Setting the stage

First -dear reader- I want to apologize for my delay in writing my blog posts. It is not easy to have Internet access in small cities. But finally, the internet is back!

My Indian trip started with the right foot but with a great amount of uncertainty. First, I met the amazing people of PRADAN in Calcutta and set the goals for my internship. We agreed on the study design and started arranging the logistics. For the next three weeks I (plus my wife and a fellow HKS student) will be on the field conducting surveys and focus groups in sixteen villages near Purulia, India.

The purpose of the study is to measure the impact of PRADAN’s programs on women’s and children's nutrition. The NGO asked us to quantify the impact depending on the type of intervention they did, whether it was a) improved agriculture and natural resources management to ensure food security; b) awareness on cleanliness and hygiene through safe drinking water, storing water, and washing hands; or c) a combination of both programs.

However, measuring the impact is not an easy thing to do. Especially if you don’t have the information of how people behaved before the intervention. For instance, if we discover that 58 percent of the children are malnourished, is that an improvement? There is no way of knowing for sure if you don’t have the initial conditions. Furthermore, if nutrition improved, how can we know it was caused by PRADAN’s programs? There are many other potential explanations that could have caused that improvement (e.g., government program, better weather to agriculture, etc.).

There are ways to solve this puzzle, but it involves taking action before you implement a program (e.g., with a randomized controlled trial). In this case, PRADAN’s programs weren’t designed in a way that it would be easy to measure their impact (PRADAN is now trying to systematically gather baseline data from new locations to facilitate the analysis).

The way we agreed to solve this is as follows: first, we decided to analyze PRADAN’s interventions in a specific –and small- region (Jhalda II) to avoid having variations caused by weather or any other geographic fixed effects. Second, we constructed a “control”: we chose certain villages -where PRADAN hasn’t worked- that are similar to the ones where they did intervene (similar in the ethnic composition, size, proportion of women, type of labor, etc.). We are assuming that both groups, PRADAN’s intervened villages and the control villages, would have had the same time-varying behavior if PRADAN hadn’t intervened. We recognize that this is a long shot, but we will try to use different econometric techniques to have the best estimation as possible.

Lastly, we decided the “final” steps: we randomly chose four villages for each group (the three different interventions and the “control”) and then we chose randomly the children and women we will survey. But as everything in life, implementation hurdles would cause us to change our plans, as we will see in our following post.