Fighting corruption with India's new national ID system has cost some low-income households their benefits in the process
In 1974, a family declaring itself heir to the historical kingdom of Oudh began a decade-long squatter protest in the VIP waiting room of the New Delhi train station. Claiming she had lost identity documents in a fire, Wilayat “Begum of Oudh” Butt and her two children refused to move until the Indian government restored their property (annexed by the British in 1856), and “unloaded a whole household there: carpets, potted palms, a silver tea set, Nepali servants in livery, glossy Great Danes.”1 Eventually, unable to verify Wilayat’s claim but fearful of sectarian violence by a public who believed her, the Indian government housed the family in a hunting lodge in the heart of Delhi. There they remained, mysterious figures shut off from the world, until each of them eventually died and an astonishing investigative report by the New York Times late last year revealed the barely believable fraud they had perpetuated.
Today Ms. Butt might have a harder time pulling off her caper. India, at the forefront of a global revolution in digital identification, has set up a national unique ID system (‘Aadhaar’) linked to biometric records, with over 1.25 billion IDs generated (91% of the population). Similar systems have proliferated worldwide, with nearly every developing country introducing national identification programmes. Two-thirds of these programmes rely on biometric technology, which allows for verification in settings with low literacy and numeracy (Gelb and Metz 2018).
Identity verification systems and social policy
Even for those of us not claiming royal descent, proving one’s identity is an important exercise, necessary to board airplanes, vote, or – critically for impoverished and marginalised communities – to claim government benefits. Governments are integrating new identity verification mechanisms into their social programmes because they see the potential for good: better targeting, and reduced corruption and fraud. Yet this approach has been contentious, with debate centering around whether requiring biometric authentication will inadvertently exclude genuine beneficiaries due to, for example, missing cards or technical failures.
In India, the integration of Aadhaar into the social welfare system represents one of the most ambitious transformations of social policy in history. The debate over requiring Aadhaar to access benefits has been litigated all the way to the Supreme Court, which in September 2018 permitted the government to mandate the use of Aadhaar for accessing public programmes. Similar debates rage over ID requirements for voting in the US, for example. In the case of India’s Public Distribution System (PDS), which distributes highly subsidised food to the poor, this could quite literally be a matter of life or death. India has the largest number of malnourished people in the world, and the PDS is the flagship programme tackling hunger and food insecurity, costing 1% of GDP. Critics have claimed that some beneficiaries, who were denied their subsidised food because they did not have Aadhaar cards, starved to death,2 and at the very least Aadhaar integration caused “pain without gain” (Dreze et al. 2017). Proponents, on the other hand, have claimed enormous fiscal savings from the combination of Aadhaar and direct benefit transfers to welfare recipients.3
Integrating Aadhaar into the PDS: Experiment and results
We examined the costs and benefits via a large-scale randomised control trial (RCT) that evaluated the impact of integrating Aadhaar into the PDS in the state of Jharkhand, the site of the purported deaths from starvation (Muralidharan et al. 2020). The reform proceeded in two stages across 132 blocks in 10 districts and 15.1 million people.
- The first stage involved the rollout of electronic point-of-sale (ePOS) devices at fair-price shops (FPSs) which enabled the Aadhaar-based biometric authentication (ABBA) of beneficiaries attempting to collect food.
- In the second stage (‘reconciliation’), the government used data from ePOS devices to determine monthly food distribution to the FPSs, by adjusting the amount of new grain disbursed based on electronic records of authenticated transactions. Belying its reputation for low state capacity, the government of Jharkhand implemented the reforms swiftly and thoroughly, while also complying near perfectly with our experimental protocols, with 91% of beneficiaries in treatment areas reporting use of ABBA six to eight months after the start of the reforms.
Matching administrative data on disbursals to nearly 16,000 original household surveys, we found little impacts on average – on corruption or the value of PDS goods received – of the use of ABBA by itself. However, the zero average impacts masked a 10% reduction in benefits for the 23% of beneficiaries who had not linked an Aadhaar card to benefit rolls, with 2.8% receiving no benefits at all. In addition, beneficiaries paid the costs of having to make multiple trips to obtain grains, with an average increase of 17% (Rs. 7) in transaction costs.
Reconciliation, on the other hand, led to big reductions in the value of grain disbursed by the government, received by households, and the difference between the two (‘leakage’). The value of grain disbursed fell by 18% in the control group (36% in treatment), with 22% (34%) of the drop representing reduced value received by beneficiaries and the remaining 78% (66%) representing a drop in leakage. The treatment group declines were larger because dealers in this group had transaction records for a longer period and were expected to have larger stocks of undisbursed grains. Realising that their ability to divert grains was now reduced, FPS owners in treatment areas reported a 72% lower bribe price that they would expect to pay to obtain PDS licenses.
Policy lessons to mitigate tradeoffs
Our results highlight the tradeoffs between lower corruption and higher exclusion. Our experience with the reforms therefore leads us to propose two specific and three general recommendations that may help mitigate these tradeoffs.
Specifically, for the case of the PDS, creating safeguards against exclusion in cases where authentication fails or beneficiaries do not (yet) have Aadhaar links is imperative. In previous work in Andhra Pradesh on the impact of biometric Smartcards for payments on two other welfare programmes, we found no evidence of exclusion – likely a result of offline authentication options as well as back-up payment methods (Muralidharan et al. 2016). Further, our calculations based on the difference between the treatment and control groups in the reconciliation phase suggests that not holding dealers accountable for past diversion – starting reconciliation with a ‘clean slate’ – would have led to most of the gains in reduced leakage without reducing value for beneficiaries.
We also draw three general policy lessons from our 10+ years of work on ID systems and social policy.
- While technology for development generates a lot of excitement, the devil is often in the details. Comparing the cases of Jharkhand and Andhra Pradesh, what mattered was not state capacity, but rather the design of the technology integration. In Andhra Pradesh, the focus was on the beneficiary experience, with built-in safeguards, getting payments closer to the beneficiary, and ensuring payments were timely; leakage reductions were a bonus that were passed on to beneficiaries. In Jharkhand, the priority was fiscal savings – which, of course, can eventually be passed on to beneficiaries – with fast implementation, but this led to exclusion, and eventually a roll-back of the programme.4
- Rigorous, independent, representative evaluations of policy reforms are imperative. In Andhra Pradesh, the highly successful ‘Smartcards programme’ was nearly shut down because of negative anecdotes fed by vested interests to higher-ups. However, our results showing the broad benefits as well as representative, near universal approval of the programme helped to reassure senior policy makers that the programme was beneficial. In Jharkhand, our representative study validated the anecdotes provided by critics of exclusion errors; yet it also showed that leakage did go down, and identified ways to reduce leakage with less pain to beneficiaries.
- Finally, understanding how reforms affect beneficiary outcomes by measuring them directly and at high frequency is essential. In Telangana, we used call centres to call beneficiaries of a large cash transfer programme to ask whether and when they had received their money; monitoring the programme in this manner improved outcomes and was hugely cost-effective (Muralidharan et al. 2019). Easily implementable at scale, this model has the promise to significantly improve last-mile public service delivery.
Editors’ note: This post was published in collaboration with Ideas for India.
References
Gelb, A and A Diofasi Metz (2018), “Identification Revolution: Can Digital ID Be Harnessed for Development?”, Center for Global Development, January.
Dreze, J, N Khalid, R Khera, and A Somanchi (2017), “Pain without Gain? Aadhaar and Food Security in Jharkhand,” Economic and Political Weekly 52(50).
Muralidharan, K, P Niehaus, and S Sukhtankar (2016), “Building State Capacity: Evidence from Biometric Smartcards in India," American Economic Review 106(10): 2895-2929.
Muralidharan, K, P Niehaus, S Sukhtankar, and J Weaver (2019), “Improving Last-Mile Service Delivery using Phone-Based Monitoring", NBER Working Paper 25298.
Muralidharan, K, P Niehaus and S Sukhtankar (2020), "Identity Verification Standards in Welfare Programs: Experimental Evidence from India", NBER Working Paper No. 26744.
Endnotes
1 https://www.nytimes.com/2019/11/22/world/asia/the-jungle-prince-of-delhi.html
2 https://www.thehindu.com/news/national/other-states/death-by-digital-exclusion/article28414768.ece
4 There are interesting and close parallels with the literature on education technology, where there has been a lot of hype about the potential for technology to disrupt and transform education. However, in practice, the results have been much more mixed with high-quality studies finding results ranging from highly positive, to negative (including several with no impact).