September 05, 2014
Diarrhea kills a million children a year, predominantly in the developing world, despite the fact that it is an easily preventable disease. The right combination of inexpensive and easy-to-use technologies, such as chlorine tablets, coupled with simple behaviors can almost entirely prevent it. Yet, despite this, there is a puzzling lack of adoption of these technologies and behaviors. Social scientists have found a consistent problem in health experiments of maintaining participation amongst poor households. Even when the technology is provided for free, households stop using it soon after the commencement of the intervention. Why do people not adopt technologies that are clearly welfare enhancing, even when these technologies are given away for free?
In a randomized controlled trial in the city of Karachi in Pakistan, I test the hypothesis that perhaps the participants are not able to learn about the efficacy of chlorine tablets because they are unable to detect a precise enough signal of tablet efficacy. In a poor urban setting, I provided households with a simple visual tool (called Info-Tool) to help them assess the efficacy of the tablets. Info-Tool allowed households to record the levels of diarrhea experienced by them using simple bar-charts that they colored in. Additionally, at the end of each month, I provided them a bar-chart of the norm of how many diarrhea days they could expect for that month. Diarrhea varies with season, so the norm that was provided was a moving monthly reference (related to the number of children under five in the household) and Info-Tool allowed households to visually compare actual recorded levels to the reference level provided.
The experiment had a simple structure, with a control arm and treatment arm, and was rolled out in three phases. In Phase 1, which lasted three months, the treatment group used Info-Tool to build up a pre-tablet record of diarrhea prevalence. Info-Tool usage allowed them to understand where their levels tracked in comparison to the provided reference. In Phase 2, which also lasted three months, the treatment continued to use Info-Tool but now both arms were offered freely delivered chlorine (where they could refuse to accept offered chlorine tablets). Finally, in Phase 3, the treatment group’s use of Info-Tool was discontinued but both groups continued to receive freely provided chlorine tablets.
The results are remarkable. Participation rates were significantly and persistently higher in the treatment group. At 57 weeks from the start of tablet delivery (at the beginning of Phase 2), the treatment group is almost twice as likely as the control group to accept offered tablets. Figure 1 shows the persistently higher uptake of tablets (weeks on the X-axis and predicted probability of tablet uptake for the control group in blue and the treatment group in red). It also indicates a distinct seasonality to uptake – as we enter the second summer season (near the 48-week mark) both groups demonstrate higher tablet acceptance. Significantly, the treatment group’s summer increase in uptake is higher than the control group, suggesting an impact of Info-Tool on the understanding of disease seasonality that households possess.
The results suggest that allowing households to track and reference their disease prevalence increased their ability to detect the efficacy of chlorine tablets thus making the intervention far more successful. More specifically, I believe that households were able to better learn about the effectiveness of tablets because Info-Tool provided a more precise signal about tablet effectiveness thus leading to higher uptake of tablets. It is also apparent that households do possess a general sense for the seasonality of the problem but with augmented learning treatment households show a higher likelihood to accept offered tablets as the “danger” (summer) season approaches.
The experiment is ongoing and, at the time of writing, is at the 57-week mark since the beginning of Phase 2 (making it one of the longest behavioral studies of drinking water treatment). Providing households with clearer signals on the effect of the tablets made the intervention far more successful. The experiment points the way forward in two important ways. First, it demonstrates a powerful new way to address a major global health challenge (under-five diarrhea). The intervention had strong and persistent effects a year after it began. Additionally, the intervention is very cost effective at close to $2 per person treated per year, which compares very favorably to interventions in the drinking water space.
Second, it provides a basis for better-structured social policy and interventions that look to provide recipients clearer signals on the benefits of participation. One can imagine other areas, especially in the domain of health, where technology adoption is critical. This idea can be applied in these areas to promote, for example, the adoption of anti-malarial bed-nets and drug regimens for diseases like TB and HIV. Not every behavioral space is amenable to an intervention of this nature, but the basic notion of making the effects of our actions immediately clearer certainly carries power.
Agha Ali Akram completed his Ph.D. in environmental economics at Yale University and is a post-doctoral researcher at Evidence Action. He can be contacted at [email protected].
This study was funded by the National Science Foundation.
