Diffusion of Media Effects: An analysis of spillovers from two large field experiments in Uganda
Two blocked and clustered placebo-controlled field experiment with 8,880 respondents in 167 villages
Donald Green, Anna Wilke, Susanne Baltes
Innovations for Poverty Action (IPA Uganda), Peripheral Vision International (PVI)
When novel information is introduced into a subset of a community, to what extent and how does this information diffuse to its other members?
Policy-makers rely on mass media campaigns as a scalable means of reducing the prevalence of harmful practices in developing countries. The attractiveness of this approach depends in part on whether the effects of such campaigns are confied to the people directly exposed or whether they spill over to the broader population. We draw upon two randomized placebo-controlled media campaigns in Uganda to assess the ways in which media effects diffuse. Two experiments exposed over 10,000 Ugandans in 56 and 112 villages to video drama on domestic violence, abortion and teacher absenteeism. Outcomes were measured via seemingly unrelated surveys months later. The campaigns produced substantial effects on the views of those directly exposed and provide a unique opportunity to study spillovers. First, the distance between villages makes diffusion across villages unlikely, which helps us home in on intra-village spillovers. Second, the novel placebo-controlled design makes it possible to identify effects among subgroups within a village, such as those who were indirectly exposed to the campaign through friends and family. Finally, we sampled teenagers and parents from the same household, which enables us to analyze intra-household spillovers.