Both Sides Are Partly Right: Measuring Aadhaar and Welfare Delivery
What biometric systems reveal, what they miss, and how to make exclusion visible
Two studies, two conclusions
A biometric system can make a public programme more accountable without making it more accessible. That tension sits at the centre of the debate around Aadhaar and welfare delivery.
In Andhra Pradesh, a randomized trial of smartcards found meaningful reductions in leakage from the rural employment programme. Leakage fell by 12.7 percentage points, household earnings rose, and government spending stayed broadly unchanged. The money had not disappeared from the system; it had moved from intermediaries to the households who were meant to receive it.
What the numbers could not see
A later study of authentication in Jharkhand produced a very different result. Authentication alone did not measurably reduce leakage in the food ration system, while collection became more costly for households. A stricter reconciliation process reduced recorded leakage, but it also left between 1.5 and 2 million people without access at some point.
The disagreement is not simply about whether the technology works. It is about what the measurement system is capable of seeing. A fingerprint can confirm who is standing at a counter. It cannot confirm how much grain leaves the shop.
The evidence at a glance
| Finding | Evidence | What it suggests |
|---|---|---|
| NREGS leakage fell 12.7 percentage points | Andhra Pradesh smartcard trial | Verification can reduce intermediary capture |
| NREGS household earnings rose 24% | Same trial | Savings reached households rather than the state |
| Government NREGS spending stayed flat | Same trial | The result was redistribution, not reduced provision |
| Authentication alone had no measurable effect on Jharkhand leakage | Welfare authentication study | Identity verification cannot solve every form of fraud |
| Collection costs rose about 17% | Jharkhand study | Access can become more expensive for households |
| Between 1.5 and 2 million people lost access at some point | Jharkhand reconciliation policy | Tight controls can create large exclusion costs |
| Quantity fraud remained outside the system's view | Field evidence from Jharkhand | A fingerprint verifies a person, not the quantity received |
Sources and further reading
The argument draws on the following published research, interviews, and policy analysis:
- Muralidharan, Niehaus, and Sukhtankar, <strong class='text-white'>Building State Capacity: Evidence from Biometric Smartcards in India</strong>, American Economic Review 106, no. 10 (2016), pp. 2895–2929.
- Muralidharan, Niehaus, and Sukhtankar, <strong class='text-white'>Identity Verification Standards in Welfare Programs: Experimental Evidence from India</strong>, Review of Economics and Statistics 107, no. 2 (2025), pp. 372–392; working paper version NBER WP 26744.
- Drèze, Khalid, Khera, and Somanchi, <strong class='text-white'>Aadhaar and Food Security in Jharkhand: Pain without Gain?</strong>, Economic and Political Weekly 52, no. 50 (2017), pp. 50–60.
- Gelb and Clark, <strong class='text-white'>Performance Lessons from India's Universal Identification Program</strong>, Center for Global Development Policy Paper 020 (2013).
- Khera, <strong class='text-white'>Impact of Aadhaar on Welfare Programmes</strong>, Economic and Political Weekly 52, no. 50 (2017), pp. 61–70.
- Khalid, <strong class='text-white'>Aadhaar, Biometrics, and the PDS in Jharkhand</strong>, Ideas for India (2018).
- The Seen and the Unseen, episode 290, <strong class='text-white'>Karthik Muralidharan Examines the Indian State</strong>.
- Himal Southasian, <strong class='text-white'>Reetika Khera on the Supreme Court Verdict on Aadhaar</strong>.
The figures are presented with their study context because the same technology can produce different outcomes when the programme, incentive structure, rollout, and grievance mechanisms change.
Exclusion has a data problem
Older systems excluded people too: through diverted supplies, short measures, missing cards, and informal refusals. Authentication did not invent exclusion. It changed its character.
When a dealer refuses a household, there is at least a visible decision and a person who can be held responsible. When authentication fails, the transaction may simply disappear from the record. A failed fingerprint is not necessarily proof of eligibility, but it is evidence that the system encountered a problem. If the problem is not logged, it cannot be counted, investigated, or corrected.
Designing a fairer record
The most useful reform is not to choose between technology and human judgment. It is to make the system record both its successes and its failures.
Every failed authentication could be logged with a reason, an available override, and a clear path to grievance redress. Published alongside leakage figures, those records would make exclusion measurable without pretending that every failed attempt is a denial.
The central lesson is simple: a system becomes accountable only when the harm it creates is visible in the same records used to measure its success.