Notes for Tamara -- on what account ownership counts, and what it misses.
I have been spending time with the 2025 Global Findex, and a handful of things surprised me enough that I wanted to write them down. They bear on the question your office has been pushing -- the move from access toward financial health -- so I am sharing them in case they are useful. None of it needs anything from you.
The starting point is genuinely good news. Account ownership climbed from % of adults in 2011 to % in 2024, and Kenya went from % to %. What got me curious was a narrower question: once almost everyone has an account, what does counting accounts stop telling us? Three things kept coming up, and by the end they turned out to be the same story.
In the economies with usable data, women are more likely than men to say they could not come up with emergency funds within 30 days. That holds in of them. The average gap is about points, and in Pakistan and Nigeria it reaches roughly and points.
The part I did not expect is that the gap holds up when you account for the things that ought to explain it away. It survives an income control, so it is not simply that women are poorer (income comes out statistically irrelevant, p=0.92). It survives the gender gap in account ownership too. Eswatini and South Africa are the clearest cases: men and women there hold accounts at essentially the same rate, and women are still 8 to 10 points more likely to be unable to raise emergency money. Closing the account gap, which the field has largely done, left this one roughly where it was.
Across these economies, account ownership runs a median points ahead of daily internet use, and the gap is largest exactly where ownership is highest. That sounds alarming on its own, so it needs a careful caveat, because mobile money complicates it.
In Kenya, % of adults have an account, % use mobile money, and only % are online daily -- so most of those accounts are reached over USSD, with no internet at all. The account-minus-internet gap is really an upper bound on "banked but offline," and where mobile money is strong it overstates how many people are actually out of reach. Tanzania looks much the same (% banked, % mobile money, % online). Sri Lanka is the genuine version of the problem: % banked, only % on mobile money, % online daily -- accounts that are real but hard to use for much beyond a balance. Its low mobile-money figure is more a design choice than missing activity -- Sri Lanka built bank-account-linked rails, LANKAQR and CEFTS, rather than USSD wallets -- which if anything sharpens the point rather than softening it.
These are daily-internet figures from the 2024 survey, so they are an upper bound in time as well: the cliff narrows as people come online. India, the widest case in the chart, had passed roughly 1.09 billion internet subscriptions by early 2026 -- subscriptions, not daily users, but a sign the offline share is shrinking -- even as rural access stays well behind the cities.
The ranking is what struck me. If you rank economies by who is still unbanked, then rank them again by this usability cliff, the two orders point in opposite directions (Spearman ). Kenya is th of on the first measure -- almost no one is unbanked -- and th on the second. The places that won the access race are first in line for the next problem, which is turning owned accounts into richly used ones.
The same gap shows up in payments. India is 8th in the world by account ownership and th by digital-payment use -- a -rank fall. Sri Lanka falls . These are Findex's own published numbers, nothing modeled by me. Worth flagging that this is the 2024 survey: India's UPI volume has roughly doubled since, so the gap here is about how many adults transact at all, not how much the already-active ones do. The breadth that is still missing sits about where the fourth finding points -- rural, women, lower-income.
Some accounts are not just lightly used; they are dormant. In Algeria, % of everyone who holds an account holds a dormant one, with Libya, Mauritius, and Ethiopia not far behind.
One clarification on the word. Dormant here means no activity, not no money, and the two are easy to conflate. India is the clearest example: the share of its accounts sitting empty fell sharply over the past decade, yet on the government's own administrative figures the share sitting unused has if anything grown, with roughly a fifth to a quarter of its mass-program accounts reported inoperative through 2025. The empty-account problem easing is not the same as the unused-account one easing.
There is a clean counterpoint, and it points back to mobile money. Economies where mobile money is widespread are nearly immune: about % of holders dormant, against % where accounts are mostly bank accounts. Where money actually moves through the account, the account stays alive.
Ethiopia, near the top of that dormancy chart, is where the transition may be starting to show. Since the survey, its banks moved onto a shared national switch and person-to-person transfers passed ATM withdrawals for the first time -- early signs the rails are catching up to the account base, which is the shift this whole section is about.
One caution I want to be explicit about. I tested whether informality, or accounts being opened for a government payment or wage, could explain dormancy across countries, and that story does not hold up -- the relationships are weak or run the wrong way once you control for income. So I am not claiming a cause here. This section is descriptive.
When I went to the public-use microdata and asked who actually falls into each of the three -- within the same country, holding the country fixed -- the answer kept coming back the same. Among account-holders across economies, a dead account is more likely to belong to someone offline (by points), in the poorest 40% (by ), with the least education (by ), and to a woman (by ).
These are not four separate maps that happen to overlap; the same traits stack on the same accounts. If you sort holders by how many of the four they carry -- none, one, two, three, all four -- the chance the account is dead climbs at every step, from % with none to % with all four. And it climbs within the same country (a -point gap holding the country fixed), so it is not just that disadvantaged people live in higher-dormancy places.
So a holder who is offline, poorer, and less-educated, and a woman, is roughly times more likely to hold a dead account than one with none of the four. That is the real pattern -- a risk that compounds as disadvantage stacks. I want to be careful not to push it further than it goes: she is only about % of all holders and holds about % of the dead accounts, not most of them. Dead accounts are spread across the whole gradient; they just grow denser the more disadvantage a person carries. What stays true is the simpler thing: the traits that go with a hard-to-get account are the same ones that go with an unused one, and the measure counts all of them as success.
Three gaps with three different maps -- but the same handful of traits drives all three, and an account is likeliest to go dead where they overlap.
I want to be honest about the strength of this last part. The public microdata cannot see dormancy in roughly 50 economies, including Algeria and several of the Gulf cases, and my reconstructed dead-account measure validates at about 0.95 against the official one on the economies it does cover. So treat the gradient as a direction, not a precise figure -- I have tried to break it a dozen ways and the shape holds, but the exact numbers will move.
If there is a thread, it is that account ownership did much of its job and is now partly behind us, and the measure that comes next looks more like usable inclusion -- reach, resilience, and real activity together, not ownership alone. That sits more in your team's field than mine, so I offer it as a starting point rather than a conclusion.
Kenya may be the clearest case: near-universal accounts in Findex, and yet the 2024 FinAccess survey put the share of adults who are financially healthy at fewer than one in five. The distance between holding an account and being financially well is what the next measure has to capture.
If any of this is useful, I would be glad to talk it through, or to take the microdata version further if it is ever helpful to you. Either way, I thought you would find it interesting.
Karthik
I would rather under-claim than have any of this fall over under scrutiny, so here is what each one rests on.
| Finding | Claim | Status |
|---|---|---|
| 1 | Women more fragile, net of income and account access | Survives every control |
| 2 | The usability cliff sits on the account-ownership winners | Holds, as an upper bound |
| 3 | Ownership rank is not usage rank; many accounts are dormant | Descriptive |
| 3 | Informality or supply-push causes dormancy | Tested and dropped |
| 4 | Dead-account risk climbs as offline, poorer, less-educated, woman stack (within country) | Exploratory |
The causal "informality drives dormancy" story slopes the wrong way once income is controlled, so I have left it out rather than show a chart that an expert would rightly knock down.