We use a randomized saturation design to investigate whether domestic election observers reduce electoral fraud in the 2012 presidential elections in Ghana. Results show that observers reduce the probability of overvoting (more votes cast than registered voters) at observed stations by 60...