The PMJDY Sign Flip: What India's Financial Inclusion Data Looks Like When You Read It Wrong

If you plotted PMJDY enrollment against digital payment use across Indian states, you would conclude that India's largest financial-inclusion program is associated with less digital payment activity, not more. The correlation is negative. Pearson r of about -0.42 in log-log. The scatter slopes downward. More PMJDY accounts per adult, fewer UPI transactions per person.

That is what the data shows, and it is completely misleading.

UPI's user base was not built by one program alone. Aadhaar provided the biometric identity layer that made mass account-opening possible. Cheap mobile data, driven partly by Jio's entry and spectrum policy, gave people a way to access their accounts digitally. Direct Benefit Transfer pushed government payments into those accounts, giving people a reason to use them. PMJDY is one piece of a larger policy stack. This analysis focuses on it specifically because it is the variable with state-level enrollment data detailed enough to test. The broader stack matters. PMJDY is the piece the data can test.

What the scatter actually looks like

The chart slopes downward. The states with the most PMJDY accounts per adult, Bihar, Jharkhand, Uttar Pradesh, Madhya Pradesh, Rajasthan, are near the bottom of the UPI distribution. The states with the least PMJDY enrollment, Goa, Delhi, Chandigarh, are near the top. A casual reader of this chart would reasonably conclude that the program isn't working, or worse, that it's associated with less digital payment activity.

That reading is wrong, and the reason it's wrong is more interesting than the chart itself.

Why the negative correlation is misleading

PMJDY was designed to reach the unbanked, and the unbanked are concentrated in poorer states. Bihar has the highest PMJDY enrollment per adult in the country at roughly 0.75 accounts per adult. Bihar is also one of the poorest states in India. Poorer states use less UPI. Not because of PMJDY, but because income is the dominant driver of digital payment adoption across Indian states. The income elasticity is close to one: a 10 percent richer state uses roughly 10 percent more digital payments.

So the negative raw correlation between PMJDY and UPI is not a PMJDY effect. It is income showing up through PMJDY. The states with the most PMJDY accounts are the states with the least income, and income is what drives UPI adoption. The correlation is real. The implied causation is backwards.

The sign flip

Once you control for income, the relationship reverses. Among states at similar income levels, those with higher PMJDY enrollment use more digital payments per person, not less. The elasticity is positive and statistically significant in two separate regressions: +0.50 in the early-UPI era (2019-21) and +0.34 in the mature-UPI era (2023-26).

The flip is not a statistical artifact. It is what the data says once income is held constant. And it shows up in both eras, five years apart, with different dependent variables (a five-rail composite in the early era, pure UPI in the mature era), different sample sizes, and different time structures. The same pattern, in both periods, with the same direction.

The mechanism is account access. PMJDY opens a basic savings account for unbanked households. An account is the necessary precondition for using a digital-payment rail at all. The data shows that this measure of account access predicts digital payment use beyond what income alone does. Not that PMJDY caused UPI adoption. The finding is associational, and there are state-level factors the data cannot fully control for. But the underlying logic is hard to argue with: you need an account to use the rail, and PMJDY is what gave most of the unbanked population an account.

Why PMJDY survives the income control when nothing else does

This is the part that makes the finding interesting beyond the numbers.

Four state-level variables were tested alongside income in both regressions: PMJDY enrollment per adult, bank-office density, urban share, and (in the mature era) internet density. Only PMJDY survives. Bank-office density correlates with income at +0.85. Urban share correlates with income at +0.7. Internet density correlates with income at +0.86. All three move tightly with state income. Once income is in the regression, they add nothing that income has not already explained. They show up through income, not beyond it.

PMJDY survives for one structural reason. Its enrollment correlates with income at roughly -0.62. Negatively. The program targets the unbanked, who are concentrated in poorer states, so its beneficiary density runs in the opposite direction from income. That negative correlation is why PMJDY still shows up after controlling for income. Income and PMJDY move in opposite directions across states, so controlling for one does not wipe out the other.

This draws a real distinction between two kinds of financial inclusion. Infrastructure that scales with the market, bank offices, telecom, urbanization, shows up through income. Programs that deliberately target people outside the market, as PMJDY does, show up separately. The policy design is what determines whether the program shows up separately from income in the data.

What this means for investors

The investor version of this is simpler.

Most of the cross-state variation in digital payment use is just income. A richer state uses more UPI. A fintech operating in a richer state gets lifted by that tide. That is the economy growing, not the product working.

The PMJDY finding suggests there is something else in the data, adoption that does not just follow income. States where PMJDY reached more people per adult show higher digital payment use than their income predicts. For a fintech operating in those states, the question becomes: is your user growth following income, or is some of it coming from the user base that PMJDY built?

If the latter, the growth may be more durable than income alone would predict, because it is drawing on a user base that was created by policy, not just by economic development. That distinction rarely comes up in how people talk about Indian fintech investing.

The open question

The finding is associational. The data cannot tell us that PMJDY enrollment caused higher digital payment use. There are state-level political, cultural, and historical factors that the regression cannot control for, and PMJDY enrollment is not randomly assigned across states.

There is also a gap between accounts opened and accounts used. A significant share of PMJDY accounts are dormant or carry zero balances. The regression measures enrollment, not usage. The distinction between "having an account" and "using digital payments" is real, and it is one of the things that needs stronger data before the inclusion story is as clean as the headline suggests.

What the data does establish is a consistent cross-state pattern that holds across two different eras of UPI, with two different dependent variables, and the same controls. The pattern is consistent with the hypothesis that targeted financial-inclusion programs leave a measurable footprint on digital-payment adoption beyond income. The data does not yet prove it.

---

A broader look at how the state-level data changes fintech evaluation is here. The full paper is in progress.