shwari, Fuliza, MoKash and MTN Qwikloan to underwrite tens of millions of micro-advances. That is alternative-data credit scoring operating at national scale — already, and without waiting for bureau reform.

The strategic reading is that the readiness question is not whether a market is digitised, but whether its digital exhaust is actually used to price risk. On that measure, Nigeria (63.3% accounts, 4.4% mobile-money borrowing) and Egypt (43.1% accounts, 1.0%) are the largest unconverted pools on the continent.

The Readiness Stack: Phone, Wallet, Payment, Loan

Readiness is not a single number. Findex 2025 lets us decompose it into four sequential layers, each measured directly: owning a mobile phone (con1), holding a mobile money account (mobileaccount.t.d), making or receiving a digital payment (g20.any) and finally borrowing on that rail (fin22b).

The attrition between layers is where strategy lives. Kenya converts 92.7% phone ownership into 87.5% mobile money accounts and 32.4% digital borrowing — a 35% wallet-to-loan conversion. Nigeria converts 83.8% phone ownership into only 32.8% wallets and 4.4% digital borrowing: a 13% conversion. Morocco owns phones at 90.2% and holds mobile money accounts at 5.9%, the sharpest break on the continent.

South Africa is the instructive middle: 87.0% phone ownership and 67.2% digital-payment usage — the second-highest digital payment rate in the sample — but 31.6% mobile money accounts and 1.7% mobile-money borrowing. The data exists and moves; it simply never reaches an underwriting model.

The practical test for any market entry or product decision is the ratio between the digital-payment column and the digital-borrowing column. Where that ratio is wide — Nigeria (54.5% vs 4.4%), Egypt (36.3% vs 1.0%), South Africa (67.2% vs 1.7%) — there is a payments rail carrying scoreable volume that nobody is underwriting against.

The Cost of the Gap: Who Cannot Raise Emergency Money

The borrowing gap is not an abstraction about market size. Findex asks whether an adult could come up with emergency funds within 30 days, and where that money would come from. The answers show what happens when formal credit is unavailable.

In South Africa, 31.8% of adults say raising emergency funds in 30 days is not possible at all — the worst result of the fourteen markets, despite 81.1% account ownership. Nigeria is at 25.0% and Egypt at 24.2%. Kenya, with the deepest digital credit rail on the continent, records 3.4%; Senegal 4.1%; Uganda 4.4%.

The source of funds is equally telling. Family and friends is the primary emergency lender almost everywhere — 55.2% of adults in Egypt, 49.0% in Morocco, 43.7% in Ghana. A bank, employer or private lender is the main source for just 2.0% of Ethiopians, 2.4% of Moroccans, 2.8% of Senegalese and 6.4% of Nigerians. Kenya (14.0%) and Tanzania (9.9%) again stand apart.

Read together with the borrowing data, the pattern is consistent: markets that turned digital payment rails into credit rails have visibly more resilient households, and markets that did not have pushed the shock-absorbing function onto family networks — an informal insurance system that fails precisely when shocks are correlated.

Where the Credit Actually Comes From

Findex separates borrowing by source, and the channel mix explains why bureau-based scoring fails. Family and friends is the dominant lender almost everywhere: 68.3% of Ugandan adults, 55.2% of Nigerians, 55.0% of Senegalese. Savings clubs — stokvels, chamas, susu, tontines — carry a further 30.4% in Uganda, 20.3% in Senegal and 16.5% in Kenya.

None of this activity is observable to a traditional credit bureau. A borrower can service a dozen informal obligations flawlessly over a decade and still present as a thin file. That is the mechanical reason exclusion persists after accounts are opened: the repayment history exists, it is simply stored in the wrong place.

The corollary matters for lenders. In Kenya, Uganda and Ghana, mobile-money borrowing already exceeds savings-club borrowing — meaning the informal ledger is being partially digitised and made scoreable. In Nigeria, Egypt, Morocco and South Africa it is not. Those four markets contain the largest volumes of unscoreable, revealed-repayment behaviour on the continent.

Who the Gap Excludes

The distributional detail in Findex 2025 is where the policy case sharpens. Formal borrowing is systematically skewed by income, gender and geography.

In Kenya, 15.8% of the richest 60% borrowed formally against 7.8% of the poorest 40% — a factor of two. Men borrowed formally at 16.7% against women at 8.8%, a 7.9-point gender gap on a base of roughly 12%. In Ghana, urban adults borrow formally at 13.8% versus 7.0% rural. In South Africa the income gradient is 14.1% versus 9.1%, and the urban-rural gradient 14.0% versus 10.6%.

Senegal is the informative exception: women borrow formally at 20.2% against men at 14.7%, and rural adults (20.7%) outpace urban (14.2%) — a pattern consistent with the country's dense mutualist and microfinance network rather than with commercial banking. It demonstrates that the gradient is a function of institutional design, not of borrower quality.

Alternative data attacks exactly these gradients, because mobile money penetration is far less income- and gender-skewed than bank branch access. That is the mechanism by which behavioural scoring converts inclusion into credit rather than merely into accounts.

Does Alternative Data Actually Work?

The claim that behavioural and transactional data can substitute for a bureau file is often asserted and rarely evidenced. The strongest published test remains Bjorkegren and Grissen's paper on behaviour revealed in mobile phone usage predicting credit repayment (World Bank Policy Research Working Paper 9074, 2019), which used mobile-phone metadata to predict default among borrowers and found predictive performance broadly comparable to conventional credit scoring — in a setting where conventional scoring was largely unavailable.

Subsequent African work is consistent. Peer-reviewed 2025 research on machine-learning credit scoring for informal African merchants, published in Human-Centric Intelligent Systems, reports that transaction-level merchant data materially improves default discrimination over demographic baselines, and Kenyan digital-lending studies using stepwise logistic regression and weight-of-evidence methods reach the same conclusion on domestic portfolios.

Three caveats belong in any executive read of this literature. First, performance decays: behavioural models trained on one macro regime degrade quickly when inflation, fuel prices or airtime pricing shift. Second, proxy discrimination is real — mobility, handset value and airtime top-up patterns correlate with protected attributes, and models must be tested for disparate impact, not just AUC. Third, consent and data-protection law is now binding, not aspirational: Nigeria's NDPA (2023), Kenya's Data Protection Act (2019) and South Africa's POPIA constrain what telco and transaction data may lawfully be repurposed for underwriting.

The honest position is therefore narrower than the marketing: alternative data reliably expands the scoreable population and improves ranking of thin-file borrowers; it does not by itself make bad credit good, and it introduces governance obligations that most lenders have not yet operationalised.

So What? Three Implications

The Findex 2025 data reframes what a credit strategy in Africa has to solve for.

Conclusion: Close the Measurement Gap, Not the Access Gap

Africa's financial inclusion agenda spent fifteen years solving access. Global Findex 2025 confirms it largely worked: most adults in most measured markets now hold an account, and Sub-Saharan account ownership has grown from roughly a third of adults to nearly three-fifths in a decade.

The next fifteen years are a credit problem, and it is fundamentally a measurement problem. Adults across the continent are borrowing at rates of 55% to 90% a year. They are repaying. The data describing that repayment sits in wallets, savings groups, merchant tills and telco ledgers — visible to almost everyone except the institutions legally permitted to lend.

The research question posed at the start of this analysis — whether digital financial readiness and alternative data can strengthen creditworthiness assessment for the previously excluded — has a defensible empirical answer in three markets already. Kenya, Uganda and Ghana did not wait for bureau reform; they underwrote against the digital ledger they had. The open question is whether the continent's four largest unconverted markets choose to do the same.