I have been critical of Collier's work before, but a new paper with him and Pedro Vicente, Votes and Violence: Evidence from a Field Experiment in Nigeria, is very interesting. They conduct an experiment in a random selection of neighborhoods in Nigeria using an anti-violence campaign during the 2007 general election.
They find that the program, no surprise, worked to reduce violence.
I am a little concerned though by their claim that "voter intimidation is effective in reducing voter turnout". I can't see how they can make such a strong claim on the results of the experiment. Yes, people in the treated communities voted more, but that could be a result of the additional attention on the election from the program. Certainly, more advertising for an election can bring out more people simply through the marketing effect.
Their sampling of what states to run the experiment in was also not random, but targeted toward the most violent states. We would expect the results of this experiment to be much larger than a scaled up program.
I also don't see any discussion of spillovers from the treated to the control neighborhoods. That though likely means their results could be biased downward.
Despite my critiques, I think its an interesting paper, and I am happy to offer some appreciation of the fine causal identification strategy.