From 2013 to 2014, I was busy. I visited 14 different countries in the year after I graduated college, and I didn’t spend more than one month in any city or more than three months in any country. I got rabies shots after being bitten by my own dog in Uganda, I was awoken by hippos tearing grass just outside my tent while sleeping next to the Nile, lived in a cinderblock room with no running water, learned to ride a motorcycle, worked on a goat farm in France, and saw an altar made of skulls in Rwanda. I had just launched out of four years of school, and I was hungry to see and try and DO as much as I could. 2013 was also the year that I began writing this blog. I’ve looked back on those entries over the last few days (some of my favorites are What Noah Thought and New Year’s Resolution), and I’m reminded of how curious and independent I felt over that year. That year I worked with purpose and dedication. Despite all of my diffuse and crackling energy, I concentrated on a project I really cared about: the Village Enterprise Randomized Control Trial in Uganda.
Village Enterprise (VE) is an American NGO that runs what’s called a Graduation Program: participants in their programs receive a cash grant to start a microenterprise, but also receive agricultural and bookkeeping training, formation into collective savings groups, and regular meetings with a coach over the course of a year. In 2013 VE’s program monitoring data indicated that this approach created substantial and sustainable improvements in the financial stability of program participants. Other NGOs, however, favored the alternative strategy of giving larger lump sum cash transfers and counting on participants’ intrinsic self-interest to use the money however suited them best. They had data and field observations saying this method was pretty good too. One program seems expensive and logistically complex to run but fully supports participants, the other seems simple, direct, and less paternalistic. Plausibly either one could be effective: how do we know which is better?
We create a Randomized Control Trial, that’s how. In an RCT you get a big group of participants and randomly sort them into different subgroups. One group gets Treatment A, another gets Treatment B, a control group gets Nothing, and after surveying these participants closely over a few years at the end you can see how their incomes, food security, and asset holdings have fluctuated. Remember that all participants start on the same level and ostensibly the only difference between them is receiving Treatment A, B, or Nothing. Do this with a large enough group of people, and at the end you can say definitively: “In this case, doing Treatment A is more effective than Treatment B.”
A key objective of the VE RCT was to test these two approaches simultaneously; Graduation Program vs. unconditional cash transfers. There is, of course, a gray area between these two extremes, and this is what I was brought in to explore. I was hired to work on behalf of a pair of behavioral psychologists to develop a behavioral intervention: a “light touch” program that paired cash transfers with two brief coaching workshops to see how the results compared to a fully hands on or hands-off approach. I lived full time in Uganda for five months piloting this intervention and finalizing the program in the regions of Hoima and Soroti. Over the course of multiple trips to the village by motorcycle to see the program in action, 12-hour journeys by minibus from one site to another, and constant tweaks and iterations, the final program design fell into place.
Our focus was on goal setting and plan making. A facilitator met briefly one-on-one with a cash transfer recipient to elucidate that person’s goal and the steps that they would take to attain it. The goal might be buying a bicycle, paying school fees for a year in advance, or buying a metal roof for their house. The cash recipient then participated in a two-and-a-half-hour workshop that reiterated plan-making exercises, encouraged participants to study good examples in their communities and to build flexible and adaptive plans, and prepared them to resist temptations and social pressure to spend their cash transfer. Participants were reconvened six months later to receive the rest of their cash transfer and to discuss their progress so far, but that was it. No one would follow up with them to make sure they stuck to their plan, no one would punish them if they wasted the money, and they were not formed into groups to encourage savings.
After working on this RCT and the Behavioral Intervention for a year, a few days ago I received my performance review. The results of the study have been published by Richard Sedlmayr with the Oxford Centre for the Study of African Economies. Some of the paper is incomprehensible to me. Consider:
“𝑦𝑖𝑗𝐹 = 𝛼𝑗 + 𝛽𝑇𝑖𝑗 + 𝛾𝑦𝑖𝑗𝐵 + 𝛿𝑋𝑖𝑗𝐵 + 𝜀𝑖𝑗
Here, 𝑦𝑖𝑗𝐹 is the per-capita outcome in household 𝑖 in village cluster 𝑗 at the time of follow-up 𝐹; 𝑇 is the randomized assignment, coded to 1 for intent-to-treat and to 0 for the counterfactual; 𝑦𝑖𝑗𝐵 is the baseline observation of the outcome; and 𝑋𝑖𝑗𝐵 is a set of socioeconomic baseline covariates. The coefficient for the intent-to-treat estimate is 𝛽.”
I include this excerpt to demonstrate just how objective, clinical, and quantitative an RCT is. Most NGOs share success stories with glossy announcements on their website about how your ten dollars can help Fatouma vaccinate her daughter, but RCT analysis is all about the numbers. And here’s what happened: the Behavioral Intervention worked. All of these affirmations and plan-making exercises—tell the group a story about a time that you overcame an obstacle, draw the steps that you will take towards your goal in sequence on this piece of cloth, tell us a mantra you can repeat to help you focus on your goal—actually paid off.
The paper observes:
“Table 8 suggests that the behavioral intervention altered the investment patterns of cash transfer recipients, leading to increased livestock investments. Income from farming increases as well, and we see some indications that income from paid employment falls. Table 32 suggests that children started working fewer hours, though no effects on schooling outcomes are discernible. We see indications of gains in subjective well-being and diverse other psychological outcomes, with a strong signal on respondents’ sense of pride (Table 14).” [Page 18]
I am so incredibly proud of the behavioral intervention, this research project, and these results. This is an innovative solution to poverty alleviation that appears to be successful, and achieves impressive gains for relatively low costs. Put another way, doing this is better than doing nothing. The project was a tremendous growth experience for me, and had a significant influence on my career choices over the last three years. I got better at working independently and as part of a team, gained experience working in rural areas in challenging conditions, and had the privilege of working with some of the sharpest, most dedicated and good-humored colleagues I’ve ever known. I would be better off for having worked on this project in any case, but I feel great satisfaction with the final results.
This brief and rudimentary description can’t do justice to a project of such broad theoretical scope and quantitative complexity. I highly encourage you to click on the link below and at least read the abstract. Time will tell, but I think we haven’t seen the last of this kind of research.