Open Defecation and Childhood Stunting in IndiaAn Ecological Analysis of New Data from 112 Districts

If you had to name the single biggest driver of why Indian children are shorter than children almost anywhere else on earth — not hunger, not disease, and not poverty alone — what would you guess? The answer this paper lands on is open defecation. The evidence behind that answer is more specific and more troubling than you might expect. India has a paradox at its center. The country has seen real economic growth over the past few decades. Child mortality has fallen. Caloric availability has improved. And yet childhood stunting — being too short for your age, a marker of chronic malnutrition and a predictor of worse cognitive development, worse health, and lower earnings across an entire lifetime — remains extraordinarily high. Spears, Ghosh, and Cumming set out to understand why. The scale of both problems is staggering. Over a billion people worldwide defecate in the open, and more than 600 million of them live in India. Of roughly 215 million stunted children globally, twenty-eight and a half percent live in India. These aren't rounding errors — they are the defining features of a public health crisis that economic growth, on its own, hasn't touched. To study the connection, the authors matched two new district-level datasets. Stunting data came from the HUNGaMA survey, carried out by the Naandi Foundation between late 2010 and early 2011, which measured over 109,000 children across 112 districts. Open defecation data came from the 2011 Indian Census for those same districts. Because only district averages were available — not individual records — the analysis is ecological: each data point is a district, not a child. Spears and colleagues are upfront about what that means. Ecological regression can reveal population-level patterns, but it can't definitively establish what's happening at the individual level, and it's vulnerable to confounding from other district-level differences. To address that, they built their models in stages, adding controls one by one: socioeconomic status measured as monthly per-capita expenditure, maternal education captured through female literacy rates, calorie availability, household size, and the proportion of the population living in urban areas. One of the methodological choices worth explaining is how they handled the open defecation variable itself. Because its distribution across districts was skewed, they ran a Box-Cox procedure — a statistical technique that finds the best power transformation to apply to a variable. You can think of it as asking: should we use the raw number, the square root, or the logarithm? The maximum-likelihood estimate pointed clearly toward the natural logarithm, and that's what they used in their main regressions. Now for what they found. A ten percent increase in open defecation was associated with a zero point seven percentage point increase in child stunting — and the same effect held for severe stunting, defined as more than three standard deviations below normal height for age. In the log-scale framing they preferred, a one-unit increase in the log of the percent of a district defecating in the open was associated with roughly a seven percentage point increase in stunting. The coefficients in the fully controlled regressions were seven point one for stunting and six point six for severe stunting, both statistically significant. To appreciate what those numbers mean, look at the contrast between the worst and best districts in the sample. Low-performing districts — one hundred of the one hundred twelve — averaged fifty-nine percent of children stunted, with seventy-six percent of households defecating in the open. The twelve high-performing districts averaged thirty-six percent stunted and thirty-four percent open defecation. That's a twenty-three percentage point gap in stunting and a forty-two percentage point gap in sanitation. Spears and colleagues calculated that differences in open defecation can statistically account for between thirty-five and fifty-five percent of that stunting gap, depending on specification. For a single environmental factor — after controlling for income, education, and calories — that's a remarkably large share. One finding that cuts against the conventional story is what didn't predict stunting: calorie consumption. In their models, calorie availability did not explain higher stunting rates. That's the heart of the paradox. Children in some of these districts aren't short because they're not eating enough. Something else is going wrong between food intake and growth. The biological mechanism the paper points to is environmental enteropathy — chronic, low-grade inflammation of the gut caused by repeated fecal-oral contamination. When children are repeatedly exposed to fecal pathogens in their environment, the lining of the small intestine can become chronically inflamed and structurally damaged. The gut loses its ability to absorb nutrients efficiently. So even when food is available, the body can't use it properly. The authors cite work by Humphrey and by Lin and colleagues showing that children exposed to worse household sanitation show markers of enteropathy and are notably shorter on average. The disease environment, not just diet, shapes growth trajectories. Spears and colleagues also note that a meta-analysis of thirty-eight food supplementation programs found the best height gains achieved were zero point seven of a standard deviation. Sanitation, by their estimates, may be doing as much work as the best nutritional interventions ever have. That reframes what sanitation infrastructure is: not just a dignity issue or a diarrhea-prevention measure, but potentially a nutrition intervention. Before accepting the main estimates at face value, there's a methodological wrinkle worth sitting with. The authors ran a Monte Carlo simulation to test what happens when you dichotomize continuous height data — turning a child's measured height into a binary yes or no stunted indicator. Most large-scale datasets do this, including HUNGaMA, which is why the main analysis had to use it. But is something lost in that simplification? The simulation drew on nearly forty-one thousand children measured in India's third National Family Health Survey. The team took one thousand random subsamples of twenty thousand children each and ran three regressions on every draw: one using mean height-for-age as the outcome, one using percent stunted, and one using percent severely stunted. They tracked the R-squared and the t-statistic on the sanitation coefficient across all one thousand replications. In nine hundred fifty-five out of one thousand draws, the R-squared was higher when continuous height was used instead of either binary measure. The t-statistic was largest for the continuous specification in eight hundred eighty-one out of one thousand draws. In plain terms: dichotomizing height throws away statistical power — consistently and substantially. Here's why that matters for the main findings. The district-level HUNGaMA data only gives dichotomized stunting rates. So the central regressions — the ones showing zero point seven percentage points per ten percent increase in open defecation — are necessarily using the weaker form of the outcome variable. The Monte Carlo results imply those estimates are conservative. The true association between open defecation and child height is plausibly stronger than what the paper's headline numbers show. The study's main finding is a lower bound, not an upper one. The authors are careful throughout about what their design can and cannot establish. Ecological analyses are susceptible to residual confounding — there may be district-level characteristics they couldn't measure that explain part of the pattern. The one hundred twelve districts weren't randomly selected; one hundred were chosen because they were low-performing, which concentrates the sample in places where both sanitation and stunting are bad. And the study cannot prove that individual children who are exposed to open defecation are the same children who end up stunted. What the study can do — and does — is show that the population-level pattern is strong, consistent across specifications, holds after controlling for income and education and calories, and accounts for a large fraction of the gap between India's worst and best districts. At the time of publication, intervention studies were underway to test causality more directly. The policy implication follows from the evidence. Over six hundred million people in India defecate in the open. Stunting in the districts studied averages fifty-six percent. A twenty-three percentage point gap separates the worst districts from the best, and open defecation explains between a third and more than half of that gap. Sanitation has already been recognized as a human right. This paper argues it should also be recognized as a nutrition-sensitive intervention — meaning that progress on toilets may do more for child growth than additional calories. Building that infrastructure isn't only about dignity or disease prevention in the narrow sense. According to this evidence, it may be one of the most direct levers available for changing the developmental trajectory of India's children. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.