The National Food Security Act, 2013 made a promise of unusual precision. It did not promise that no Indian would go hungry, which no statute could deliver on its own. It promised something more mechanical and therefore more testable: that identified persons, up to a stated share of the population, would receive a stated quantity of grain at stated prices through a stated network of fair price shops, and that a stated machinery of grievance officers, food commissions and social audits would watch over the delivery. NFSA implementation can be judged against that promise almost clause by clause, and the record that emerges is neither the triumph of the government’s telling nor the failure of the critics’ telling. The shops largely work. The grain largely moves. What does not move is the number at the top of the system: the count of persons the law covers, which was fixed by a census and has been overtaken by the population it was meant to describe.

This article applies to the food security statute the same discipline this series has applied to enforcement statistics elsewhere. In the record of the anti-terror laws, the governing lesson was that an arrest figure and a conviction figure measure different things and cannot be traded for one another, a point set out in the implementation record of the anti-terror laws. In the insolvency cluster, the lesson was that a recovery percentage means nothing until the denominator, admitted claims or liquidation value, is named, as the record of the Insolvency and Bankruptcy Code’s implementation shows in detail. The food security record raises the same problem in a sharper form. Here the contested figures, how many people are excluded, how much grain leaks, how many cards were bogus, all depend on a denominator, and in this statute the denominator itself is the story: coverage is a percentage of a population figure that has not been revised since the census of 2011. Every gap examined in this article is measured against that frozen base, and the largest of them is created by it.

Four quantities recur through everything that follows, and the reader who keeps them separate will not be misled by any of the controversies in this field, because most of those controversies are one quantity masquerading as another. The first is allocation: the grain the Centre assigns to a state in a month, a figure of pure arithmetic on the identified list. The second is offtake: the grain the state actually lifts from the Food Corporation’s godowns against that allocation, which can fall short for reasons of logistics, finance or stocks, and whose shortfall says nothing by itself about what households received. The third is distribution: the sales recorded at the fair price shops, which in the computerised system are the authentication-attached transactions at the point of sale. The fourth is receipt: what households report, to a surveyor months later, that they actually obtained, which is the only one of the four that measures the entitlement as the beneficiary experienced it, and the only one not generated by the system itself. Leakage is the distance between the second and the fourth of these numbers. Exclusion is the distance between the list behind the first and the population outside the building. A government reporting success quotes the first and third; a critic reporting failure quotes the fourth and the population; both are describing the same system at different stations along its chain, and the implementation record, taken whole, is the account of what happens between the stations.

The NFSA implementation architecture the Act required

The entitlement scheme itself, the quantities, the prices, the maternity benefit and the nutrition chapters, belongs to the statute’s own article, the guide to the National Food Security Act 2013, and is compressed here to what the implementation record needs. The Act, Act 20 of 2013, was signed on 10 September 2013 and deemed to have come into force on 5 July 2013. Section 3 entitles every person in a priority household to five kilograms of foodgrains per month at the subsidised prices in Schedule I, rice at three rupees a kilogram, wheat at two rupees and coarse grains at one rupee, and entitles Antyodaya Anna Yojana households to thirty-five kilograms per household per month. Section 3(2) caps the entitled population at up to seventy-five per cent of the rural population and up to fifty per cent of the urban population. Around that entitlement the Act built a delivery and accountability architecture, and it is the architecture, more than the entitlement, that implementation had to construct.

The first piece of the architecture is the coverage machinery of Chapter IV. Section 9 provides that the percentage coverage in rural and urban areas for each state shall be determined by the Central Government, subject to the ceilings in Section 3(2), and that the total number of persons to be covered in a state shall be calculated on the basis of the population estimates as per the census of which the relevant figures have been published. Section 10 obliges each state government, within the number so determined, to frame guidelines and identify the Antyodaya households and the remaining priority households, and its proviso contemplated identification within three hundred and sixty-five days of commencement, with the older targeted distribution continuing in the meanwhile. Section 11 requires the list of identified households to be placed in the public domain. The second piece is the reform programme of Section 12, which directs the Centre and the states to endeavour progressively to reform the distribution system, naming doorstep delivery of grain to fair price shops, the application of information and communication technology including end-to-end computerisation, the leveraging of Aadhaar with biometric information of beneficiaries for proper targeting, full transparency of records, preference to public institutions and women’s collectives in the licensing of shops, diversification of commodities, and schemes of cash transfer or food coupons in prescribed areas. The third piece is the accountability machinery: an internal grievance redressal mechanism under Section 14, a District Grievance Redressal Officer for every district under Section 15 with an appeal lying to the State Food Commission, the Commission itself under Section 16, disclosure of distribution records under Section 27, social audits under Section 28 and vigilance committees under Section 29. Section 8 adds a food security allowance, payable by the state when the entitled grain or meals are not supplied, and Section 22 obliges the Centre to allocate grain from the central pool in accordance with the number of persons identified in each state.

One feature of the blueprint’s drafting explains more of the implementation record than any administrative failing, and it should be registered before the record begins. The Act’s obligations come in two grades. The entitlement itself is hard law: quantities, prices, coverage ceilings and the identification duty are stated as commands, and a beneficiary denied them has, in principle, a claim. The reform programme and much of the accountability machinery are drafted a grade softer: Section 12 obliges the governments to endeavour progressively to undertake the reforms, a formulation that states a direction without a deadline and a duty without a readily enforceable content. Doorstep delivery, computerisation, diversification and cash transfer all live in that softer grade, and so, in their operational detail, do the social audits and vigilance committees. The implementation pattern of the following decade maps onto the drafting grades with uncomfortable fidelity: what the Act commanded was eventually done almost everywhere, late but done; what the Act commended was done where a state government, a court or a crisis supplied the compulsion the text withheld, and discharged as formality where none did. The courts’ repeated interventions, in the drought litigation and the migrant proceedings alike, can be read as the judiciary converting endeavour-language into deadlines case by case, and the unevenness that remains in the record is, in large part, the measure of where that conversion has and has not reached. A statute’s verbs are its first implementation decision, and this statute chose its verbs in a way that guaranteed the shape of everything that followed.

Measured against that blueprint, the physical and digital parts of the architecture were built, late but substantially. The identification of households was the first deadline and the first miss. The Comptroller and Auditor General’s performance audit of preparedness for implementing the Act, Report No. 54 of 2015, submitted in April 2016, found that only eleven of the thirty-six states and Union territories had identified eligible households within the statutory three hundred and sixty-five days, that eighteen states and territories had begun implementation by March 2015, and that only fifty-one per cent of the eligible beneficiaries in the country had been identified by then. The audit also found that, Chhattisgarh, Delhi and Jharkhand apart, most states had not carried out a fresh identification at all: they had re-stamped the old targeted distribution cards as compliant with the new categories, so the new law inherited the old lists with their accumulated errors, and in Karnataka the audit found some 8.90 lakh ration cards identified by the state itself as bogus or ineligible against which grain was still being issued. Rollout continued state by state after the audit period, and the Food Ministry announced in November 2016 that the Act stood implemented in all states and Union territories, Tamil Nadu and Kerala among the last to come on board. The entitlement was therefore nationwide within a little over three years, but the list it ran on was, in most states, the pre-Act list wearing a new stamp, and that inheritance shapes everything the later record shows about exclusion at the identification stage.

The pattern of the rollout contains a lesson about statutory deadlines in a federal system. The states that met the Act’s timetable most easily were, in the main, the states that had the least to change: the universal systems of the south, whose lists already covered their populations and for whom the Act was largely a financing and categorisation exercise, and the reforming states of the centre and east, whose identification and delivery machinery had been rebuilt in the previous decade for their own reasons. The states that missed the deadline by the widest margins were the large states whose targeted systems were weakest, which is to say the states with the most eligible people waiting on the outcome. A uniform statutory deadline, three hundred and sixty-five days for every state regardless of its starting point, thus distributed its failures regressively, falling hardest exactly where the entitlement was most needed, and neither the Act’s design nor the Centre’s monitoring distinguished between a state that was late because it was unwilling and a state that was late because the task, honestly attempted, was larger than the year allowed. The preparedness audit that documented the failure was itself an innovation worth recording: the Comptroller and Auditor General audited the system’s readiness before the system had fully begun, an audit of preparation rather than of expenditure, and its findings set the terms in which the rollout’s failures were understood for years afterwards. Implementation records are usually written from audits of what money bought. This one begins, unusually and usefully, from an audit of what the machinery was when the promise fell due.

Gap one: a coverage ceiling fixed to a stopped census

The centre of the implementation record is an arithmetical fact with legal force. Section 9 ties the total number of covered persons in each state to the population estimates of the census whose figures have been published. The census taken in 2011 remains the last census whose figures have been published; the decennial census due in 2021 was postponed and no later count has replaced it. Applying the Section 3 ceilings to the 2011 population produces the figure the Food Ministry uses as the intended coverage of the Act: 81.35 crore persons, about two-thirds of the population counted in 2011. That figure is a ceiling computed on a fixed base. It does not grow as the population grows, because the statute’s formula points at a published census and no newer census has been published to point at. A child born into a covered household enters the card if the state’s list has room within its ceiling; a household formed after the ceiling filled waits for a deletion to free a place. The number of entitled persons is therefore frozen while the population it was drawn from rises by well over a crore a year.

How many people does that freeze leave outside? The estimates differ because the projections differ, and the honest way to state them is as a range with its makers named. Jean Dreze, Reetika Khera and Meghana Mungikar, projecting the 2011 census forward with state-wise birth and death rates, calculated that on 2020 population figures some 92.2 crore persons ought to have been covered at the statutory percentages, against about 80.9 crore actually covered by food security ration cards, leaving of the order of ten crore people outside the net who would have been inside it on a current count. Dreze, updating the same method, put the figure for 2025 at more than twelve crore, and Khera has described the plausible band across recent years as ten to fourteen crore. Against these, the Food Ministry’s position, stated in Parliament and in its releases, is that the Act’s intended coverage is 81.35 crore persons and that the states have identified close to that number: a Lok Sabha answer recorded 79.40 crore persons identified, about 97.6 per cent of the intended coverage, and the Department of Food and Public Distribution’s year-end release for 2025 put the number actually receiving grain at about eighty crore. Both statements can be true at once, because they use different denominators. The ministry measures performance against the statutory ceiling and finds it nearly full. The researchers measure the ceiling against the living population and find it short by a tenth or more. The gap between the two statements is not a dispute about administration at all; it is the census gap wearing two different descriptions.

The counter-reading deserves its strongest form. A defender of the frozen base points out that two-thirds coverage is generous by international standards, that the Act never undertook to cover a rising share of a rising population between censuses, and that every decennial system has a lag built into it, since a census-based formula is always youngest on the day the census is published and oldest the day before the next one. All of that is true, and none of it meets the point. The lag the draftsman contemplated was a lag of years within a functioning decennial cycle. The postponement of the 2021 census broke the cycle itself, so the base is not merely old but unscheduled for replacement, and the exclusion it produces grows every year without any legislature voting for it and without any administrator erring. The critical use of the exclusion estimate makes the opposite mistake, treating the figure as a deliberate denial of named people by a government that chose to refuse them. It is not a refusal. It is an automatic consequence of a formula operating exactly as written, on a count that has not been refreshed, and the remedies are correspondingly institutional: a new census, or a decision by Parliament or the executive to compute coverage on official population projections instead. That distinction matters for the remedy a reader should demand, and it matters for the verdict at the end of this article.

The courts have been brought to the edge of the question without displacing the formula. In the migrant labourers proceedings described later in this article, the Supreme Court observed in its judgment of 29 June 2021 that the Central Government may undertake the exercise under Section 9 to re-determine the total number of persons to be covered, and in an order of 21 July 2022 the Court directed the Union to look into the matter and to formulate a policy so that benefits under the Act are not restricted to the 2011 census figures, keeping in view that the right to food is a fundamental right. The Union did not re-determine the coverage on projections, and the ceiling remains the 2011 figure. A separate policy current runs the other way: a NITI Aayog discussion paper circulated in 2021 proposed reducing the coverage ratios themselves, from seventy-five and fifty per cent to sixty and forty, which on the projected population of 2020 would have brought the covered number down to about 71.62 crore and saved, on the paper’s estimate, up to Rs 47,229 crore. That proposal was not adopted either. The coverage of the food security law therefore sits where the statute left it, squeezed between an unadopted proposal to shrink the percentages and an unimplemented suggestion to refresh the base, and the squeeze tightens by demography alone.

One paradox inside the ceiling deserves separate statement, because the two halves of it are usually quoted by opposing sides as if only one existed. At the ministry’s own count, about two crore places inside the ceiling stood unfilled: the intended coverage is 81.35 crore persons, the identified number recorded in Parliament is 79.40 crore, and the difference is headroom that any state may fill with eligible households without touching the statute. In states where the ceiling is not the binding constraint, exclusion is therefore not a census problem at all but an identification problem, and the remedy lies in state list-making rather than in New Delhi. In states where the ceiling binds, and Bihar and Uttar Pradesh are the standard examples in the exclusion literature, no amount of list-making admits the household outside until a place is freed or the base moves. A single national exclusion figure conceals which of these two situations a given excluded family is in, and the two situations call for different pressure on different governments. The researchers’ estimate of ten crore or more outside the net and the ministry’s figure of two crore unfilled places are both drawn from the same arithmetic; they describe different margins of the same frozen list.

The fiscal dimension of refreshing the base is arithmetic of the same kind. Coverage is a multiplier on the food subsidy: each additional crore of covered persons carries a monthly entitlement of five kilograms per person, drawn from the central pool at an economic cost to the Food Corporation well above the Schedule I issue price, and the difference is the subsidy. Restoring the statutory percentages on a refreshed population base would therefore raise the annual food subsidy bill roughly in proportion to the added coverage, before any change in prices, and shrinking the percentages on the NITI Aayog pattern would reduce it by the corresponding proportion, which is why the paper could attach a savings figure to its proposal.

A second feature of the frozen base is less often noticed: the freeze applies to the distribution of coverage between states as well as to its total. The seventy-five and fifty per cent figures in Section 3 are national ceilings, and the percentage actually assigned to each state under Section 9 was computed state by state, from the consumption and poverty patterns of the same period, so that poorer states received coverage percentages well above the national average and richer states well below it. Inter-state migration and differential population growth have been moving people between those frozen shares ever since. A state whose population has grown faster than the national rate, or which receives large numbers of in-migrants, finds its assigned percentage describing a smaller fraction of its resident poor each year, while a slow-growing state accumulates slack inside its ceiling. Portability, examined below, relieves the individual migrant’s problem without touching the accounting one: the grain a migrant draws in the destination state is adjusted between the two states’ allocations, but the migrant’s name stays on the home state’s list and in the home state’s count. The census gap is therefore also a federal gap, a slow redistribution of entitlement away from growing and receiving states toward shrinking ones, produced by the same postponement, and it will be corrected, when the count resumes, in the same automatic way it was created, by percentages recomputed on a living base. Until then, the map of coverage in the Act is a map of where India’s population lived in 2011, and every year of delay redraws the real country further from it. The coverage question is usually argued as a question of rights, and it is one; but it is settled, when it is settled, as a question of the subsidy budget, and a reader of the record should expect the percentages to move, if they move, in a budget speech rather than in a judgment. The same arithmetic explains the executive preference for working the existing ceiling harder: verification drives that remove ineligible cards free places for the eligible at no additional subsidy cost, which is why so much of the implementation energy of the past decade has gone into cleaning the list rather than enlarging it.

Gap two: exclusion at the identification stage

Below the ceiling, a second gap opens at the point where states decide which households fill the permitted number. The Act gave that task to state guidelines under Section 10, and most states worked from the Socio-Economic and Caste Census of 2011, applying exclusion and inclusion criteria of their own design to a database that was itself a snapshot. Identification errors run in both directions, and the record documents both. Inclusion error, the presence of the ineligible on the list, is what the CAG found in Karnataka’s unreconciled bogus cards and in government employees identified as beneficiaries in Bihar. Exclusion error, the absence of the eligible, is harder to count precisely because the uncounted leave no administrative trace, but its mechanisms are well documented: households formed or divided after the survey, migrants whose names sit on a card in a village they have left, widows and single persons whom household-based enumeration missed, and families in states whose ceilings filled before their applications were processed. The updating duty in Section 10(2), under which states must refresh the list within the Section 9 number, converts identification from an event into a process, and the deletion drives described below are that process working; but a deletion only helps an excluded household if the freed place is actually reallocated to it, and the waiting lists that many states maintain are not published with the prominence Section 11 requires for the list itself.

Section 11’s transparency duty is the identification system’s intended corrective, and its record is mixed in an instructive way. Digitisation made the list publishable at a scale the paper system never allowed, and the state portals that emerged from the computerisation programme do put the card database online, searchable by card number and village, so that a household can verify its own presence and a researcher can test the list against the population. What the portals do not publish with the same completeness is the queue: the applications pending, the households refused and the reasons, and the places freed by deletion and their reallocation. The asymmetry is characteristic of administrative transparency as practised, which publishes the stock and not the flow. A list can therefore be simultaneously public and unaccountable, every name on it visible to the world, while the decisions that keep a name off it remain inside the office. The exclusion that matters most to the household outside is of the second kind, and it is the kind the Act’s publication duty, read by administrators as a duty to display the list, does not reach; the Swaraj Abhiyan line of cases pressed on precisely this class of gap, between a transparency that displays outcomes and an accountability that must explain decisions.

The deletion drives are the largest single fact in the identification record, and they carry the same double face as everything else in this system. The Food Ministry stated in November 2024 that about 5.8 crore ration cards had been removed from the system through digitisation, de-duplication and Aadhaar-based verification, and its answers in Parliament have since reported annual deletions of 41.99 lakh cards in 2023, 48.85 lakh in 2024 and 41.41 lakh in 2025, as states matched their databases against the records of the income tax department, the transport authorities, the corporate registry and the Aadhaar database. The government’s reading of these numbers is that they measure ghosts exorcised: cards in the names of the dead, the duplicated and the never-existent, whose grain was being drawn by someone else or sold. The critics’ reading, stated most carefully by Dreze and Khera in their rejoinder to the Jharkhand evaluation literature, is that when Jharkhand cancelled cards that had not been linked to Aadhaar, most of the cancelled cards belonged to real households, not ghosts, so deletion drives at speed convert an accuracy exercise into an exclusion event. The two readings are not equally supported for every state and every drive, and the record does not permit a single national split between ghost cards and genuine ones. What it permits is a discipline: a deleted card is evidence that a record failed a check, not evidence of what the household behind the record was. A further category the Department itself flags illustrates the ambiguity, the silent ration card, a card against which no grain has been drawn for a long period. Silence can mean a ghost. It can also mean a migrant household drawing grain elsewhere under portability, a family that buys in the market in good months, or a household defeated by the authentication process described below. Treating silence as proof of ineligibility would manufacture exclusion out of a measurement artefact, which is why the verification exercises built around such cards have paired flagging with a window for the household to establish itself by electronic know-your-customer verification before any final removal.

One more identification channel opened from an unexpected direction. The e-Shram portal, the national database of unorganised workers whose construction the Supreme Court ordered in the migrant proceedings, accumulated registrations in the tens of crores, and comparison with the ration card database showed a large body of registered workers who had no ration card at all. On 19 March 2024, in the continuing migrant proceedings, a bench of Justices Hima Kohli and Ahsanuddin Amanullah directed the states and Union territories to issue ration cards, within two months, to about eight crore e-Shram-registered workers who did not have them, so that they could draw benefits under the Act and other schemes. The direction exposes the ceiling problem in its purest form: eight crore additional cards cannot be fitted inside an 81.35 crore ceiling that is already nearly full unless ineligible cards are removed at the same pace or the ceiling itself gives way, and the implementation of the direction has accordingly proceeded through the same verification-and-deletion machinery, at the speed that machinery allows.

Two features of the identification machinery repay attention because they recur in every state’s record. The first is the design of the criteria themselves. Section 10 left the guidelines to the states, and the standard pattern combines automatic inclusion categories, such as households without shelter and households headed by widows or the disabled, with automatic exclusion categories, such as income tax payers, owners of motorised four-wheelers and government employees. The exclusion categories are proxies, chosen because they can be checked against other databases, and their errors are the errors of proxies: a household that once bought a used car, or in which one member holds a minor government post, can be swept out with the genuinely affluent, while a prosperous household that matches no proxy stays in. The CAG’s finding of government employees drawing benefits in Bihar is the proxy working as designed, a database match catching what a survey missed; the hardship cases in the exclusion literature are the same proxy working on households whose visible markers overstate their means. The second feature is the head of household rule. Section 13 designates the eldest woman of the household, of eighteen years or above, as the head of the household for the purpose of issuing ration cards, a deliberate attempt to put the entitlement document in a woman’s name. Because most states met the Act by re-stamping their old cards rather than issuing new ones, as the CAG recorded, the change of headship the section ordered largely did not happen on the ground: the document kept the name it had always carried, and a provision aimed at the household’s internal standing was defeated by the same administrative shortcut that defeated fresh identification. The provision matters to the record because it shows the implementation gap operating even where no money, no technology and no census stood in the way, only the convenience of the existing card stock.

The poorest fifth: the Antyodaya record within the record

Inside the priority list sits a smaller and older category whose implementation record has its own shape. The Antyodaya Anna Yojana, begun in 2000 for the poorest of the poor, was folded into the Act as its most protected tier: thirty-five kilograms per household per month, a household entitlement rather than a per-person one, for the households the states identify as the poorest within their lists, landless labourers, marginal farmers, artisans, widows and the elderly without support, households headed by the terminally ill or the disabled, and the primitive tribal groups the identification guidelines name. The design contains a quiet inequity that the record bears out: because the Antyodaya quantity is fixed per household, it is generous to a single elderly person living alone, for whom thirty-five kilograms exceeds any plausible consumption, and thin for a household of eight, for whom it works out below the priority per-person norm. The per-person design of the priority tier was the Act’s correction of exactly this defect for the general list, and its retention for the poorest tier is an inheritance the draftsman left standing.

The identification of Antyodaya households reproduced, in miniature, every problem of the larger list: the category’s value made its cards objects of local competition and patronage, state surveys of Antyodaya identification over the years found cards held by households plainly above the category’s definitions alongside poorest households left in the priority tier or off the list entirely, and the re-stamping exercise the CAG documented carried the old Antyodaya errors into the new law undisturbed in most states. The tier also interacts with the ceiling problem in a way worth recording: Antyodaya numbers are counted within the state’s Section 9 total, so every Antyodaya card in ineligible hands consumes a place in the same frozen arithmetic as any other card, while drawing a larger quantity against it. What the Antyodaya record shows, in the end, is the limit of categorisation as an instrument of targeting: each refinement of the category improves the fit between the entitlement and need only as far as the identification machinery can actually distinguish the households, and the machinery’s demonstrated accuracy, in the CAG’s samples and the survey literature, is coarser than the categories drawn on top of it. A system with two tiers honestly administered delivers more to the poorest than a system with five tiers administered at the accuracy the record shows, and the Act’s choice of a protected inner tier inside a broad outer one is best understood as a compromise between those pulls, generosity at the top of need and breadth across the rest.

Portability: the entitlement learns to travel

For most of the ration system’s life, an entitlement was an address. The card bound the household to one fair price shop, and a worker who crossed a district or a state line for employment left the entitlement behind, to be drawn by the family at home or not drawn at all. Section 12 does not name portability, but the reform it mandates, end-to-end computerisation with Aadhaar-based identification, is the machinery portability requires, and the One Nation One Ration Card plan built that machinery into a national platform through the Integrated Management of Public Distribution System. Portability began as a pilot in four states on 9 August 2019, spread state by state through the pandemic period, and was completed in June 2022, when Assam became the thirty-sixth and last state or Union territory to join, making the entitlement in principle drawable at any fair price shop in the country. The Supreme Court’s migrant judgment of 29 June 2021, which directed the states that had not implemented the scheme to do so by 31 July 2021, converted a programme rollout into a judicially watched deadline for the states then outstanding.

What does a migrant worker need to draw grain away from home?

The worker needs the household’s ration card to be digitised and seeded with its members’ Aadhaar numbers, and must authenticate by fingerprint or iris at the electronic point of sale device of any fair price shop. No new card or permission is needed. The family at home draws the balance on the same card.

The national platform was not the first portability in the system’s history, and its prehistory explains its shape. Several states had experimented with portability inside their own borders once their card databases were digitised, allowing a card to be used at any shop in a district or across the state, and the Andhra Pradesh and Telangana experience in particular demonstrated that intra-state portability was administratively unglamorous and heavily used, because most movement for work is short-distance and much of the demand for a second shop comes from households whose assigned shop is distant, irregular or dishonest rather than from households that have left the state. The national scheme generalised that discovery: it is, in engineering terms, a set of shared rails, a central transaction repository and state systems that remain responsible for their own lists, stocks and dealers, knit together by the Integrated Management of Public Distribution System programme and visible to beneficiaries through the Annavitran transaction records and the Mera Ration application. The architecture matters to the record because it locates the failure points correctly. A portable transaction can fail at the card’s seeding, at the device’s connectivity, at the destination shop’s stock, or at the settlement between the two states, and only the first of those is the beneficiary’s problem in any sense she can act on; the rest are the platform’s, and the platform’s published statistics, which count successes, are structurally silent about the distribution of failures among them.

The transaction volumes the platform has recorded are large, and they are the Food Ministry’s figures, to be read as platform counts rather than as counts of migrants fed. By June 2022, when the platform became nationwide, the ministry reported about seventy-one crore portable transactions since the launch, 43.6 crore of them against the statutory entitlement and 27.8 crore against the pandemic free-grain scheme, with about sixty-four crore of the transactions recorded in the Covid period from April 2020; by August 2022 the reported cumulative figure was 77.88 crore, with portability transactions averaging about three crore a month. Two cautions govern the use of these numbers. The first is composition: a portable transaction includes a card used at a different shop in the same town or district, and the great majority of portability is intra-state, much of it intra-district, reflecting ordinary convenience and the correction of bad shop assignments as much as migration. The inter-state stream, the stream the reform was designed for, is smaller by orders of magnitude, though visible where migration runs: the ministry recorded beneficiaries from twenty-one states drawing grain in Delhi in a single month, April 2022. The second caution is the difference between a transaction and an entitlement fulfilled. The platform counts successful authentications. It does not count the worker who was turned away because the shop’s stock was exhausted, whose fingerprints failed, or who never attempted the journey to an unfamiliar shop, and field reporting from migration corridors documents all three. Portability is therefore best described as a genuine and heavily used capability whose benefits concentrate where migration is short-distance and circular, and whose hardest case, the long-distance single migrant in a city, remains the least completely served.

The design detail that matters most to migrant households is the split draw. Because the entitlement attaches to the card and is debited electronically, a part of the household’s grain can be lifted at the destination and the remainder at the home shop within the same month, which matches the actual geography of migrant families, part at the worksite and part in the village. Before portability, that split was impossible in law and common only through informal arrangements with the home dealer. The Mera Ration mobile application, which lets a beneficiary see entitlements, recent transactions and nearby shops, supports the arrangement in the destination language. None of this enlarges the entitlement by a kilogram; it converts an entitlement that migration used to suspend into one that migration merely complicates.

Portability also rewired the economics of the fair price shop, in ways the transaction counts do not show. A dealer’s income depends on the offtake through the shop, since the dealer’s margin is earned per quintal distributed. Under the address-bound system, that income was guaranteed by captivity: the assigned households had nowhere else to go, and the dealer’s incentive to keep them satisfied was correspondingly weak. Portability introduces an element of competition into a system that never had any. A shop near a worksite, a railway station or an industrial estate can draw custom from cardholders assigned elsewhere, and a home shop whose dealer short-weighs or opens irregularly can lose offtake to the shop down the road. The same mechanism creates the reform’s quiet losers: shops in districts of heavy out-migration, whose assigned households increasingly draw at the destination, see their offtake and their margins shrink toward the point where the licence is no longer viable, and a closed shop in a village is an access loss for exactly the households, the old, the non-migrating, who remain. State governments have answered the viability problem in places by raising dealer margins and by consolidating shop areas, and the tension is structural rather than accidental: portability optimises the entitlement for the mobile at some cost to the infrastructure of the stationary, and both populations are entitled. The honest summary is that portability is a reform whose benefits are counted in crores of transactions and whose costs appear in no series at all, surfacing only as a shuttered shop or a longer walk for the household that never migrates.

Computerisation and authentication: the machinery and its two records

The statutory basis for the digital build is Section 12(2), which names end-to-end computerisation to ensure transparent recording of transactions at all levels and to prevent diversion, and the leveraging of Aadhaar, including biometric information, for proper targeting. It is worth being precise about what the text does and does not say. The Act lists Aadhaar leveraging among reforms the governments shall endeavour progressively to undertake; it does not itself make biometric authentication a condition of receiving grain. The compulsion came administratively, through the reform programme, through notifications under the Aadhaar framework, and through state practice, a development traced in the article on the Aadhaar Act 2016. The build itself is not in dispute. Successive answers by the Food Ministry in Parliament describe a ration card and beneficiary database digitised in full across all states and territories, Aadhaar seeding completed for more than 99.7 per cent of ration cards and about 99.26 per cent of beneficiaries on the figures given to the Lok Sabha, electronic point of sale devices installed at 5.50 lakh of the country’s 5.51 lakh fair price shops, and, on the figures given to the Rajya Sabha in March 2026, 99.2 per cent of beneficiaries Aadhaar-seeded with 98.75 per cent of foodgrain distribution carried out through digital authentication, biometric or iris. An earlier ministry statement of November 2024 put 5.33 lakh devices in place and credited the digitisation and de-duplication exercise with the removal of about 5.8 crore ration cards. These are the state’s own series and they are internally dated; they should be cited with their dates and not merged into a single undated claim that the system is digitised.

What does end-to-end computerisation of the PDS actually mean?

It means every stage of the chain is recorded on networked systems: the beneficiary database and ration cards digitised, allocations generated from that database, grain movement tracked from depot to shop, each sale recorded on the shop’s electronic point of sale device against an authenticated beneficiary, and the records visible on public portals for inspection and reconciliation.

Does biometric authentication stop a dealer from giving less grain than the entitlement?

No. Authentication verifies who is collecting grain, not how much grain changes hands. A fingerprint match proves the beneficiary was present; it does not weigh the grain. Short-weighing, the classic quantity fraud of the ration shop, is untouched by identity verification and is addressed, if at all, by transaction records, weighing integration and inspection.

The case for the authentication layer, in its strongest form, is that identity was the binding constraint on the old system’s honesty. A paper system could not distinguish a real household from a fabricated one at any scale, so ghost cards were a structural feature, and the deletion of crores of records after de-duplication, together with the fall in measured leakage examined in the next section, is the return on removing them. The Andhra Pradesh evidence supports the technology family: the randomised evaluation by Karthik Muralidharan, Paul Niehaus and Sandip Sukhtankar of biometric smartcards in the employment guarantee and pension programmes found leakage of payments substantially reduced and genuine beneficiaries receiving more, without an increase in exclusion, in a design that authenticated identity while building in offline and fallback capacity. The case against the layer as actually deployed in the food system is that its errors fall on the entitled. The same three economists, evaluating the later imposition of Aadhaar-based biometric authentication on the Jharkhand distribution system in a randomised design covering some fifteen million beneficiaries, found that the authentication mandate produced no measurable reduction in leakage, raised the transaction cost borne by beneficiaries by about seventeen per cent in repeat journeys and lost time, and left nearly ten per cent of genuine beneficiaries denied grain in the study period through failures of linking or authentication. A household survey by Jean Dreze, Reetika Khera and their co-authors across some one thousand households in thirty-two Jharkhand villages, reported in the Economic and Political Weekly in 2017, found exclusion errors as high as twenty per cent in villages where biometric authentication was required for every transaction, alongside the persistence of the quantity fraud that authentication cannot reach, and the State of Aadhaar Report for 2017-18, drawing on surveys in Andhra Pradesh, Rajasthan and West Bengal, estimated Aadhaar-related exclusion from the distribution system in the millions of persons each month across the states surveyed. Both records are real. They measure different deployments: authentication as one component in a redesigned payment system, with fallbacks, against authentication as a monthly gate at a rural shop with worn fingerprints, intermittent connectivity and a dealer whose cooperation the beneficiary still needs. The verdict the evidence supports is not that authentication failed or succeeded, but that its gains came from the one-time cleaning of the database, which seeding and de-duplication achieved, while its recurring monthly gate at the point of sale has costs that fall on genuine beneficiaries and benefits that the best-designed study could not detect. A system that seeds identity once and authenticates flexibly thereafter captures the first without the second, and the drift of practice, iris authentication, one-time passwords and exception registers for failed authentication, is an unannounced concession that the point is well taken.

Between the database and the shop stands the less discussed half of the computerisation programme, the digitisation of the supply chain itself, and it deserves a paragraph because it is where computerisation touches leakage directly. In the paper system, a consignment of grain passed through a chain of handwritten records: the allocation order from the state to the district, the release order at the Food Corporation godown, the truck challan, the receipt at the block or taluk godown, the issue to the shop and the shop’s stock register, and a diversion anywhere along that chain could be concealed by adjusting the next document. The computerised chain replaces each link with a recorded transaction: allocations are generated from the card database and published, godown issues are entered against them, doorstep delivery is acknowledged by the shop on receipt, and the shop’s opening stock each month is supposed to be the arithmetical consequence of what it received and what its recorded sales say it sold. The point of the architecture is reconciliation: three records that were once kept by three different hands, the godown’s, the transporter’s and the dealer’s, are made to describe the same bags, and a discrepancy among them is an event that can be investigated rather than a rumour that can be denied. States built this chain at different speeds and to different depths, and its completeness varies, but its direction is uniform, and it explains a feature of the leakage record that authentication cannot: the fall in diversion began, in the reforming states, when the supply chain was redesigned and made visible, before any beneficiary was asked for a fingerprint. The fingerprint guards the last metre of the chain. The reconciliation guards the hundreds of kilometres before it.

The reforming states: where delivery works, and why

The Act did not invent its reform template; it nationalised one that a group of states had built and tested in the decade before 2013, and the implementation record reads differently once that parentage is kept in view. The state that supplied the template most completely was Chhattisgarh. Beginning in the mid-2000s, the state took the running of fair price shops away from private licensees and transferred them to cooperatives, panchayats and self-help groups, so that the dealer’s profit motive was replaced by a community institution’s accountability to its own members. It delivered grain directly to the shop’s door, bypassing the intermediate storage and transport layers at which diversion had been easiest, and it computerised the movement of grain from procurement centre to shop, with dispatch information made visible through text messages and public records, so that a consignment that left a godown and never arrived left a visible gap. In December 2012 the state went further than the Centre was yet willing to go and enacted its own food security legislation, the Chhattisgarh Food and Nutrition Security Act, covering some ninety per cent of its population and extending beyond cereals to pulses and iodised salt. The measured result, on the survey-based estimates of Dreze and Khera, was a fall in leakage from fifty-two per cent in 2004-05 to nine per cent in 2011-12, among the lowest in the country, achieved before the national Act existed and therefore evidence of what the template could do in committed hands.

Odisha and Bihar supply the other two panels of the triptych, and they answer different scepticisms. Odisha’s reform, built on expanded coverage and cheap grain, cut leakage on the same estimates from seventy-six per cent to twenty-five per cent over the same period, showing that a poor state with difficult terrain could move the number by administrative effort. Bihar answers the objection that the template only works where it starts from strength: Bihar’s distribution system in 2004-05 was the worst measured in India, with leakage of ninety-one per cent, a system that existed on paper and barely in fact, and its fall to twenty-four per cent by 2011-12, through expanded coverage, computerised coupon and record systems and political attention from the top of the state government, demonstrated that the floor was not destiny. Against these, the older universal systems of the south present a different model altogether. Tamil Nadu and Kerala never adopted targeting in the targeted era, running universal distribution in which every household could draw, financed in part from state revenues; both were among the last states to implement the Act, in November 2016, and both retained coverage wider than the Act’s ceilings at their own cost rather than cut their lists to the statutory pattern. The universal model buys freedom from identification error at the price of subsidising the non-poor, and the targeted model buys fiscal economy at the price of the exclusion errors this article has documented; the Act chose the targeted model, and the states that had refused it complied on their own financial terms.

The lesson the reforming states teach is about the order of operations, because their sequence was the reverse of the national one. In Chhattisgarh, the political decision to make the system work came first: prices were cut and coverage expanded so that most households had a stake in the shops’ honesty, delivery was taken to the doorstep so that diversion required stealing from a visible consignment, management was handed to institutions answerable locally, and computerisation was layered onto a supply chain that had already been redesigned. Where the national rollout after 2013 deployed the technology first and in some states only, computerising records of a delivery chain whose incentives were unchanged, the results were correspondingly thinner, as the authentication record in the previous section shows. The comparative point generalises beyond food. The same decade’s other great Indian delivery success, the cooking gas subsidy reform that moved subsidy into bank accounts nationally, worked because the product was standardised, the beneficiary’s alternative was a functioning market and the transfer replaced a physical chain with a payment; grain distribution has none of those advantages, being a physical chain by design, in places where the market alternative is thinnest. A reader comparing implementation records across Indian welfare law should therefore resist crediting or blaming the instruments, Aadhaar, computers, cash, in the abstract. The record here shows instruments succeeding where an administration had already decided what they were for, and idling where they were installed as a substitute for that decision.

Gap four: leakage between allocation and the household

Leakage is the difference between the grain the state releases for the distribution system and the grain households report receiving from it, and everything about its measurement follows from that definition. The numerator is administrative: the offtake of grain by states from the central pool, published in the Food Ministry’s bulletins. The denominator is estimated: household purchases from fair price shops as reported in the National Sample Survey’s consumption rounds and their successor, the Household Consumption Expenditure Survey. Neither side of the subtraction is a direct observation of theft, and the residual absorbs survey recall error, grain drawn for other welfare schemes, festival and tide-over allocations, and genuine diversion alike. That is why estimates for the same year differ by author and method, and why this article reports the leakage record as several named series that are never merged.

The oldest official benchmark is the Planning Commission’s Programme Evaluation Organisation study of 2005, which concluded that fifty-eight per cent of the subsidised grain issued from the central pool did not reach the intended below-poverty-line families. The survey-based series constructed by Jean Dreze and Reetika Khera put all-India leakage at fifty-four per cent in 2004-05 and 41.7 per cent in 2011-12, with other research groups landing close by for the same years: Bhattacharya and co-authors at 58.6 and 43.1 per cent, and Ashok Gulati and Shweta Saini at 46.7 per cent for 2011-12. The High Level Committee on restructuring the Food Corporation of India, chaired by Shanta Kumar and reporting in January 2015, worked with a leakage figure of about forty-seven per cent for 2011-12 and built large recommendations on it; Dreze and Khera disputed that figure in print, pointing to estimates from the India Human Development Survey that put 2011-12 leakage nearer thirty-two per cent, and the episode is a standing lesson in this field, that a committee’s number is a choice among methods, not a measurement of the world. For the years after the Act, the first large survey is the consumption survey of 2022-23, and Reetika Khera’s estimates from it put leakage at 22.1 per cent for that year, with a separate estimate by Das and co-authors at twenty-eight per cent, and Raghav Puri and Prabhu Pingali, in the Tata-Cornell Institute’s analysis, at 24.1 per cent for 2022-23 and 8.8 per cent for 2023-24. Puri and Pingali attach to their own later figure a caution that belongs in any honest account: the steepness of the final fall coincides with the withdrawal of the pandemic free-grain addition, which had enlarged the administrative numerator, and with features of the survey data that may overstate household receipts, so the 2023-24 estimate should be read as the optimistic end of a range, not as a settled level.

Three conclusions survive the methodological quarrels, and they are the ones this article will defend. The direction is not in serious dispute: on every survey-based series, leakage in the early 2020s is far below its level in the mid-2000s, roughly halving between 2011-12 and 2022-23 on the most conservative reading. The timing matters for attribution: much of the fall predates the Act and tracks the state-level reform wave of the 2000s, when states such as Chhattisgarh, Odisha and Bihar expanded coverage, cut prices, computerised records and took over doorstep delivery, with leakage in Bihar falling, on Khera’s estimates, from ninety-one per cent in 2004-05 to twenty-four per cent in 2011-12, in Chhattisgarh from fifty-two to nine per cent, and in Odisha from seventy-six to twenty-five per cent; the Act generalised the reform template nationally, and the further fall after 2013 is consistent with that generalisation, assisted by the computerisation and authentication build described earlier. And the residual is material by any series: even the lowest recent estimate leaves close to a tenth of the nation’s foodgrain entitlement unaccounted between the depot and the household, and the survey-based consensus leaves closer to a fifth or a quarter. A reader who quotes a single leakage percentage without its author, its survey and its year has not simplified the record; they have falsified it.

Where the residual leakage sits matters for what would reduce it further, and the literature’s decomposition is consistent enough to state. Diversion historically concentrated at two points. The first is above the shop: grain diverted in transport or storage, or recorded as issued to shops that received less, a leakage of the supply chain, addressed by doorstep delivery and the reconciliation architecture described in the computerisation section. The second is at the shop itself: sales recorded against cards whose holders received less or nothing, short-weighing of genuine sales, and the charging of prices above the issue price, none of which any identity technology detects, since the authenticated beneficiary can be, and in the survey record sometimes is, short-changed in the same transaction that authenticates her. A third historical channel largely closed with the Act itself: before 2013, the system’s division into above-poverty-line and below-poverty-line quotas, with different prices and thin above-line allocations that states frequently did not lift in full, generated accounting leakage on a large scale, grain allocated against entitlements nobody was meant to exercise at the offered price; the Act’s broad coverage and single low price structure removed most of that artificial component, which is one reason the measured fall after 2013 cannot be credited to technology alone. The free-grain layer compressed the shop-level channel further by a different mechanism: with the issue price at zero, the margin between what the dealer could charge the beneficiary and what diverted grain fetched in the market narrowed on the legitimate side of the ledger, though diversion for resale into the market remained possible wherever grain itself, rather than margin, was the prize. The policy conclusion the decomposition supports is unfashionable in both camps: the leakage that remains is disproportionately the kind that requires inspection, weighing, audit and prosecution at the level of the individual shop and the individual consignment, the unglamorous enforcement the Act’s vigilance machinery was built for, and not the kind that any further layer of beneficiary-side technology addresses.

The pandemic stress test

Every implementation record contains a moment when the system is loaded beyond its design assumptions, and for the food security system that moment arrived with the lockdown of March 2020, which closed workplaces across the country in a matter of days and set off the largest internal migration in independent India’s history, as workers who had lost their livelihoods in the cities walked or crowded transport home to their villages. The distribution system’s performance in that emergency is the hardest test in this article, and it cuts in two directions at once, which is why it needs its own section rather than a paragraph in someone else’s argument. On the supply side, the machine performed. The Food Corporation’s stocks were ample, the railways moved grain at record pace, and the Food Corporation’s own report of 11 June 2020 recorded some 117 lakh tonnes of foodgrain lifted in the lockdown period and moved in over four thousand rail rakes, with distribution to identified beneficiaries running at double entitlement under the Garib Kalyan scheme described in the next section. A distribution system designed for monthly routine carried a doubled load, in a closed economy, without the stocks or the logistics failing. Whatever the system’s defects, physical incapacity was not among them.

On the entitlement side, the emergency exposed the list as the system’s true frontier, exactly as the gap analysis predicts. The doubled grain flowed to ration card holders, because the entitlement attaches to the card; the workers who most visibly needed food in April 2020 were those whose cards were in villages they had left, or who had no cards at all, and the portability system that would eventually answer their case was, at that moment, a four-state pilot’s national descendant still being rolled out. The executive’s answer for the cardless was a separate scheme under the Atma Nirbhar Bharat package, announced in May 2020: five kilograms of grain per person per month, free, for May and June 2020, to an estimated eight crore migrant workers covered neither by the Act nor by any state card, with an allocation of eight lakh tonnes whose cost of about Rs 3,500 crore the Centre bore. The outcome, in the Food Ministry’s own reporting of September 2020, was that an average of about 2.67 crore persons received grain in each of the two months, against the eight crore estimate, with states initially indicating some 2.8 crore migrants as potentially coverable. The ministry explained the gap by noting that the eight crore figure had been a liberal estimate, set at a tenth of the covered population to avoid under-allocation, and that many workers had returned home, where their home-state cards covered them. Both things are true, and the episode remains instructive: the state could move any quantity of grain, but it could not name the people the grain was for, because no list of them existed, and building one in an emergency reproduced, at speed, every identification error the ordinary system makes slowly. Several state governments, facing the same invisible population, ran parallel distributions of their own for households without cards, from state stocks and state funds, with identification improvised through employers, police stations, resident associations and self-declaration; the quality of those efforts varied with the administrative depth of the state, and they left almost no comparable statistics, which is itself a datum about emergency provision outside the entitlement framework. The litigation that followed, and the Supreme Court’s directions on dry rations, community kitchens, portability in all states and a national database of unorganised workers, are set out in the monitoring section of this article; their administrative residue is the e-Shram portal and the card-issuing direction of 2024, which belong to the identification record. The stress test’s verdict is therefore double, and both halves generalise. A logistics system can be surged. An entitlement list cannot. The strength of a rights-based distribution system in a crisis is exactly the strength of its list on the day the crisis begins, which is an argument, made by events rather than advocacy, for treating the unglamorous work of list maintenance as the core of the implementation enterprise.

The layered entitlement: when the scheme covers the statute

Everything so far has measured the statutory entitlement. What a beneficiary actually paid and received changed, between 2020 and 2023, under layers of scheme that sat on top of the statute without amending it. The statutory prices in Schedule I, three rupees for rice, two for wheat and one for coarse grains, were fixed for the first three years of the Act, after which the Centre could revise them subject to the minimum support price ceiling the Schedule sets. The Centre never revised them; the prices of 2013 simply continued, an early example of the pattern this article keeps meeting, in which the operative position is set by the absence of a decision the statute contemplated.

The Centre’s layer was not the first free grain in the system, and the prehistory matters because it shows the direction of travel before the emergency made it universal among the covered. Tamil Nadu had supplied rice free of charge through its universal system since 2011, treating the issue price as an anachronism in a system whose administrative cost of collecting it was a visible fraction of the sum collected, and other states had cut their effective prices below the Schedule I levels from their own budgets for their own cardholders. The economic argument for the zero price had been available all along: at three rupees a kilogram, the price’s function was never revenue, it was a token of contribution, and its collection imposed costs, on the dealer who accounted for it, on the beneficiary who assembled it, and on the system’s honesty, since a priced entitlement creates the margin inside which short delivery hides. Setting the price to zero removed the token and the margin together. What the state-level precedents could not show, because their scale was bounded by state budgets, was whether a zero price across eighty crore people would change the politics of coverage itself; the years since 2023 supply the answer recorded earlier in this article, that generosity within the list and silence about the list’s edge are purchased with the same money.

In March 2020, as part of the pandemic relief package, the government announced the Pradhan Mantri Garib Kalyan Anna Yojana, under which every person covered by the Act received an additional five kilograms of grain free of cost each month, over and above the statutory entitlement bought at Schedule I prices, with a kilogram of pulses per household in the first phase. The scheme ran in seven phases between April 2020 and December 2022, with a pause in the winter of 2020-21; the Department of Food and Public Distribution’s year-end release for 2025 records about 1,118 lakh metric tonnes of grain allocated across those twenty-eight months, with a planned financial outlay of about Rs 3.91 lakh crore. When the last phase ended, a Union Cabinet decision of December 2022 folded the free element into the statutory distribution itself: from 1 January 2023, the full entitlement under the Act, Antyodaya and priority alike, was supplied free of cost for one year under the Garib Kalyan name, and the arrangement was then extended for five years from 1 January 2024, at an estimated outlay of Rs 11.80 lakh crore recorded in the Food Ministry’s answers in Parliament, which carries it to December 2028. The quantities a household receives and the price it pays are therefore, for the period of the decision, more generous than the statute: the full statutory quantity at a price of zero.

A second layer, quieter than the price layer, changed what is in the bag rather than what it costs. From 2021 the government moved to supply fortified rice through the public distribution system and the other food schemes, rice kernels blended with iron, folic acid and vitamin B12 at the mill, rolled out in phases with the stated aim of covering the distribution system in full. The intervention illustrates the layered method in its purest form: the Act says nothing about fortification, no amendment was sought, and the nutritional content of a statutory entitlement was altered for eighty crore people by executive decision and procurement specification. Supporters within the public health establishment presented the layer as the distribution system’s scale finally put to nutritional use; critics among nutrition researchers questioned the evidence for the fortificant’s benefit at population scale and the safety of unmonitored iron dosing for particular groups, a disagreement the implementation record can report without resolving. What is not in dispute is the institutional fact: a beneficiary’s legal entitlement is to five kilograms of foodgrains, and what those grains contain is, like their price, a matter of the layer rather than the law.

The distinction between the statute and the layer is not pedantry, and the implementation record turns on it. The Act’s prices remain in Schedule I, unrevised; the zero price rests on executive decisions taken year to year and term to term, and what an executive decision gives, an executive decision can end, without Parliament amending anything and without any beneficiary acquiring a right that survives the decision’s expiry. The same is true in miniature of the additional quantities that states supply from their own budgets, and of the scheme grain that flowed to persons without cards during the pandemic under separate announcements. A beneficiary reading the Act learns what the law guarantees. A beneficiary reading the ration shop learns what the layering provides under the decisions in force. The two have diverged before, when scheme grain stopped in the pause between phases, and the divergence is the standing risk of governing a statutory entitlement by scheme. It also cuts the other way politically: the free layer has made the frozen coverage ceiling easier to defend, because the argument that the covered receive everything free blunts the argument that the uncovered have grown in number. Generosity to the included and exclusion of the remainder are, in this system, two faces of one budget, and an honest assessment has to hold both in view at once.

The cash alternative: recommended, piloted, and left at the margin

The most ambitious attempt to redesign delivery came from within the state’s own machinery. The High Level Committee chaired by Shanta Kumar, constituted in August 2014 to restructure the Food Corporation of India, reported in January 2015 with recommendations that went well beyond its subject: coverage under the Act should be brought down from sixty-seven per cent of the population to forty per cent, the covered should receive seven kilograms per person instead of five at prices linked to half the minimum support price, procurement should be handed to the experienced states, stocking outsourced, and cash transfers should replace grain distribution beginning in cities with a population above one million. The coverage and price recommendations were not accepted, and the Act’s coverage has never been legislatively reduced; the committee’s durable legacy is the parts of its programme that were already in Section 12, end-to-end computerisation and cash transfer pilots, pursued as administration rather than as redesign.

The committee’s own subject, the Food Corporation, deserves its paragraph in any implementation record, because the distribution system’s performance rests on the Corporation’s balance sheet and godowns. The committee found a corporation procuring far beyond the distribution system’s requirement in the surplus states, holding stocks well above the buffer norms, and carrying the excess at an economic cost inflated by old storage, cover-and-plinth stacking in the open, and a workforce and depot structure built for a different era; its remedies ran to handing procurement to the experienced states and confining the Corporation to deficit regions and buffer management, outsourcing storage to private and public warehousing agencies, building modern silos in place of conventional godowns, and moving grain increasingly by container and rail rake rather than by the bag-and-truck chain. The record of adoption is the familiar one of this article: the direction was accepted in principle and implemented in part and slowly, silo capacity was contracted out in tranches over the following years, decentralised procurement expanded state by state, and the Corporation’s stocks continued to swing with harvests and politics rather than with any norm. For the food security entitlement, the consequence is indirect but real: every rupee the Corporation spends storing grain the system does not need is a rupee of food subsidy that buys no household any grain, and the subsidy figures quoted in coverage debates inherit that inefficiency silently. The committee’s deepest point was accounting, that the subsidy should be judged by what reaches the household per rupee spent, a measure on which the system’s improvement over the decade, as leakage fell, is genuine and under-credited.

The pilots were real and their record is instructive. Under the Cash Transfer of Food Subsidy Rules, 2015, made under the Act, the food subsidy was paid in cash into bank accounts in lieu of grain in Chandigarh and Puducherry from September 2015 and in the urban areas of Dadra and Nagar Haveli from March 2016, the transfer amount fixed by reference to one and a quarter times the minimum support price. MicroSave Consulting’s baseline assessment, conducted in August 2015 before the first transfers, found the amount already contentious, since one and a quarter times the procurement price does not track the retail price a household must actually pay, and found market infrastructure in Dadra and Nagar Haveli too thin for the experiment, on which the rural part of that pilot was dropped. NITI Aayog’s evaluation of the three territories in 2017 found that on average about two-thirds of beneficiaries were receiving the transfers, that roughly a fifth reported periods of non-receipt, that beneficiaries spent more time and money reaching banks and markets than the shop had cost them, and that preference for the cash mode grew over time as the machinery settled. The pilots were never extended to a state, and the urban areas of Dadra and Nagar Haveli remain, on the Food Ministry’s answers in Parliament, the standing exception where the food subsidy runs in cash mode within a grain system. The reasons for the stall are written across the record of the other sections. A cash system dissolves the fair price shop network, and with it the dealer constituency and the state supply administrations built around physical grain; it dissolves the Food Corporation’s distribution role while leaving its procurement role politically untouchable in the surplus states, so the reform threatened the delivery half of the grain economy while sparing the half with the strongest defenders; and it transferred the price risk of the food market from the state to the household, a transfer whose fairness depends entirely on indexation the pilot design did not convincingly supply. Against those obstacles, the evaluation’s findings, workable machinery, mixed beneficiary experience, growing acceptance, were enough to keep the option alive in the statute’s text and in committee recommendations, and not enough to move a single state government whose voters collect grain monthly and judge it in the hand. The cash question in Indian food policy is therefore best read as settled for the present by revealed institutional preference rather than by evidence: the system’s actors, given a tested alternative, chose the grain. The lesson the record supports is narrow and useful: cash transfer of the food subsidy is administratively possible and fails, where it fails, on amount adequacy against retail prices and on the thinness of banking and markets in poor areas, not on the concept; and a reform whose strongest evidence comes from two small urban territories has not earned generalisation to the countryside on that evidence alone.

The household’s month: the entitlement as experienced

The statistics of the previous sections describe the system from above. From below, at the level of the household, the entitlement is experienced as a monthly sequence of small negotiations, and the survey literature that has watched households through that sequence supplies a texture the aggregate numbers cannot. The month begins with information: whether the grain has reached the shop, on which days it will be distributed, and whether the household’s card will be served, knowledge that in the paper era travelled by word of mouth and in the computerised era arrives, unevenly, by text message and application. It proceeds through the queue at the shop, where the costs the Jharkhand evaluation measured in rupees and percentages are experienced as journeys repeated when the machine fails or the stock runs out, as wages forgone for the day spent, and as dependence on the dealer’s goodwill for the exceptions, the failed fingerprint, the missing member, the card left in the village, on which the formal entitlement silently depends. It ends with the quantity in the bag and the quality of the grain, the two things authentication does not measure: household surveys in the reforming states recorded, over the reform decade, a marked rise in the share of households reporting that they received their full quota, the single most consistent finding of the ground-level literature, alongside persistent complaints about grain quality, the mixing of inferior stock, and the gap between the grain the system issues and the grain the household would choose.

Three features of the experienced system deserve record because they explain behaviour the administrative data treats as noise. The first is partial lifting: households do not always draw their full entitlement, drawing less in months when their own harvest or wages are adequate, when the grain on offer is poor, or when the journey costs more than the saving, and the system’s administrators have repeatedly misread this rational behaviour as evidence that the cards are ghosts, most consequentially in the silent card exercises examined earlier. The second is the household’s internal allocation of the collection burden, which falls overwhelmingly on women, so that the time costs of authentication failure, repeat journeys and queueing are in practice a tax on women’s labour, a distribution of the system’s friction that the transaction-cost studies measure in aggregate and the design of Section 13, putting the card in a woman’s name, acknowledged without remedying. The third is the relationship with the dealer, which the reform literature treats as a principal-agent problem and the household experiences as a dependency to be managed: the dealer is at once the system’s most likely point of fraud and the household’s only source of flexibility, credit against the next month’s grain, a portion drawn early, a failed authentication overlooked, and households protect the relationship even when they know it costs them grain, which is why complaint rates against dealers run far below the survey-measured incidence of the conduct complained of. Any reading of the grievance statistics in the next section should carry that discount: a district reporting few complaints is reporting the price of complaining as much as the quality of delivery.

The monitoring machinery: built late, staffed thinly

The accountability chapters of the Act were the slowest to be built, and the record of their construction is largely a record of judicial prompting. The design is complete on paper: an internal grievance mechanism in every state under Section 14, a District Grievance Redressal Officer in every district under Section 15, empowered to hear complaints about non-distribution of entitled grain and to enforce entitlements, with an appeal to the State Food Commission; the Commission under Section 16, a chairperson, five members and a member-secretary, charged with monitoring the Act, inquiring into violations of entitlement on complaint or of its own motion, advising the government and laying annual reports before the state legislature, armed under Section 20 with the powers of a civil court for summoning witnesses and documents; a statutory alternative under Section 18 by which a state may designate another statutory commission or body to discharge the Food Commission’s functions; social audits under Section 28; vigilance committees under Section 29; and the food security allowance under Section 8 as the financial sanction when the grain does not arrive.

What the states initially did with this design was, in the main, to designate rather than to build. District officers were designated from the existing district administration, carrying the grievance function as an additional charge on top of the supply administration whose conduct was the subject of the grievances, and the Food Commission’s functions were widely handed under Section 18 to an existing body, a human rights commission, a women’s commission or a consumer commission, as an extra portfolio. The Supreme Court took the matter up in Swaraj Abhiyan v. Union of India, the drought litigation. In directions issued in 2016 the Court required the states before it to put the internal grievance mechanism and district officers in place and to constitute State Food Commissions, and in its judgment of 21 July 2017, delivered by Justice Madan B. Lokur, the Court went further: district officers should be independent officials rather than officers of the implementing machinery, a separate State Food Commission should be constituted, with designation of another commission acceptable only when unavoidable and as a last resort, social audit machinery should be established, vigilance committees actually formed, and the Centre should use its direction power, under Section 38 of the Act and Article 256 of the Constitution, to bring defaulting states into compliance. The judgment recorded the Court’s dismay at the position taken by Haryana, whose affidavit had suggested there was hardly any work for a State Food Commission to do, in a state running a full food security programme.

The humblest element of the design, the vigilance committee, repays a paragraph because it is the element whose failure is most complete and least lamented. Section 29 contemplates committees at the level of the state, the district, the block and the individual fair price shop, composed with prescribed representation of the scheduled castes, the scheduled tribes and women, charged with supervising the distribution in their area and reporting violations to the grievance officer. The fair price shop committee was the design’s point of contact with the household’s month described earlier: the body before which the dealer’s register could be challenged by the people in the queue, without a journey to the district headquarters or a written complaint. In most states the committees were constituted on paper in the first flush of compliance, met rarely if at all, kept no published minutes, and lapsed without remark; the social audit, where an external agency or a mobilised community performed the same function episodically, supplied in practice what the standing committee was meant to supply continuously. The pattern generalises across the monitoring chapters and is the deepest of the implementation findings in this article. The Act’s accountability design assumed a density of local institutional life, committees, hearings, published lists, that the administrative system was willing to legislate and unwilling to animate, and the functions survived, where they survived, by being taken up intermittently by courts, auditors and researchers rather than by the standing bodies built for them. A reader assessing any Indian welfare statute’s accountability chapters should ask the question this record teaches: not whether the body exists in the Act, but whether anyone has ever seen it meet. The older right to food litigation, People’s Union for Civil Liberties v. Union of India, begun in 2001, had taken the complementary route of appointing its own Commissioners to report on compliance, an extra-statutory monitor for a pre-statutory system; the Act absorbed the monitoring function into law, and the Swaraj Abhiyan directions were needed because absorption into law had not produced absorption into administration.

What is a social audit under the food security law?

A social audit is a collective public verification of the distribution record: the beneficiary list, the shop’s stock and sale registers are read out and checked in a public hearing, so ghost entries, false records and short deliveries surface before the people they affect. Section 28 obliges states to conduct periodic social audits and publish the findings.

Behind the statutory machinery stands an older, judicial one, because the Act itself grew out of litigation. The right to food case begun by the People’s Union for Civil Liberties in 2001, in the shadow of starvation deaths in Rajasthan, ran for years as a continuing mandamus in which the Supreme Court converted the existing food schemes into enforceable entitlements by interim order and appointed Commissioners, among them the economist N. C. Saxena, whose office reported on the state of the schemes on the ground and supplied the Court with an evidence base independent of the governments being monitored. The Commissioners’ reports documented, scheme by scheme and state by state, the distance between entitlement on paper and grain in hand, and much of the Act’s design, its grievance officers, its commissions, its social audits and its vigilance committees, is that litigation’s agenda written into statute: the monitoring chapters are the Commissioners’ office, institutionalised and handed to the states. The hand-over is where the design weakened. The Commissioners answered to the Court and reported in public; their statutory successors answer to the state governments whose performance they monitor, are appointed by those governments, and in the designation states are other bodies altogether. The migrant proceedings of 2020 to 2024 showed the older machinery reasserting itself when the newer one did not bite: it took a suo motu case and repeated Supreme Court directions to move portability deadlines, card issuance and the worker database. A monitoring system that requires a constitutional court’s continuing attention to perform its routine functions has not failed exactly, but it has not yet become what the Act drew.

The honest summary of the monitoring record is that the forms exist almost everywhere and the substance varies from thin to real. Commissions have been constituted in the states, annual reports are laid in some legislatures, and grievance channels, call centres, online portals and mobile applications, exist in profusion; but commissions constituted by designation, or staffed on additional charge, or left without the inquiry staff Section 16 contemplates, monitor in the way their construction predicts. The food security allowance illustrates the gap in miniature: it is the Act’s own remedy for non-supply, payable by the state when the entitlement does not arrive, and there is no published series of allowance payments at any scale, because the machinery that would establish non-supply, certify it and pay against it is the same machinery whose weakness causes the non-supply. A household pursuing its own card, complaint or allowance should treat the general position described here as information rather than advice on its case, since the procedure, the forum’s practice and the state rules differ and limitation periods in grievance and appeal are unforgiving. The monitoring chapters are best read as a standing rebuke the statute addresses to its own administrators: Parliament anticipated that the delivery system would fail in identifiable ways and built the correctives in advance, and the implementation question has never been whether the design exists but whether the correctives would be staffed as institutions or discharged as formalities.

The supply side: allocation, procurement and stocks

Delivery also depends on grain existing in the central pool and reaching the state, and the allocation mechanics are the hinge between the food security statute and the procurement system. Section 22 obliges the Centre to allocate to each state the quantity its identified population requires at the Schedule I prices, and the monthly allocation follows the card database: so many identified persons, so many kilograms, adjusted for the Antyodaya households drawing thirty-five kilograms per household.

How is the foodgrain allocation for a state worked out?

The allocation is arithmetic on the identified list: the state’s identified priority population multiplied by five kilograms per person per month, plus thirty-five kilograms for each Antyodaya household, allocated as rice and wheat from the central pool in the state’s consumption ratio, at Schedule I prices and, in the free-grain period, at no charge.

Behind the allocation stands procurement at the minimum support price and the buffer stock, the subject of the article on the support price and procurement framework, and the implementation record on this side is comparatively strong: procurement and stocks were sufficient through the pandemic expansion to supply both the statutory entitlement and the free additional layer for twenty-eight months, a logistical fact that the system’s critics and defenders alike have to accommodate. The strains appear at the edges of the machine rather than its centre, in the storage and movement findings of the CAG’s preparedness audit, which recorded godown construction lagging sanctions and railway rake availability falling short of requirement by thirteen to eighteen per cent in the years it examined, and in the recurring state-level shortfalls when allocation, lifting and shop-level stock fall out of alignment in a particular month. A beneficiary experiences those strains as the shop that has not received its consignment, which is why the grievance machinery matters; the aggregate record, though, is that the Indian state has become reliable at moving very large quantities of grain to half a million shops every month, and unreliable mainly at the margins of who is on the list, how they prove it, and what happens when something goes wrong.

The central pool that feeds the distribution system also feeds the state’s other grain commitments, and the sharing matters to the measurement questions this article keeps returning to. The midday meal programme in schools and the supplementary nutrition programme for young children draw foodgrain from the same procurement and storage system under their own allocations, as do welfare institutions and the calamity and festival allocations states request from time to time. The administrative statistics of the pool therefore describe a family of schemes, not one, and the leakage estimates constructed by comparing offtake with household reports have to net out the grain that lawfully left the pool for a school kitchen before any residual can be charged to the ration shops. The careful estimators do this, with varying success, and it is one of the quieter reasons their figures for the same year differ: the boundary of the system being measured is itself a judgement. The sharing has a delivery consequence as well. The school meal and the ration shop are, for a poor household, parts of a single food security experience, and failures correlate: the district whose supply chain cannot keep its shops stocked is rarely the district whose schools serve the meal reliably, because both draw on the same district administration, the same transport contracts and the same godowns. The Act’s nutrition chapters have their own coverage elsewhere in this series; they enter this record only to mark that the pool, the chain and the district are common property of all the entitlements at once, and the implementation of each is evidence about the others.

Between allocation and the shop sits one more measured quantity that the public debate routinely confuses with delivery: offtake, the grain states actually lift against their allocations. Offtake of the statutory allocation runs high in the aggregate, because the entitlement is a standing claim that states dishonour at political cost, but the margins of the system show the slippage plainly: additional and tide-over allocations announced for festivals, calamities and special categories are lifted unevenly and sometimes barely at all, and in the years before the Act the above-poverty-line quota was lifted so partially by so many states that its nominal coverage overstated the operating system by a wide margin. The distinction matters for reading announcements. An allocation is a permission, expressed in tonnes; an offtake figure is an event; a distribution figure is a claim about shops; and a household’s receipt is the only term of the four that the entitlement itself promises. Much of the confusion in the food security debate, and not a little of its bad faith, consists of quoting the largest available number in that sequence as if it were the last one, and the discipline this article has applied, naming which quantity each figure represents and whose count it is, dissolves a good part of the apparent disagreement between the system’s defenders and its critics before any question of values is reached.

The states pay too: the federal shape of the record

The food security system is a central statute delivered by state machinery on shared money, and its implementation record cannot be closed without the fiscal federalism that shapes both. The Centre’s part is the grain itself: procurement at the support price, storage and movement to the states, and the subsidy that covers the distance between the economic cost of that grain and the issue price, or, in the free layer, the whole cost. The states’ part is everything after the state godown: intra-state transport, handling, the fair price shop dealer’s margin, and the administration of identification, distribution and grievance. The Act contemplated central assistance to the states for the intra-state leg on norms the Centre prescribes, and the adequacy of those norms, against hilly terrain, dispersed habitations and long supply lines in the north-east, is a standing subject of state complaint, recorded in the readiness findings of the early audits. Where the norms fall short, the difference is absorbed in the ways this article has documented: thinner dealer margins and the shop viability problem, slower doorstep delivery, and monitoring staffed as an additional charge.

On top of the shared structure, the states run layers of their own, and the record of the layers is the natural experiment in what voters reward. Drèze, Khera and Mungikar counted, alongside the statutory list, some ninety million people covered in 2020 by state-issued cards carrying entitlements equivalent to or better than the Act’s, financed from state budgets: states covering households beyond the ceiling, states issuing their own cards to the ceiling’s overflow, and states adding commodities, pulses, edible oil, salt, sugar, to the statutory cereals, in the pattern Chhattisgarh’s own food security legislation set before the national Act. The political economy is legible. Grain delivery is the most visible thing an Indian state government does for the poor, its failures are felt monthly and its successes are attributable, and incumbent governments that visibly repaired their distribution systems were, in the reform decade, repeatedly returned to office on the strength of it. The centre-state structure also explains the record’s characteristic unevenness: a central statute can fix quantities, prices and percentages for the whole country, but it cannot fix the administrative attention of thirty-six governments, and the distance between Bihar’s trajectory and a laggard state’s is a distance between governments, not between laws. That is the final structural lesson of the implementation record, and it applies to Indian welfare statutes generally: the Act nationalised the entitlement, but delivery remained a state subject in everything but name, and the entitlement’s real content in any given district is the product of both facts at once.

The four gaps

The table below is this article’s findable artifact. It sets the four delivery gaps side by side on the same discipline: what the gap is, how it is measured, the range of published estimates with their makers, what causes it, and what would close it. It is designed to be usable by readers arguing opposite conclusions, because the figures are given with their methods attached.

Gap	What the gap is	How it is measured	Published estimates, with sources	What causes it	What would close it
The census base and coverage	Persons who would qualify at the statutory percentages on the present population but fall outside the ceiling computed on the census of 2011	Statutory percentages applied to population projections, compared with the covered number on ration cards	Ceiling of 81.35 crore persons on the 2011 base (Food Ministry); about ten crore excluded on 2020 projections (Dreze, Khera and Mungikar); more than twelve crore by 2025 on the same method (Dreze); a ten to fourteen crore band across recent years (Khera)	Section 9 ties coverage to the last published census; the decennial census due in 2021 was postponed and no successor count has been published	A new census, or a decision to compute coverage on official population projections, as the Supreme Court invited in 2021 and 2022
Exclusion at identification	Eligible households omitted from the state list within the ceiling, and ineligible households retained on it	Household surveys against the card database; audit comparison of lists with eligibility criteria	Only fifty-one per cent of eligible beneficiaries identified by March 2015, most states re-stamping old cards (CAG Report No. 54 of 2015); about 5.8 crore cards deleted in the digitisation and de-duplication drives (Food Ministry, November 2024)	Identification run once, on the Socio-Economic and Caste Census of 2011, by state criteria of varying quality; lists updated by deletion more energetically than by addition	Continuous enrolment with published waiting lists, periodic re-survey, and reallocation of freed places as Section 10(2) contemplates
Exclusion at authentication	Genuine beneficiaries unable to complete Aadhaar seeding or biometric authentication at the point of sale	Randomised evaluation and household surveys at fair price shops; administrative failure data, which is not published as a series	No measurable leakage gain and nearly ten per cent of genuine beneficiaries denied in the study period (Muralidharan, Niehaus and Sukhtankar, Jharkhand evaluation); exclusion errors as high as twenty per cent in villages requiring authentication for every sale (Dreze, Khera and co-authors, 2017)	Worn fingerprints among manual workers and the elderly, connectivity failure, seeding errors, and a monthly gate administered by the dealer the system distrusts	One-time seeding with flexible authentication thereafter, working exception registers, iris and one-time password fallbacks, and publication of failure rates
Leakage in delivery	Grain released for distribution that households do not report receiving	Administrative offtake compared with household consumption reported in national surveys	Fifty-four per cent in 2004-05 and 41.7 per cent in 2011-12 (Dreze and Khera); about forty-seven per cent for 2011-12 on the Shanta Kumar Committee’s method; 22.1 per cent for 2022-23 (Khera), twenty-eight per cent (Das and co-authors), 24.1 per cent for 2022-23 and 8.8 per cent for 2023-24 with stated data cautions (Puri and Pingali)	Diversion in transport and storage, false recording at the shop, short-weighing, and the arbitrage between the issue price and the market price	Doorstep delivery, transaction recording, stock reconciliation and inspection; compressed further when the issue price fell to zero, removing part of the arbitrage

Reading the record: five rules for using these numbers

The statistics in this field are quoted badly more often than they are contested honestly, and a reader who works with them, in a newsroom, a legislature or a court filing, can avoid most of the standard errors by holding to five rules that the record itself teaches. The first rule is to name the denominator before quoting any coverage figure: against the statutory ceiling, the system is nearly full; against the living population, it is short by a tenth or more; and a sentence that does not say which comparison it is making is not a fact but a posture. The second rule is to quote leakage only with its maker, its survey and its year attached, because the figure is a residual between an administrative numerator and a survey denominator, and different constructions of the same year have differed by fifteen percentage points; the honest range for the early 2020s runs from under ten per cent to over a quarter, and choosing one end of it silently is advocacy. The third rule is to refuse spliced series: administrative counts whose definitions changed mid-stream, seeding percentages measured on cards in one answer and on beneficiaries in another, transaction counts accumulated over different periods, cannot be joined into a single rising or falling line, and the ministry’s own figures in this article have been kept in their separate dated sets for exactly that reason. The fourth rule is to distinguish announced targets from achieved counts with special care in the emergency schemes, where the eight crore estimate for the cardless migrant scheme and the 2.67 crore average monthly recipients are both official numbers describing different things, the size of the provision and the size of the delivery. The fifth rule is to check, for any statement about price or quantity, whether the statement is about the statute or about the executive layer on top of it, because the two diverge: the Act charges three rupees for rice and the layer charges nothing, and a report that attributes the zero price to the Act, or the Act’s prices to the shop counter, has mis-stated the law in one direction or the generosity in the other.

These disciplines are easier to apply when the underlying material is assembled in comparable form, and the research practice in this field has adapted accordingly: scholars and journalists who track the estimates maintain the committee reports, evaluation studies and survey presentations as slide decks and data appendices, circulating them for comment before publication, and a reader who receives such material in presentation form can open it directly in a browser-based PowerPoint viewer for committee and survey decks to check a quoted figure against the table it came from. The habit matters more than the tool. Every contested number in this article became contested at the point where it was separated from its table, its method note and its date, and travelled onward as a bare percentage. Keeping the figure attached to its source is not pedantry; in a field where a single percentage point of leakage is several lakh tonnes of grain, it is most of what accuracy consists in.

Verdict

The honest verdict on this record has two limbs, and neither can be amputated without falsifying the whole. On delivery, the food security system is a substantially reformed machine: leakage has roughly halved on the most conservative survey series and fallen further on others, the entitlement travels with the beneficiary in a way the pre-Act system never attempted, the lists have been digitised and repeatedly cleaned, and the supply side held through the largest free distribution exercise in the country’s history. A verdict that calls the Act a failure has to explain away that record, and cannot. On entitlement, the system is frozen at its point of greatest consequence: the number of persons the law covers was fixed by a census that the decennial cycle should have replaced years ago, the percentages have been neither raised to meet the population nor lowered as the NITI Aayog paper proposed, and the monitoring institutions that were supposed to police the boundary between the list and the population were staffed as formalities until the courts intervened, and thinly after that. The statute is indexed to a census that stopped. That is the namable claim of this article, and it converts the loudest argument in Indian food policy, the argument about exclusion, into its correct institutional form: the largest single cause of exclusion from India’s largest entitlement is not a decision about who deserves grain, and not theft of grain, but the absence of a count. Until the count is made, or the formula is pointed at a living estimate, every other reform in this article operates inside a ceiling that demography makes wronger each year.

Three decisions will determine what the next implementation record looks like, and the record assembled here identifies each of them precisely. The first is the census itself: a published count re-bases Section 9 automatically, enlarges the ceiling by the growth of the population and redistributes the state shares, and no administrative reform substitutes for it. The second is the percentages: Parliament can hold seventy-five and fifty, adopt the NITI Aayog paper’s sixty and forty, or choose another pair, but the choice belongs to the legislature that wrote Section 3, and it is a fiscal choice wearing a welfare argument’s clothes. The third is the free layer: the zero price runs, on the Cabinet’s decision, to December 2028, and its renewal or lapse will be taken as an executive act under no statutory compulsion either way, which means the price of India’s food entitlement for the years beyond that date rests on a decision the Act does not govern. A system that began by converting a welfare scheme into a legal right has ended, on these three points, depending on decisions taken outside the right’s own text. That is not a counsel of despair; the delivery record shows what the Indian state can do when it decides the machinery matters. It is a direction for attention. Watch the count, watch the percentages, and watch the layer, and the rest of the record will explain itself.

Frequently Asked Questions

Q: Why is food security coverage still based on the 2011 Census?

Because the statute says so in terms. Section 9 of the National Food Security Act, 2013 requires the total number of persons covered in each state to be calculated on the basis of the population estimates as per the census of which the relevant figures have been published. The census of 2011 is the last census whose figures have been published, the decennial census due in 2021 having been postponed, so the coverage ceiling for every state is still computed on the 2011 population, producing the national intended coverage of 81.35 crore persons. The base changes only when a new census is published or Parliament or the executive adopts projections in its place.

Q: How many people are excluded from the food security net?

The estimates form a range because they rest on population projections. Jean Dreze, Reetika Khera and Meghana Mungikar, projecting the 2011 census forward, calculated that on 2020 population figures about 92.2 crore persons ought to have been covered at the statutory percentages against about 80.9 crore covered by food security ration cards, a gap of the order of ten crore people. Dreze put the figure for 2025 at more than twelve crore, and Khera has described a band of ten to fourteen crore across recent years. The Food Ministry measures against the statutory ceiling instead, recording about eighty crore persons receiving grain against the intended 81.35 crore.

Q: Does One Nation One Ration Card actually work?

As a capability, yes, at national scale. Portability began in four states in August 2019 and covered all thirty-six states and Union territories by June 2022, and the Food Ministry reported about seventy-one crore portable transactions by then, rising to 77.88 crore by August 2022, at an average of about three crore a month. The qualifications are composition and experience: most portability transactions are inside the beneficiary’s own state, the inter-state stream is far smaller, and the platform counts successful authentications rather than the attempts defeated by failed fingerprints, exhausted shop stock or an unfamiliar dealer. For short-distance and circular migrants the entitlement genuinely travels; for the long-distance single migrant it works least completely.

Q: How much foodgrain is lost between allocation and delivery?

On the survey-based estimates, the share has fallen steeply without reaching a settled level. Dreze and Khera put leakage at fifty-four per cent in 2004-05 and 41.7 per cent in 2011-12. For 2022-23, Khera’s estimate from the Household Consumption Expenditure Survey is 22.1 per cent, Das and co-authors estimate twenty-eight per cent, and Puri and Pingali estimate 24.1 per cent for 2022-23 and 8.8 per cent for 2023-24, with cautions about the survey data attached to the later figure. Every estimate is a residual between administrative offtake and survey-reported household purchases, so the author, survey and year must travel with any figure quoted.

Q: Has Aadhaar seeding excluded genuine beneficiaries?

Yes, on documented evidence, alongside documented gains. Seeding and de-duplication removed about 5.8 crore ration cards on the Food Ministry’s November 2024 statement, and the ministry credits the exercise with eliminating ghost and duplicate records. Against that, the randomised evaluation of Aadhaar-based biometric authentication in Jharkhand by Muralidharan, Niehaus and Sukhtankar found nearly ten per cent of genuine beneficiaries denied grain in the study period through linking or authentication failure, and the 2017 Jharkhand survey by Dreze, Khera and co-authors found exclusion errors as high as twenty per cent in villages where authentication was required for every sale. Both effects are in the record; the weighting between them is the argument.

Q: Are State Food Commissions functioning?

They exist in form in the states, with functioning that varies from real to nominal. Section 16 requires every state to constitute a Commission to monitor the Act, inquire into violations and hear appeals from District Grievance Redressal Officers, yet many states initially discharged the function by designating an existing commission under Section 18 or staffing the body on additional charge. In Swaraj Abhiyan v. Union of India, in 2016 and again in July 2017, the Supreme Court directed that commissions be constituted separately, with designation of another body acceptable only as a last resort, and that district grievance officers be independent. Constitution has broadly followed; staffing, inquiry capacity and published annual reports remain uneven.

Q: What did the Supreme Court say about rations for migrant workers?

In In Re: Problems and Miseries of Migrant Labourers, the suo motu proceedings on the pandemic exodus, the Court held the right to food to be part of the fundamental right to life and, on 29 June 2021, directed states to implement One Nation One Ration Card by 31 July 2021, to run schemes giving dry rations to migrant workers without ration cards, to run community kitchens, and to build the national database of unorganised workers that became the e-Shram portal; it also observed that the Centre may re-determine coverage under Section 9. In July 2022 it directed the Union to address the 2011 census restriction, and in March 2024 it directed ration cards for about eight crore e-Shram-registered workers who lacked them.

Q: Are foodgrains under the food security law still priced?

The statute still carries prices, and the beneficiary pays none. Schedule I of the Act fixes issue prices of three rupees a kilogram for rice, two rupees for wheat and one rupee for coarse grains, rates set for the first three years and never revised afterwards. Since 1 January 2023, however, a Cabinet decision has supplied the entire statutory entitlement free of cost under the Pradhan Mantri Garib Kalyan Anna Yojana, an arrangement extended for five years from 1 January 2024. The zero price is therefore a scheme decision running to December 2028, not a term of the Act; the Schedule I prices remain in the statutory text beneath it.

Q: What is the difference between allocation and offtake of foodgrains?

Allocation is the quantity the Centre assigns to a state from the central pool for a period, computed under Section 22 from the state’s identified population and its entitlements. Offtake is the quantity the state actually lifts against that allocation. The two diverge when a state cannot store, move or distribute its full assignment, when its identification is incomplete, or when demand at the shops falls short of the entitlement on paper. The distinction matters because leakage is measured against offtake, not allocation: grain allocated but never lifted cannot leak, and a state can show a wide gap between its allocation and what households receive without any grain being stolen, if the shortfall happened at the lifting stage.

Q: How do researchers estimate leakage from the public distribution system?

By subtraction between two independent records. The administrative record gives offtake, the grain states lifted from the central pool for distribution, published in the Food Ministry’s foodgrain bulletins. The consumption record comes from the national sample surveys, whose households report how much grain they bought from fair price shops, scaled up to the population. The difference between grain released and grain households report receiving, expressed as a share of the grain released, is the leakage estimate. The method’s weakness is that the residual absorbs survey recall error, grain routed to other welfare schemes and honest losses alongside diversion, which is why estimates by different teams for the same year can stand several points apart and why the method’s users insist on naming the survey round behind any figure.

Q: Did the Shanta Kumar committee recommend reducing food security coverage?

Yes. The High Level Committee on reorienting the role and restructuring the Food Corporation of India, chaired by Shanta Kumar and reporting in January 2015, recommended that coverage under the National Food Security Act be brought down from sixty-seven per cent of the population to forty per cent, with the covered population receiving seven kilograms per person at prices linked to half the minimum support price, and that cash transfers replace grain distribution starting in cities with a population above one million. It supported the retrenchment with a leakage estimate of about forty-seven per cent for 2011-12, higher than the survey series of Dreze and Khera, who disputed the figure in print. The coverage and price recommendations were not accepted.

Q: What happened when cash replaced grain in the food subsidy pilots?

The pilots ran in three Union territories under the Cash Transfer of Food Subsidy Rules, 2015: Chandigarh and Puducherry from September 2015, and the urban areas of Dadra and Nagar Haveli from March 2016, with the subsidy paid into bank accounts at one and a quarter times the minimum support price. MicroSave Consulting’s baseline work found the amount did not track retail prices and that market infrastructure in Dadra and Nagar Haveli was too thin, on which its rural pilot was dropped. NITI Aayog’s 2017 evaluation found about two-thirds of beneficiaries receiving transfers on average, roughly a fifth reporting periods of non-receipt, higher time and travel costs than the ration shop, and preference for cash growing over time. The pilots were never extended to a state.

Q: Why have crores of ration cards been deleted?

Because the digitisation of the beneficiary database let the state check every record against other databases at scale. The Food Ministry stated in November 2024 that about 5.8 crore ration cards had been removed through digitisation, de-duplication and Aadhaar seeding, and Parliament has since been given annual deletion figures of 41.99 lakh for 2023, 48.85 lakh for 2024 and 41.41 lakh for 2025, as states matched cards against tax, transport, corporate and death records. The government reads the deletions as ghost, duplicate and ineligible records removed. Critics, including Dreze and Khera in their work on the Jharkhand cancellations, respond that a large share of cancelled cards in the drives they examined belonged to real households, so a deletion count measures records failing checks, not fraud proved.

Q: Can a state cover more people than the food security ceiling allows?

Yes, from its own budget. The ceiling in Sections 3 and 9 governs the entitlement the Centre finances at Schedule I prices; nothing in the Act prevents a state from issuing its own cards, widening its list or enlarging quantities at its own cost, and several states run broader systems on top of the statutory floor. Dreze, Khera and Mungikar’s review of state lists counted, alongside the statutory cards, some ninety million people in 2020 holding state cards carrying entitlements equal to or better than the statutory ones, and further cards below that standard. The legal significance is that the top-up is a state scheme benefit, revocable by the state, while the statutory core is an entitlement the beneficiary can enforce.

Q: Did NITI Aayog propose cutting the coverage percentages?

Yes, in a discussion paper circulated in 2021. The paper proposed reducing the national coverage ratios from seventy-five per cent of the rural population and fifty per cent of the urban population to sixty and forty per cent, a change that would require amending Section 3(2) of the Act. On the projected population of 2020, the paper’s arithmetic brought the covered number to about 71.62 crore persons, with estimated savings of up to Rs 47,229 crore. The proposal ran in the opposite direction from the exclusion literature, which argues the covered number should rise with population, and it was not adopted; the ratios in the Act and the 2011 census base both remain in place.

Q: Can a family draw part of its grain in two different places?

Yes, under portability. The entitlement attaches to the ration card and is debited electronically at the electronic point of sale device, so a member working away from home can draw part of the household’s monthly grain at a fair price shop near the worksite and the remaining members can draw the balance at the home shop in the same month, each against biometric authentication. Before One Nation One Ration Card, the entitlement could be drawn only at the single shop to which the card was assigned, and a split draw of this kind was not possible within the rules. The facility serves the common migrant pattern of a divided household, part at the destination and part in the village, without issuing a second card.

Q: Will registration on the e-Shram portal get a worker a ration card?

Not automatically, but it has become a route the Supreme Court has pressed into service. The e-Shram portal, the national database of unorganised workers built after the Court’s directions in the migrant labourers proceedings and launched in August 2021, records the worker; it does not itself confer a food security entitlement, which still depends on identification by the state within its coverage ceiling. In March 2024, however, the Court directed states and Union territories to issue ration cards to about eight crore e-Shram-registered workers who did not have them, within two months, precisely so that registration would translate into access to the Act’s benefits. Implementation runs through the states’ identification and verification machinery and its pace is set by the ceiling problem.

Q: Is free foodgrain under the food security law permanent?

No. It rests on executive decision, not on the Act. The Act entitles beneficiaries to grain at the Schedule I prices of three, two and one rupees a kilogram. Free supply began as the pandemic scheme of additional grain from April 2020, and from 1 January 2023 a Cabinet decision folded the free element into the statutory entitlement itself for one year; the decision was then extended for five years from 1 January 2024 at an estimated outlay of Rs 11.80 lakh crore, which takes it to December 2028. When the decision lapses, the position in law is the Schedule I price, unless Parliament amends the Act or the executive extends the decision again. A scheme can be withdrawn by the government that made it; an entitlement cannot.

Q: What is a silent ration card?

A silent ration card is a card against which no grain has been drawn for a long period, a category the Department of Food and Public Distribution uses in its data-matching and verification exercises when it flags cards for checking. Silence is ambiguous evidence. It can indicate a ghost card or a household that has died out or moved away, and it can equally indicate a migrant household drawing under portability elsewhere, a family buying in the market in a good season, or a household repeatedly defeated by authentication failure. The verification exercises built on the flag have therefore paired it with electronic know-your-customer confirmation and a window for the household to establish itself, rather than treating silence alone as proof that the card should be deleted.

Q: Why do most portability transactions happen inside the beneficiary’s own state?

Because most movement is short-distance and most portability is convenience rather than migration. The Food Ministry’s transaction counts treat any draw away from the assigned shop as portable, including a different shop in the same town, and the bulk of draws are of that kind: a shop closer to work, a shop with stock when the assigned shop has none, movement within a district or between neighbouring districts. Inter-state draws are a small share of the total, though concentrated where migration runs, as when beneficiaries from twenty-one states drew grain in Delhi in April 2022 on the ministry’s count. The platform’s headline totals therefore measure the portability habit across the whole system, and only a fraction of them represent the long-distance migrant the reform was designed around.