Human Development

Giving up too early on malnourished children? Catch-up growth and Midday Meals

  • Blog Post Date 14 April, 2014
  • Articles
  • Print Page
Author Image

Abhijeet Singh

Stockholm School of Economics

It is widely believed that malnourishment in the first few years of childhood adversely affects cognition and adult economic outcomes. This column presents new research which shows that full recovery from early malnourishment is possible. Based on data from the state of Andhra Pradesh, it is found that the Midday Meals programme of the Government of India has been successful in compensating for early nutritional deprivation.

Childhood malnutrition casts a long shadow on individual outcomes well into adulthood (Almond and Currie 2011). Periods of malnourishment in the first 2-3 years lead to poorer cognition (Glewwe et al. 2001) and poorer adult economic outcomes (Maccini and Yang 2009), and these nutritional shocks tend to persist. The conventional view is that this damage is irreversible once the critical window of “a thousand days” from conception has passed. But what if this view is gloomier than the evidence warrants? Here I present evidence that “catch-up growth” is possible even after this critical window, and that policy might still make a difference.

Catch-up growth – the biomedical literature

Children in developing countries are born only slightly below the international reference standard for height and weight but quickly fall behind in the first 24 months of life (Victora et al. 2010). Most children who are stunted at two years of age continue to lag behind nutritionally at later ages (Martorell et al. 1994); this is frequently taken to imply that nutritional insults in early childhood cannot be remedied at a later age. However, the persistence of nutritional deprivation is not in itself evidence that these insults are irreversible. An alternative hypothesis is that full recovery from early malnourishment is possible, with faster than normal growth in malnourished children (this is called “catch-up growth” (Boersma and Wit 1997) in the medical literature). That this is not observed in practice could simply be because the correct conditions are not satisfied – environment and diet don’t change in most populations (Golden 1994).

In developing countries, a number of studies with panel data on individual children do find at least some degree of reversibility in stunting (height-for-age more than 2 standard deviations below the international reference)1 – in the Philippines, Peru, Senegal, Ethiopia and the Gambia, even until relatively late in childhood. A recent study, which utilised comparable child-level panel data from Ethiopia, India, Peru and Vietnam, not only documents recovery from stunting for some children but also suggests that this recovery “might have significant benefits on schooling and cognitive achievement”2. Even if catch-up growth is possible, as the evidence suggests, our understanding of what policy measures might lead to such recovery remains very limited. This is a gap that my co-authors, Stefan Dercon and Albert Park, and I seek to address in a recent paper (Singh et al 2014).

The Midday Meals Scheme and catch-up growth

In November 2001, the Supreme Court of India directed the Government of India to provide cooked Midday Meals in all government and government-aided primary schools “within six months”. By 2003, most states started providing cooked meals in primary schools. Covering an estimated 120 million school children by 2006 (Khera 2006), this is now the largest school-feeding programme in the world.

We used panel data from Andhra Pradesh collected by the Young Lives study to assess the nutritional effects of the Midday Meals Scheme. Previous work by Farzana Afridi had demonstrated that the school meals improved the daily nutritional intake of children, but it wasn’t clear whether it could have any effect on malnutrition, given that children are typically around five years old when they begin primary school, long after the critical 1,000 day window of intervention was over. More to the point, we wanted to evaluate whether a scheme as large as this, which had been successfully extended at scale across the country, could help in compensating for nutritional deprivation in early childhood.

Untangling the effects of a nationwide programme

Evaluating these effects is, however, far from straightforward. All government schools offered the meals to all children enrolled in primary grades starting at the same time. In the language of the impact evaluation literature, there was no clear relevant ‘comparison group’ that differed from the ‘treatment group’ only in the fact that the children in the treatment group were receiving the meals.

Importantly for our purposes, at the time of the second round of data collection (2007), only around half of the 4-6 year old children had yet started schooling. So, we could compare children who were already enrolled in government primary schools (and thus getting Midday Meals) with children who would join these schools later but hadn’t yet.

Merely defining this comparison group is not enough: the decision of when to enrol a child in schooling is not random and, in particular, we were concerned that some parents might enrol their children in school earlier to benefit from the meals. If these children were on a different nutritional trajectory than the comparison group – perhaps if their parents not only enrolled them early but also invested more in their nutrition - any estimates that we would get would be biased.

In order to address this problem, we used a feature of the relationship between age and enrolment in our data. Specifically we noted that there was a jump in enrolment between children born in December 2001 and January 2002 in whether they were enrolled in the 2006-2007 school year; we argue this is because parents and teachers use the calendar year of birth in which a child turns five as a ‘rule of thumb’ for enrolment – so children born in 2001 would be more likely to be enrolled in 2006-2007 than children born in 2002. This ‘discontinuity’, which leads to a child born in December 2001being more likely to be enrolled than a child born in January 2002, allows us to reliably estimate the effects of the meals on nutritional status. Essentially, our analysis depends on the assumption that children born in these adjacent months are similar in all other ways except that the December-born children are more likely to be enrolled. In particular, we do not expect the nutrition status of December-born or January-born children to be systematically different except through the effects of enrolment (and so Midday Meals).

School meals at age five compensated entirely for malnourishment from droughts in early childhood

Our findings were surprisingly positive. We found that the Midday meals had an effect both on the weight and the height of children who received these meals. In particular, the effect was largest for children whose nutrition had suffered as a result of droughts in early childhood.

India had suffered from a severe drought in 2002-2003, the worst nation-wide since at least 1987; rainfall in the Young Lives survey districts in Andhra Pradesh was up to 40% below normal. The drought, we found, had a large negative effect on the weight and height of children – almost a full standard deviation on height-for-age z-scores based on the World Health Organization (WHO) reference. We also found, however, that the Midday Meals compensated for this nutritional shock entirely – put simply, there was no difference in 2007 between the nutritional status of children whose households were not affected by the drought and those who had suffered from the drought but had later received the school meals.

Is this finding on the effect of school meals, compensating four years later for the effect of droughts in the first two years of life, plausible? In the light of the literature on catch-up growth, we think it is. More importantly, it provides perhaps the first example where a nationwide policy, evaluated at scale, can be shown to lead to catch-up growth. Given the high levels of child malnutrition in India, and the unpredictability of monsoon rains, these are hopeful findings.

What do these results mean for policy?

The direct connection between specific research findings and policy is rarely clear. At the very least, as I wrote in the Guardian last summer in the wake of the Bihar Midday Meals poisoning tragedy, these findings (and others on the programme) provide evidence on the benefits of the Scheme. They suggest further that there may be some hope for correcting nutritional deprivations which have been uncorrected by the crucial window of a thousand days from conception to the second birthday. Prevention and remediation of nutritional insults till two years of age is likely to be the most effective strategy, in costs and in impact, for dealing with child malnutrition; but for children who remain malnourished outside this critical window, some remediation might still be achievable.

This column is based on a paper that was published in the latest issue of Economic Development and Cultural Change (ungated version here).


  1. Stunting reflects that a child is in the bottom 3 percentiles of the international growth standards developed by the World Health Organization (WHO). For more details, see
  2. This is similar to older findings here but importantly different from Barham and coauthors’ recent experimental study.

Further Reading

No comments yet
Join the conversation
Captcha Captcha Reload

Comments will be held for moderation. Your contact information will not be made public.

Related content

Sign up to our newsletter