Two Surveys, One Country

Pritiraj BrahmaFounding Partner, Pritiraj & Partners · 9 August 2026

India’s two official sources on child undernutrition disagree about wasting by fourteen percentage points. One of them is now used to report progress, and it is not the one designed to measure it.

On stunting, India’s two official measurement systems broadly agree. On wasting — the indicator that signals acute, immediate, treatable hunger — one reports 19.3 per cent and the other about 5 per cent. Both are published by the Government of India.

India has two instruments for measuring child undernutrition and they do not agree. The National Family Health Survey is a large, population-representative household survey with trained measurement teams, conducted roughly every five years. The Poshan Tracker is an administrative platform on which anganwadi workers record the height and weight of children enrolled at their centres, monthly, at near-universal coverage.

The most recent NFHS, NFHS-5 (2019-21), found 35.5 per cent of children under five stunted, 19.3 per cent wasted and 32.1 per cent underweight — an improvement on NFHS-4 (2015-16), which recorded 38.4, 21.0 and 35.8 per cent respectively. The Poshan Tracker, as cited by the Ministry of Women and Child Development in a Rajya Sabha reply in December 2025 using data up to October, reported that of more than 6.44 crore children aged 0 to 5 measured, 33.54 per cent were stunted, 14.41 per cent underweight and 5.03 per cent wasted.

Stunting: 35.5 against 33.54. Close enough to be reconcilable. Wasting: 19.3 against 5.03. That is not a discrepancy to be reconciled; it is two different accounts of whether Indian children are acutely hungry.

Five findings anchor this analysis:

1. The divergence is concentrated in weight-based indicators. A peer-reviewed comparison published in The Lancet Regional Health – Southeast Asia (2024) restricted NFHS-5 to the subset of children reported as measured at an anganwadi centre in the previous year — 56 per cent of the sample — to make the two sources comparable. Even on that like-for-like basis, Poshan Tracker estimates were on average 1.86 percentage points lower for stunting, 13.72 points lower for underweight and 12.16 points lower for wasting. Overweight prevalence, by contrast, was 0.68 points higher.
2. The Tracker’s own figures have moved substantially in three years. PIB reported Poshan Tracker data for November 2023 showing 37.51 per cent stunted, 17.43 per cent underweight and 6 per cent wasted among roughly 7.44 crore children measured. A subsequent Rajya Sabha reply cited 38.9 per cent stunted, 17 per cent underweight and 5.2 per cent wasted among 7.31 crore children under five measured. The December 2025 reply put stunting at 33.54 per cent. These are not consistent trends; they are a series still finding its feet.
3. Coverage expanded extremely fast, which affects comparability. The Lancet analysis records 4.01 crore children aged 0 to 6 measured in July 2022 — 45 per cent of those registered — rising to 8.36 crore, or 94 per cent, by September 2023. A prevalence rate computed on 45 per cent of registered children and one computed on 94 per cent are not the same statistic.
4. State-level figures show wasting reported at implausibly low levels. Poshan Tracker data for June 2025, released in reply to parliamentary questions, recorded wasting at 3.21 per cent in Odisha, 3.83 per cent in Haryana, 2.95 per cent in Punjab, 1.55 per cent in Jammu and Kashmir and 0.25 per cent in Ladakh. For context, NFHS-5 put national wasting at 19.3 per cent.
5. India has no current population-representative measurement. The most recent nationally representative anthropometric data for Indian children remains NFHS-5, fieldwork for which concluded in 2021. Results of a subsequent round are not in the public domain.

Why wasting is the hard measurement, and the important one

Stunting is height-for-age. It reflects chronic deprivation accumulated over years and moves slowly, which makes it forgiving of measurement error: a child stunted last month is stunted this month, and a centimetre of error rarely flips the classification.

Wasting is weight-for-height. It requires two accurate measurements taken at the same visit, and it moves fast — a child can enter and exit wasting within weeks. It is therefore the indicator that identifies children needing treatment now, and it is the indicator most sensitive to equipment calibration, technique, rounding and the digit preference that appears in any hand-recorded dataset.

This ordering of difficulty maps exactly onto the divergence: smallest gap on stunting, largest on the two weight-based measures. That pattern is itself evidence about what is happening — not about Indian children changing, but about how they are being measured.

Administrative data cannot audit itself

The Poshan Tracker is a real advance in Indian governance, and the firm has no interest in disparaging it. Near-universal, monthly, name-based measurement of eight crore children is a delivery capability almost no comparable state possesses. Used to identify a specific child for referral, it is exactly the right instrument.

The failure is one of institutional design, and it is a familiar one. The same worker who measures the child records the measurement, and the aggregate of those measurements is used to assess the programme she works for. There is no adversarial check anywhere in that loop. No malpractice needs to be assumed — rounding towards the normal range, deferring a difficult measurement, or simply measuring a struggling toddler imprecisely under time pressure will produce a systematic downward bias without anyone intending one. That is why every serious learning-assessment system in the world separates the body that teaches from the body that tests.

The consequence is now material. Ministerial statements report improvement in malnutrition indicators citing Tracker prevalence, while the survey designed to measure prevalence has not reported since 2021. A programme is being marked by its own record.

The counter-case, honestly stated

Three arguments deserve real weight.

First, NFHS is not error-free. Household surveys have their own measurement problems, sample design constraints and non-response patterns, and its wasting estimates in particular have been questioned in the technical literature for implausibly high levels relative to India’s mortality and morbidity profile. It is not obvious in every case that the survey is right and the Tracker wrong.

Second, the two instruments measure different populations. Poshan Tracker covers children enrolled at anganwadi centres — 8.82 crore of an estimated 16.1 crore children under six in 2021, per a ministry reply. NFHS covers all children. If anganwadi enrolment is not random with respect to nutritional status, some divergence is expected and legitimate. The Lancet restriction addresses this only partially.

Third, a five-yearly survey cannot manage a programme. A minister who must act this quarter cannot wait for NFHS-6. The answer to a flawed real-time system is a better real-time system, not a retreat to a survey that reports every half-decade.

All three are true, and none of them justifies the present position, which is that two official numbers differ by fourteen points on the most urgent nutrition indicator and no published document reconciles them.

What we would do

1. Publish a reconciliation note with every Poshan Tracker release. One page: the Tracker figure, the most recent survey figure, the population difference, and the known measurement caveats. The Lancet authors did this work in 2024; the ministry should do it as routine.
2. Institute an annual independent re-measurement audit. A random sample of anganwadi centres, re-measured within days by teams that do not report to the ICDS chain, published as a bias adjustment for each state. This is the model that made India’s learning assessments credible, and the sample required is small.
3. Stop using Tracker prevalence for trend claims against NFHS. Two series with different instruments, populations and coverage histories cannot be differenced. Report Tracker change against Tracker baseline, and only from the point at which coverage stabilised above 90 per cent.
4. Make wasting the audited indicator, and publish the treatment cascade behind it. Children identified as severely wasted, referred, admitted to nutrition rehabilitation, discharged recovered. Prevalence without a cascade is a statistic; the cascade is the delivery.
5. Publish the NFHS-6 timeline and stand by it. A named release date for fieldwork completion and for results. As long as the most recent independent measurement of Indian child nutrition is five years old, every claim of progress rests on the programme’s own paperwork.
6. Publish measurement metadata. Device type, calibration date, and measurer identity are already implicit in the system. Making them explicit turns an unauditable aggregate into a dataset whose quality can be assessed — and lets a district officer find the centres whose numbers cannot be right.

Nothing here suggests Indian child nutrition is not improving. It probably is. The point is narrower and harder: on the indicator that identifies which children need food this week, India currently cannot say what the truth is — and has arranged its measurement so that no one is required to find out.