A Conundrum of Efficiency And Inclusion: Aadhaar and Fair Price Shops

The deployment of digital technologies in public-service delivery has always been premised on the expectations of transparency, efficiency, and accountability (Sapru and Sapru 2014, Barthwal 2003, Second Administrative Reforms Commission 2008). In the last two decades, most e-governance programmes in India have claimed to have made service delivery more efficient and transparent for the citizens (Garg 2008, Barthwal 2003). Even though such claims have received a fair bit of critical attention from several scholars in terms of its limited impact on actual experiences of service delivery and governance (Mohan et al 2013, Veeraraghavan 2013, Madon 2009), the trend of using digital technologies for public-service delivery continues to flourish in India. India’s digital identity project, Aadhaar, is also framed around similar goals of achieving “efficient, transparent and targeted delivery of subsidies, benefits and services ... through assigning of unique identity numbers” (UIDAI 2017). One of the important use cases of Aadhaar-enabled service delivery is to achieve end-to-end computerisation of public distribution system (PDS) operations. First approved as plan scheme in 2012 under the 12th Five Year Plan and carried forward by subsequent union governments, end-to-end computerisation encompasses (Standing Committee on Food, Consumer Affairs and Public Distribution 2016): digitisation of PDS data for removal of bogus cards and better targeting, computerisation of supply chain to check leakages/ diversion, grievance redressal portal for transparency and accountability, and automation of fair price shops (FPS) to ensure that correct beneficiaries receive their rations as per eligibility.

State governments, each at different stages of implementing the plan scheme, loosely refer to this project as the Aadhaar-enabled public distribution system (AePDS)^[1]. The call for computerisation can be traced back to the Wadhwa Committee report submitted in 2009, which recommended introducing “an end-to-end automated system in the PDS chain...to minimise human intervention in the PDS operations and reduce the scope for manipulation” (HRLN 2009). It is the same line of thinking underlying FPS automation. “Having transparency in the functioning of a FPS is critical for having greater transparency in the overall PDS value chain” (National Informatics Centre 2016). The manual nature of transactions at FPS is understood to lead to “diversions as it is not possible to probe whether actual sale happened at FPS or not” (National Informatics Centre 2016). Hence, end-to-end computerisation of transactions at the FPS was proposed. FPS transactions could then be recorded, transmitted, and beneficiaries could be authenticated with the Unique Identification Authority of India (UIDAI) server to ensure that disbursement happened only to the intended beneficiary (National Informatics Centre 2016).

In this article, we unpack these claims to evaluate the extent to which AePDS systems have altered service-delivery processes at FPS and how they impact beneficiaries. For this, we draw upon our fieldwork in Karnataka and Tamil Nadu (TN). Karnataka has a targeted PDS mechanism with Aadhaar-based Biometric Authentication (ABBA), which verifies the identity of beneficiaries through fingerprint capture. TN has a universal PDS with smart ration cards, which, when swiped, verifies the identity of the beneficiary. Fingerprint capture is required only during the linking of Aadhaar to ration cards (RCs).

Under ABBA in Karnataka, beneficiaries tell their RC number to the FPS owner/system operator, who then enters that into the PDS Portal of Karnataka (colloquially called Ahara),^[2] on a laptop. If it is a valid RC, the page requiring biometric capture will load.^[3] Then the beneficiary will give their fingerprint via a biometric capture machine connected to the laptop (Figure 1). If the capture is accepted, the FPS owner will then be prompted to type into the laptop, the quantity of ration that is being "lifted," that is, disbursed. The FPS owner submits this, thus ending the authentication process for each beneficiary, which is signalled by the flashing of a dialogue box at the top of the web browser. Then the beneficiary approaches the FPS staff in charge of weighing the rations and receives that month’s quota of rations.

In TN, beneficiaries need to produce the smart RC at the FPS. The system operator swipes the credit card-sized smart RC in the point of sale machine. Every time the RC is swiped, the beneficiary’s data, including whether ration for that month has been collected or not, shows up on the screen. If the ration has not been collected, the machine produces a printed receipt with details of beneficiary and quantity distributed. After moving to a second queue, this receipt is handed over by the beneficiary to the FPS worker who weighs and hands over the rations.

Figure 1: A fair price shop in Karnataka. Fingerprint capture machine (indicated by yellow box)

The data is largely in the form of qualitative semi-structured interviews of beneficiaries and our observation of the disbursement process in the ration shops. Additionally, we also interacted with FPS owners (Table 1). The choice of two sites with different approaches to FPS automation and to PDS itself provides an opportunity to examine whether and how these approaches substantially differ with respect to the claims justifying the need for computerisation.

Table 1: Categories of Interviews

Increased turnaround time for each disbursement: Infrastructural constraints concomitant to end-to-end computerisation are well known, such as server issues, network-connectivity issues and power cuts (Allu et al 2019, Bhatnagar 2016, Khera 2017, Saini et al 2017, Somanchi et al 2017). These constraints lead to an increased turnaround time for each disbursement. To then manage the delays and queues, many FPS owners in Karnataka now provide ration distribution in two phases. Authentication, which is an online process, is completed in the first three–five days, successful disbursement is recorded. During the next one–three days, beneficiaries just need to collect rations, making it an entirely offline process, which means they are no longer subjected to infrastructure-related delays. Since they use the same website and server infrastructure, even the food and civil supplies offices face the same issue. To manage the delays and queues, one food office decided to allocate specific days for specific villages (Figure 2).

Figure 2: This notice outside a food and civil supplies office in Karnataka informs beneficiaries of three hobilis (revenue divisions) to come on Tuesday, Thursday, Saturday for the purpose of correcting, updating, and deleting RC.

If a card swipe or fingerprint is rejected after three to five attempts, the beneficiary is told to step aside and try their luck after some time. Multiple authentication attempts combined with the aforementioned infrastructural constraints cause delays and require beneficiaries to make multiple visits to the FPS. One reason for multiple attempts and delays is that fingerprints of beneficiaries are not always machine-readable. This sort of “biometric failure” (Menon 2018) or “fingerprint authentication error” (Somanchi et al 2017) was found in, but was not restricted to, aged beneficiaries, those with prolonged illness, disability or who do hard labour involving their hands (cement factory workers, for instance). Such beneficiaries may require 10–20 attempts before the machine finally authenticates them, and that is not a guarantee. The fingerprints of some beneficiaries are never authenticated at all.

Distinction between identity authentication and ration disbursement: The process design envisions authentication and disbursement to be two separate steps, but in practice, they are performed as one step. In Karnataka, after successful fingerprint authentication, the FPS owner enters the quantity of ration disbursed on the laptop and clicks the "submit" button, which is then recorded as a successful disbursement by the software. The beneficiary then goes to the staff in charge of weighing rations and collects the same. In TN, a receipt is printed after swiping the card that authenticates the beneficiary, who then takes the receipt and stands in another queue where rations are weighed and handed over to then.

In both the cases above, the software records successful disbursement without actual disbursement taking place. So, even if an FPS owner denies rations entirely or gives a lesser quantity of a rations than allotted, the system’s records show that ration has been disbursed as per allotment. This creates opportunities for malpractices at the FPS level. On the one hand, it allows FPS owners to exercise discretion regarding the quantity of grains disbursed, while on the other hand, it helps them maintain official records without any traces of discretion. This problem has been persisting from the pre-computerisation days (Khera 2011). The quantity of rations collected by the beneficiary can depend on their relationship with the FPS owner. We came across beneficiaries of the same village who received varying proportions of their eligible quotas, similar to what Drèze et al (2017) reported from Jharkhand.

Hence, the claims of end-to-end computerisation stand on rather shaky grounds, when it comes to actual practices at the FPS. In fact, they could actually legitimise a false reality as FPS owners can use the electronic records to argue against the allegations of corruption. Furthermore, in the last step before completing the transaction, the FPS owner can type in the quantity of rations being disbursed. The manual nature of transactions leading to diversion of rations justified the need for end-to-end computerisation. Why, then, is this option available at all, and why is it available to FPS owners? Why is there no mechanism to verify what the FPS owner types in?

Perils of expert-centric systems: The software and hardware infrastructure supporting Aadhaar makes it an expert-centric system, making it hard for the last-mile service providers and beneficiaries to navigate the system. For example, the system automatically suspends an RC if rations have not been collected for three consecutive months. This action is automated— the system does not account for any other situation. One of our respondents in Karnataka faced suspension when he missed the disbursement dates for three consecutive months while visiting his pregnant wife at her maternal home in another district. The same could happen due to seasonal migration, illness, calamity, etc. It is not just that the system does not account for these situations, but that local-level intermediaries who are more familiar with these situations do not have the mandate or the expertise to manoeuvre around the system. Therefore, human expertise of FPS owners, who have been doing this work for several years gets overridden by a system relying heavily on technological expertise.

The AePDS, whether targeted or universal, requires the overlaying of information communication technology infrastructure on the existing administrative infrastructure of welfare distribution in different states. This means a complex process integration between identity authentication and welfare disbursement by bringing in heterogeneous institutions and actors into the PDS system. The authentication is done using UIDAI’s Application Programming Interface (API), processed through the PDS server (Figure 3) and verified against the Central Identities Data Repository in the UIDAI server (UIDAI 2017).

Figure 3: Process flow of FPS automation (National Informatics Centre 2016). 

Arrangements like these render the entire system more of a technological black box with a series of “messy” moments for users and ad hoc negotiations by last-mile service delivery personnel (Singh and Jackson 2017). This implies fuzzy boundaries of responsibilities and accountability with PDS. For example, it is unclear, for both beneficiaries and intermediaries (government officials and FPS owners), who should bear the responsibility for authentication failure, and what should be the status of one’s welfare entitlements in the event of repeated authentication failure.

Conclusions

As shown by the examples of beneficiaries getting varying proportions of their eligible quotas, requiring monthly biometric authentication does not prevent leakages or diversion. Successful disbursement is recorded even if intended beneficiary did not receive the commodities as per eligibility. The electronic ledger is thus not an accurate reflection of FPS-level transactions. End-to-end computerisation of transactions at the FPS does not, in practice, make for a transparent system. Computerisation of transactions was supposed to do away with the inefficiencies of paper-based systems. The need for multiple visits and authentication attempts, the increase in turnaround time and concomitant infrastructural issues go against the idea of a seamless, efficient service-delivery system.

After pointing out how a window for leakages opens up due to a flaw in the disbursement process, it seems logical to suggest procedural tweaks that record successful disbursement only after rations have been actually collected by the beneficiary. However, the same flaw also allows FPS owners some leeway in managing the delays by splitting the ration distribution into two phases, as explained earlier. Moving forward, procedural changes and reforms must account for the role of human infrastructure that sustains the welfare apparatus and use of technology in service delivery. Thus, instead of relying on digital technologies as a way to “quick fix” governance issues, policymakers need to reflect on administrative reform in a more nuanced manner.