Tamil Nadu’s decision on September 22 to extend the Chief Minister’s Breakfast Scheme to classes 6 to 8 is important for more than just the additional children it will feed. The expansion adds about 15.30 lakh students, bringing overall coverage to roughly 34.64 lakh children. The larger significance is institutional: the scheme illustrates a form of welfare in which access follows participation in a public service, rather than an administrative judgment about whether a household is poor enough.
That distinction deserves serious attention because India’s welfare debate remains trapped in an unhelpful binary: targeted benefits versus universal benefits. The more difficult question is not whether targeting is desirable, but how targeting is designed in practice.
School meals offer one clear answer. Eligibility is broadly attached to school participation. The benefit is purpose-specific, visible, and relatively easy to administer. Tamil Nadu’s State Planning Commission has reported improvements in attendance, punctuality, classroom attention, and learning-related behaviour following earlier phases of the breakfast programme, while emphasising its core nutritional and educational objectives.
Cash transfers operate under a very different informational constraint. Governments rarely observe household economic capacity perfectly, particularly where earnings are informal, seasonal, or volatile. Means tests and proxy means tests evaluate a household's level of financial need based on reported income, assets, occupation, and living conditions.
That approximation inevitably produces errors. Some eligible households are excluded, while some ineligible households are included. The IMF notes that means-tested and proxy-means-tested systems face precisely these inclusion and exclusion problems, as well as administrative, compliance, and social costs.
However, the more important problem arises one step later. Why should an imperfect estimate of household welfare produce a perfectly sharp eligibility boundary? Suppose two households differ only marginally in measured income. One falls just below the administrative threshold and receives the full benefit, while the other lies just above it and receives nothing. Economic vulnerability may differ by Rs 1,000, but public support may differ by Rs 12,000. That is the eligibility cliff.
The difficulty is not targeting per se. Scarce public resources often require some concentration on those with greater need. The difficulty is discontinuous targeting: converting small differences in imperfectly measured economic capacity into large differences in entitlement.
Such cliffs can also generate severe behavioural distortions. When a modest improvement in earnings causes the complete loss of a benefit, the effective marginal cost of earning more becomes unusually high. Welfare economics has long recognised this problem through the concept of implicit marginal tax rates. IMF analysis similarly notes that withdrawing benefits as income rises can weaken work incentives if the withdrawal rate is too steep.
Administrative burden compounds the issue. Means-tested programmes require households to understand complex rules, produce documents, demonstrate eligibility, and periodically prove that their circumstances have not changed. Research on administrative burden identifies learning, compliance, and psychological costs as critical determinants of whether eligible citizens actually receive benefits.
None of this makes universalisation automatically superior. Universal benefits minimise screening errors and reduce administrative friction, but they also spread scarce fiscal resources across households with vastly different capacities. For a fixed budget, highly targeted transfers can deliver larger benefits to poorer households. The IMF therefore treats universal and targeted systems as a genuine trade-off involving fiscal cost, administrative capacity, work incentives, targeting errors, and social legitimacy.
Tamil Nadu’s next welfare debate should therefore move beyond rigid binaries. A far more coherent welfare architecture would consist of three distinct layers. The first comprises universal or near-universal basic services where exclusion carries high social costs and broad access generates extensive public benefits. This includes public education, school meals, primary healthcare, immunisation, essential nutrition, water, and sanitation. These programmes build capabilities rather than merely compensate for income shortfalls.
The second is graduated cash support. Eligibility need not operate as an on-off switch. Assistance can decline progressively as economic capacity rises. A household marginally above a benchmark would receive calibrated support, not lose it entirely. This would not eliminate targeting errors; it would greatly reduce the consequences of being placed just on the wrong side of an uncertain administrative line. The crucial design variable here is the taper rate: withdraw benefits too rapidly and the eligibility cliff reappears in another form; withdraw them too slowly and fiscal costs rise. That is a technical policy choice capable of being modelled, audited, and periodically recalibrated.
The third layer is outcome-linked support where a clear behavioural objective exists: scholarships for educational progression, maternal-health incentives, skill completion, or targeted assistance that removes identifiable barriers to employment. However, conditionality should remain limited. Poorly designed conditions can impose compliance costs that defeat the purpose of the benefit itself.
Such an architecture would also enhance political legitimacy. Welfare programmes command greater trust when eligibility rules are predictable, transitions are gradual, and losing a benefit does not appear arbitrary. Conversely, even an accurately targeted scheme may be perceived as deeply unfair if neighbouring households with similar circumstances receive radically different treatment.
This is why the most useful question for Tamil Nadu is no longer, "Are freebies fiscally irresponsible?"
A mature welfare state must ask instead: How should entitlements be designed when both exclusion and universality carry high costs? The breakfast scheme offers one viable answer because it links access to a public purpose rather than a disputed poverty threshold. Cash assistance will still be necessary for risks that public services cannot address. But where targeting is required, it need not produce sharp cliffs.
The real test of a modern welfare state is not how precisely it excludes people, but how intelligently it designs entitlements in a world where economic vulnerability exists on a spectrum, rather than on either side of a bureaucratic line.
Thakur is Professor and Director, Centre of Excellence (CoE) for Public Policy, Sustainability, and ESG advances research, Alliance University, Bengaluru