Final Post

Before mentioning our results, it is important to emphasize one thing: malnutrition matters because it is an effect and cause of poverty. The short-term impact on children is overwhelming: 56% of infant mortality in India is caused by malnutrition. Unfortunately, for the children who survived malnutrition, the long-term impact of undernourishment -lower human capital- will reinforce their poverty cycle by diminishing their future earnings.  

Once we recognized the importance of malnutrition, our results become scarier: one third of the mothers -in our sample- are malnourished, with half of those being severely underweight. Furthermore, on average, 70.6% of the children under six years old are stunted and half of those are severely stunted. If that wasn't enough, 35.9% of the children are severe wasted (weight for height).

Several insights we discovered help explain this crisis: illiteracy, belonging to a Scheduled Caste, and being employed in daily labor are related to malnutrition in mothers. For children, on the other hand, many factors worsen nutrition indicators: 1) most children are already stunted at birth; 2) children that are breastfed after their first six months eat solid food fewer times per day, compared to those that are not breastfeeding; and 3) having a young or underweight mother makes a child more likely to be stunted and wasted.

Given these facts what can PRADAN do to improve the situation? Although there is no one single universal solution, three things that are important should be included: 1) promote increased energy intake, especially in girls; 2) emphasize the importance of mothers’ nutrition; and 3) promote health and hygiene practices. 

In addition to these recommendations, one thing should be recognized: it is better to develop local solutions that can be sustained over time. Otherwise, as we have witnessed in the past, improvements will not be permanent. Furthermore, as mentioned in the previous post, women must be at the center of all the efforts. If women stay held back one thing will happen for sure: all efforts will fail.  

Finally, I want -again- to give special thanks to Harvard’s Women and Public Policy Program (WAPPP) and to supporters of the Cultural Bridge fellowship for providing me the grant that made my summer trip possible. I am convinced that their support will help PRADAN develop new policies that will improve the lives of women. That's why we need to keep fighting; closing the gender gap in economic participation, political opportunity, health, and education is not only possible but also necessary. We need to insist: #NoWomanHeldBack. 

Are girls more malnourished than boys?

One of the main questions we were trying to answer with our research was this: are girls more malnourished than boys? Specifically, before our trip to India we knew that –on average- girls were more underweight than boys. Additionally, we also knew that severe underweight prevalence fell more for boys (by 23.7%) than for girls (by 10.8%). But this was the national average. It wasn’t certain what we would find in Purulia, West Bengal.

What we surprisingly found was that girls weren’t more likely to be stunted (short for their age) or severe wasted than boys. This -for sure- was a surprise because we were doing our research in poor rural villages. Thus, we were expecting to observe a big difference between boys and girls, as in the average poor rural parts of India. But this revelation motivated us to dig deeper to try to find possible explanations for this surprising result.

Our first discovery was that girls weren’t more malnourished than boys because all of them -boys and girls- born stunted! That is, the situation in the villages we visited was worse than what we expected. As you can see in the following graph, the average girl (and boy) in our sample was born shorter than what is expected for babies. But shorter in our case means stunted: more than two standard deviations (Z-scores in the graph) of what should have been the length of the babies, using the World Health Organization standards.

Stunting (height for age) Z-Scores

This led us to our second crucial finding: as opposed to malnourished boys, having malnourished women will translate to a malnutrition trap. Let me explain. First, if girls (and boys) stay stunted after their first two years of life they will be more likely to stay stunted the rest of their lives because the damage done during these initial years is likely to be irremediable. Second, as you can see with our result, stunted mothers are more likely to give birth to stunted children, reinforcing the malnutrition trap. Why this happens? Because shorter mothers are more likely to develop Intrauterine Growth Restriction (IGR): they are not tall enough to have an adequate space to allow their children to grow as they should during pregnancy.

In addition to these findings, it is important to mention a limitation of our research: because we didn’t have enough resources, we didn’t measure all the children in the villages (we only survey a representative –of these villages- random sample). Why I mention this as a limitation? Because research done by Rohini Pande found that birth order matter, especially for girls: “girls born before the family’s eldest son actually fare better than boys born after him, while girls born after the eldest son fare worst of all”. This means that girls are in fact receiving fewer resources for their development than boys, negatively affecting their economic prospects.

No matter the limitations of our research one thing is clear: women are malnourished and this in turn is triggering a malnutrition trap. If we want to solve India’s nutrition problem, women must be at the center of the efforts. Without admitting this fact every policy will fail.

Correlation is not Causation!

These days are exciting for people interested in statistics. Never before have scientists been able to use as many data as today. In fact, it seems that scientists are flooded with data; there is not enough time and capabilities to analyze data in time. That’s how Big Data was born: data seems to be too big, too fast, or too hard for current technologies to process.

A prime example of how Big Data can be tamed is Google Trends. As many of you know, using the incredible amount of information that is been gathered every second by their search engine, Google created a tool named Google Flu Trends to predict this disease. The promise was startling: to predict a disease outbreak in real time. Without any doubt, to predict any event (and disease outbreaks!) in real time would be a great accomplishment. Who would not want something like this? NO ONE.

But there was a little problem: Google was using that information without a clear idea of why people were searching for things related to flu. Furthermore, there is no theory (serviceable) to understand how information is propagated in today’s data-flooded days; could it be that social media overreacts to news creating the notion that –for example- a flu outbreak is higher than it is? No one can know for sure. What we do know is that Google Flu trends failed. According to research done by Harvard University and Northwestern University, Google Flu Trends prediction system overestimated the number of influenza cases in the US in one hundred out of 108 weeks during the 2011 - 2012 flu season (starting on August 2011).

Why I mention this? Because it is important to emphasize the importance of having a causal model (or a theory of change) before you analyze any kind of data. Without a true model describing how and why information is searched and propagated, it is easy to wrongly estimate everything.

Consequently, before analyzing the nutrition status of children and women in our study, it was important to first understand the direct causes of malnutrition. To illustrate this, look at the following picture (taken from Lancet):

Conceptual Model of Pathways to Death and Disability

In this conceptual model, you can see that breastfeeding practices are not linked to malnutrition: it impacts weight by affecting the number of infections. But when we started analyzing our data, we found that breastfeeding was highly correlated with malnutrition. Specifically, we found that -all else being equal- a child that is breastfed is 16.6% more likely to be underweight, 28.3% more likely to be severely underweight, and 12.9% more likely to be stunted. If we stopped here our analysis, we would conclude –wrongly- that breastfeeding was causing malnutrition.

But we knew better thanks to our causal model. We knew that we would have to look for another variable that was varying with breastfeeding. That is, the relationship that we found between breastfeeding and malnutrition must be disguising a true relationship between a third unknown variable with both breastfeeding AND malnutrition.

That’s precisely what we found: for children older than six months old, breastfed children also ate solid foods fewer times a day. Hence, our missing variable was the amount of solid food eaten by children! Specifically, breastfed children ate solid food only twice a day (and 8.5% of them were not given any solid food). In contrast, those who were no longer breastfed ate solid food three times per day.

In sum, before you analyze anything try to have a theory of change. Otherwise, you will be concluding wrongly because correlation is not causation!

Tails or Average

When you read a blog using statistics you will be –mostly- reading about averages. Why? Because when researchers try to grasp a relationship between two variables, the best way to uncover this is through what economists call conditional expectation. What does conditional expectation mean? It means -in common parlance- (conditional) average. Let me explain this concept with an example.

Many economists have tried to find the relationship between years of completed education and earnings. The common sense would say that if you study more years you would earn more money. To find if this is true, however, is not trivial. For each year of schooling there can be enormous variation; while some people with 8 years of schooling earn $850 USD per month, other people with the same years of schooling earn $6,500 USD. This situation poses a problem to the researcher. How can she/he detect a relationship between schooling and earnings with that enormous variation? The answer to this question is using averages; if you estimate the average earnings per year of schooling, it will be easier (more likely) to detect this economic relationship. The following graph from the excellent book Mostly HarmlessEconometrics explains how this works.

Why I bring this to the blog? Because of something that caught my eye during my internship: the appalling situation of some (few) women in India. It interested me because when I read about Indian women, they almost never describe the lives of these women. Why? Because these women are a minority –they are on the tail of the distribution. Therefore, you will find it harder to detect an economic relationship by focusing on these women.

What exactly caught my eye?

1. That some women didn’t know their birthday.
2. That some women got married when they were between 8 and 12 years old.
3. That some women had their first child when they were 13 years old.
4. That some women didn’t know how to read, write or count.
5. That some women didn’t go to the hospital to deliver their baby because she didn’t understand the pain (they didn’t know that they were experiencing contractions and would soon be in labor).
6. That some women didn’t have a name: they were referred to as someone’s daughter, then someone’s wife, to end up as someone’s mom.
7.  That –unsurprisingly- these women were more likely to be stunted or underweight.  

These facts describe the lives of some women but not the average Indian women. The less unfortunate will have a name, will get married at a more reasonable age, and will know their birthday. That’s why I posed the question tails or average? Because sometimes it is important to design policies to help not the average women but the one’s in the tail of the distribution, those who are living a true nightmare.

Implementation hurdles

Every time I start a new task, it is not carried out exactly the way I planned. This time was no exception. How could I conceive every hurdle I would encounter?

Let's begin with the basics. I don’t speak Bengali, the spoken language in West Bengal, India.  So, to interact with everyone we need an interpreter. But as you can imagine, you can never be sure if the message you are trying to give is given the way you planned. In focus groups, this is a critical issue. Initially, we planned to conduct two focus groups per village to gather as much information as possible. To make things easier, we chose different questions that would convey the insights we needed for our study (breastfeeding practices, access to food, hygiene practices, etc.). The idea was that our interpreter would carry out the focus groups instead of translating every answer to us.

But after the first two villages we knew our interpreter wasn’t steering the conversations in the focus groups, as we needed. We knew that gathering insights from focus groups wasn’t an easy task, but we thought –mistakenly- that the interpreter would carry out the job well enough with a script. Once we discovered she wasn’t generating the insights we needed, we decided to change the dynamic: instead of two we would carry only one focus group and we would be the ones directing the conversation instead of her. She would be translating every answer to us and then –depending on the answers- we would ask more questions. This approach proved to be more successful, as we are gathering interesting insights that I will share in the future.

Another hurdle we encountered was in the actual process of taking the surveys and measuring the children and their mothers. First, we changed survey questions that didn’t work as planned. Second, as we chose randomly the people we would survey, many people in the villages complained they weren’t being measured. This is clearly not an easy problem to solve because it involves telling people that we can’t measure their potentially malnourished daughter because it would generate bias in our estimations. So, we decided that we would measure the complainers without taking their data in our analysis.

Finally, something that I couldn’t know beforehand was how difficult it would be to weight and measure the children. As you can see in the following video, many children are afraid of any kind of contact with strangers. It has been extremely difficult to complete our surveys in each village we have visited.

Now, every time I read a paper, I’ll be thinking in these and other hurdles every researcher must confront. Certainly, as my fellow researcher told me, I will doubt their data is as perfect as it looks while reading the paper.