Sonja Vogt and I made the first of many trips to Sudan in 2011. It was summer. I remember looking at a thermometer one morning in Khartoum. 9:29 am. 49 degrees Celsius. I didn't know that was possible. We headed north that day, into the desert, and my re-education continued.
We were there to begin a research program on cultural traditions that harm women and girls. Our focus in Sudan was female genital cutting. Before we could begin, we had to deal with an important challenge. We were concerned that people in Sudan know that other people in other places view cutting as a shocking practice. What, then, does a local participant do? Does she hide her support for cutting to produce a favorable impression? Or, does she perhaps exaggerate her support to push back against outsiders passing judgment? Either way, the empiricist has a problem.
For this reason, we invested heavily in designing new methods to collect reliable data. For example, we spent well over a year developing an implicit association test about cut versus uncut girls. In addition, after months looking for a new way to estimate the prevalence of cutting, we settled on a method that relied on the tradition of applying henna to a girl's feet when she is cut.
When we started collecting data, the surprises came quickly. In 2015, we published a paper showing pronounced heterogeneity in attitudes and practices at extremely local scales. In effect, cutting and non-cutting families were neighbors. That same year, using completely different kinds of data from 13 countries in West Africa, Marc Bellemare, Lindsey Novak, and Tara Steinmetz came to similar conclusions.
These results were surprising because an insightful and influential model of cutting, first proposed by Gerry Mackie, posits that families face incentives to coordinate their choices with other families. These incentives should typically lead people to conform to local practice. Two possible outcomes follow, one in which most families cut and one in which most families do not cut. The model further suggests that an intervention can disrupt a cutting tradition so that conformity switches from supporting cutting to supporting abandonment. In this way, a policy maker can rely on social mechanisms within the cutting society to do much of the work required to reverse the tradition.
A number of empirical studies, including our own work in Sudan, have provided support for the model at the individual level; families do seem to care about conforming to local practice. The second part, however, namely homogeneity at the aggregate level, seems to be missing or at least unreliable. This is crucial because aggregate outcomes are the primary concern of policy makers and development organizations.
In any case, about a year later we published another paper showing that, whatever the reason for local heterogeneity, one can actually design an effective intervention that exploits this kind of variation. Instead of showing the tension between cutting and abandonment as a clash between cultures, our interventions were movies that dramatized the tension as a conflict among individuals within a culture.
Still, we were left with nagging questions about the apparent disconnect between individual choices and aggregate outcomes. Conformity seemed to be shaping whether families cut, but aggregate outcomes did not suggest that policy makers could rely on conformity to provoke an endogenous reversal of the tradition. This is the point of departure for our current paper.
Specifically, we developed and analyzed a number of empirically informed models. Each model begins by assuming that everyone, to some extent, values conforming to local tradition. People vary, however, in other ordinary ways. They vary in terms of whether they prefer to coordinate on cutting versus not cutting. They vary in terms of how they respond to a policy maker's intervention. They vary in terms of whom they know and the importance they attribute to maintaining a distinct cultural identity.
We find that local conformity can dramatically amplify the effects of an intervention, but it has no general tendency to do so. The relevant policy questions, perhaps unfortunately, are more challenging than generic notions about conformity reinforcing behavior change. In particular, the policy maker's choices interact strongly with the pre-existing heterogeneity in a population in some situations but not in others. Our paper, in turn, is an analysis of how and why the amplifying effects of conformity can range from almost nothing to nearly everything.