This is the 3rd in this year’s series of posts by PhD students on the job market.
Nearly half of all children under age five in India are undernourished, yet parents often fail to recognize malnutrition in their children. For instance, in my study setting, while nearly 40% of children were stunted, 84% of their mothers described their height as “normal” or “tall”. My job-market paper examines a potential explanation: in high-malnutrition settings, parents may systematically overestimate their children’s nutritional status because they benchmark growth against local peers rather than global standards. This, in turn, can suppress demand for improved nutrition and sustain a self-reinforcing low-nutrition equilibrium, where widespread malnutrition becomes normalized and parents see little need for additional nutritional investment. These dynamics matter not only for individual child well-being, but also for long-run human capital and economic development: early-life undernutrition is linked to lower educational attainment, productivity, and earnings in adulthood.
A growing literature shows that beliefs play a central role in shaping decisions, and that inaccurate beliefs lead to suboptimal choices across various settings. Most existing work focuses on misperceptions that arise from unobserved information about others (e.g., Bursztyn and Yang 2022), or from information that is hard to access or understand (e.g., Jensen 2010, Fitzsimons et al. 2016, Dizon-Ross 2019). This study instead highlights how the source of information—specifically who individuals get information signals from—can itself generate systematic misperceptions. In other words, misperceptions can arise even when individuals correctly interpret the information they observe, if this information comes from a selected or non-representative reference group.
The Experiment
Design: I conducted an individual-level field experiment with 1,021 mother–child pairs across 168 government childcare centers (Anganwadi Centers, or AWCs) in the state of Telangana. Eligible children were 7-24 months old at baseline, an age when growth faltering is acute. Centers were classified into “high-nutrition” and “low-nutrition” groups based on average child anthropometric z-scores using government administrative records, allowing me to examine how beliefs and treatment effects vary across different nutrition environments. Within each center, children were randomly assigned to control and treatment groups, stratified by sex and malnutrition status.
Setting: Telangana has experienced a 75.5% increase in per-capita income over the past decade and large improvements in sanitation coverage, yet child malnutrition rates remain stubbornly high. The government provides substantial supplementary food for children under age 3: 16 eggs and 2.5 kilograms of Balamrutham (a fortified, ready-to-eat therapeutic food) per child each month, covering roughly 40 % of recommended calorie needs and the full protein requirement for a 16- to 20-month-old child. While nearly all mothers reported receiving these rations at baseline, fewer than half said their child regularly consumed them—pointing to demand-side, rather than supply-side, constraints.
Treatment: The intervention aimed to highlight gaps between perceived and actual child growth, and shift benchmarks for healthy development. Enumerators used simple picture cards to first elicit mothers’ perceptions of their child’s height- and weight-for-age percentiles relative to other children in their own sub-district and around the world. Mothers were asked to imagine 100 children of the same age and sex as their own child, arranged from shortest to tallest height (and lowest to highest weight), and point to where their child would rank in that line-up. They were then shown the child’s true percentiles relative to WHO growth standards, helping them visualize exactly where their child stood in the distribution and highlighting any overestimation. Mothers were also provided height and weight reference values at different percentiles to recalibrate benchmarks for healthy growth and were encouraged to set incremental goals to move up the ranks. The script emphasized that average child growth in the study villages lagged behind that in developed regions and that comparisons only to local children could obscure deficits. Mothers were also provided brief, general nutrition advice comparable to information they already obtain from AWCs (e.g., encouraging use of government supplementary food), but the core intervention focused on correcting beliefs on growth.
The information was delivered immediately after the baseline survey, with one round of reinforcement messages sent via Whatsapp three months later. Endline data were collected six months after baseline, allowing for analysis of short-run impacts on outcomes across three domains: (i) beliefs, (ii) feeding practices, and (iii) child growth.
Key findings
Baseline data revealed large distortions in how parents perceive their children’s growth. On average, mothers believed their children were at the 47th percentile for height and 37th for weight, when true values were 14 and 13, respectively. They reported nearly identical rankings between global and local comparisons, implicitly viewing local children as representative of children worldwide. Perceived percentiles more closely tracked true values in local comparisons. Mothers were also asked to use a measuring tape to indicate what they considered to be the “ideal” height for their child’s age. This reported ideal height was well below global norms, corresponding to an average height-for-age z-score (HAZ) of -1.83. Despite 38% of the children being stunted (HAZ < -2 or height percentile < 3), 84% of mothers described their height as “normal” or “tall”, and 74% believed their child was of average or above-average height (Figure 1).
Figure 1.
Importantly, these misperceptions were not random. They were systematically larger in villages where the average child was shorter. Mothers in low-nutrition centers held more optimistic views of their children’s growth, reported lower benchmarks for ideal height, and were more likely to describe malnourished children as normal. In centers with the lowest average nutrition, reported ideal height even fell below the stunting threshold (Figure 2). This highlights the key paradox: parents in areas with the highest malnutrition rates are also the least likely to recognize it.
Figure 2.
The intervention shifted beliefs and feeding practices. The treatment substantially reduced misperceptions, raised ideal height benchmarks, and improved recognition of undernutrition. Mothers in the treatment group reported greater diet adequacy, increased protein consumption, and higher utilization of government supplementary food. Treated children consumed 5.6 grams more protein (a 71% increase) and were about 50% more likely to have eaten Balamrutham or an egg, as measured using a 24-hour diet recall module; monthly recall data corroborate these patterns, indicating sustained increases in consumption. Mothers’ weight and meal frequency were unaffected, and treatment effects were similar for children with and without siblings, suggesting that the improvements were not driven by diverting food away from mothers or other children.
Within 6 months, treated children gained 0.09-0.15 SD in anthropometric z-scores, and the fraction of underweight children fell by 25%. Stunting and wasting rates were also 10% and 22% lower, though not statistically significant at conventional levels. These are large effects: a back-of-the-envelope calculation suggests that, if scaled nationally, the intervention could close roughly 7% of India’s height deficit among children under 3 years, and 10-17% of the gaps in weight-for-age and weight-for-height; a more conservative approach targeting rollout only in low-nutrition areas could still close 5-11%.
The treatment effects on both diets and growth were strongest among children whose mothers initially had the largest misperceptions, defined in terms of the gap between perceived and true height percentiles at baseline. This heterogeneity provides the key causal evidence on the mechanism: it is precisely where misperceptions were largest that information had the biggest impact. Mechanism tests suggest these effects operate primarily through belief correction, rather than the brief general nutrition advice.
Cost-effectiveness
This belief-correction intervention is low-cost and scalable. It cost about USD 12 per child in the randomized evaluation but would fall below USD 1 if scaled and integrated into existing public programs. With an estimated cost of USD 140 per 1 SD improvement in HAZ and USD 390 per stunting case averted, it is far more cost-effective than typical nutrition and cash transfer interventions (~USD 900-8,000 per case). Because belief distortions are most severe in the poorest growth environments and among the most malnourished children, it is also inherently progressive.
Policy implications
India, and many other low- and middle-income countries, operate large-scale public nutrition programs providing free food and growth monitoring services, yet malnutrition rates remain high. The findings suggest that demand-side constraints, rooted in parental misperceptions of child growth, may mute the impact of these programs. Embedding belief-correction into frontline nutrition platforms, and delivering information in a form that is easily interpretable and salient for parents, could strengthen demand for improved nutrition, increase utilization of services, and ultimately improve child health.
Key Takeaway: when local norms obscure true deficits, correcting reference points can unlock underused, high-return investments—a lesson that generalizes beyond child nutrition to many other domains.
Sneha Nimmagadda is a PhD candidate in the Department of Economics at the University of Southern California (USC).
Join the Conversation