Voting and Peer Effects : Experimental Evidence from Mozambique

Voter education campaigns often aim to increase voter participation and political accountability. We follow randomized interventions implemented nationwide during the 2009 Mozambican elections using a free newspaper, leaflets, and text messaging. We investigate whether treatment effects were transmitted through social networks (kinship and chatting) and geographical proximity. For individuals personally targeted by the campaign, we estimate the reinforcement effect of proximity to other targeted individuals. For untargeted individuals, we estimate the diffusion of the campaign depending on a proximity to targeted individuals. We find evidence for both effects, similar across the different treatments and across the different connectedness measures. We observe that the treatments worked through the networks by raising the levels of information and interest about the election, in line with the average treatment effects of voter education on voter participation. We interpret this result as a free riding effect, likely to occur for costly actions