To Give or Not to Give: Academics Study Bednet Distribution
On Wednesday, I attended a presentation by Jessica Cohen on her doctoral research experiment (co-authored with Pascaline Dupas) about distribution schemes for anti-malarial long-lasting bednets. The randomized experiment, conducted in rural, western Kenya, aimed to test the difference between free distribution and cost-sharing schemes in terms of their direct impact on malaria prevention in pregnant women and their infants. A draft of their paper is available through the Brookings Institution.
In the past, I have been an advocate for the continued use of social marketing (cost-sharing) and for profit-based BoP business models to produce and distribute insecticide treated nets (ITNs). Because of this history, I was a little apprehensive about what Cohen’s presentation would reveal. After all, the report’s summary stated clearly that the experiment produced “no evidence that cost-sharing reduces wastage on those that will not use the product: women who received free ITNs [insecticide-treated nets] are not less likely to use them than those who paid subsidized positive prices.“It turns out, though, that the real conclusion of the experiment in no way supports any blanket claims, such as those made to the New York Times by Dr. Arata Kochi, Director of the WHO’s malaria program: “Virtually the only way to get the nets to poor people is to hand out millions free.” Instead, the study arrives at this cautious and tempered conclusion: free distribution of bednets is not the way but also a way, in some cases, for lowering malaria rates in a cost-effective manner.
The reason for such caution does not stem from intrinsically poor experimental design, or from unclear results – the experiment was conducted thoroughly, with admirable attention to detail, appropriate variables, and robust results, especially given funding and information constraints. Rather, the authors’ caution may have more to do with the difficulties of applying the results. This is a localized study, and it carries the baggage of a number of inextricable circumstantial factors. As a result, applying the results to reach a verdict on best practices for bednet distribution or malaria prevention is difficult if not disingenuous. There are two important points here that should raise the suspicions of anyone who sees this study being used to fuel the anti-social marketing or anti-cost-sharing fires.
First, PSI (Population Services International) has been active in the same area where the experiment was conducted for years, and, according to a PSI representative also in attendance at the presentation, PSI had sold a large number of nets in that area prior to the beginning of the experiment – amounting to a theoretical coverage of 1.4 nets per household. Unfortunately, it is impossible to determine the actual pre-existing coverage and usage rates of nets prior to the experiment (or the positive externalities in terms of awareness or general impact on public health from their use). The only figures available on coverage are from a 2003 Kenya Demographic and Health Survey, which reports that 6.7 percent of households had an ITN – an estimate that Cohen believes is completely outdated.
The experiment clearly shows high usage rates (above 60 percent) among women that received nets for free, and no significant difference between their usage rates and those of women who purchased the nets at prices varying between 10 and 40 Kenyan shillings. However, it is impossible to know if the high usage rates found in the experiment are evidence that no correlation exists between a person’s valuation of a net (reflected by their willingness to pay) and their usage rate, or if previous social marketing efforts had already eliminated much of this effect. If pregnant women already knew that bednets would be valuable to protect their health and the health of their newborns, then usage rates might be high even when nets were given to women for free.
The other important detail I’d like to highlight is that the experiment was conducted exclusively through 20 public health clinics. I would assume that this decision for the experimental design was done out of necessity or quality control, but it presents a problem: only those who sought their care through a public health clinic were studied. It says nothing of those who sought no care at all, or those who sought care from other services. This may not be exactly the same as studying the willingness to pay for coffee by only surveying Starbucks’ customers, but one sees where representation might be an issue.
Cohen and Dupas’ study is certainly valuable, in that it challenges the development community to look into possibilities of not choosing one model or the other, but of experimenting with combinations of cost-sharing vs. free distribution. This is a smaller reflection of an overall trend that I hope will happen – that the development community will find ways to use charitable solutions not to inhibit or kill off for-profit BoP business models, but to enhance them and fill the gaps where even they cannot reach.