Tuesday, February 10, 2009

How about randomly evaluate random evaluations?

The Center for Global Development has posted about a conference in Kuwait in which 3ie, an evaluation group that specializes in increasing local capacity for impact evaluations, announced it has received proposals for evaluations that total three times their initial funding available.

I've always been curious about the political economy implications of randomized evaluations. Do good randomized evaluations decrease the value and increase the cost of corruption? Do they have larger political implications for local officials, who have to tell their constituents they were randomly chosen for selection?

3ie could randomly select which proposals to fund, then evaluate the implications of the evaluations.

1 comment:

Matt said...

Here's another question: do randomised trials actually lead to informed policy? Randomised trials, of the sort managed by 3ei and the Poverty Action Lab are extremely precise, well-crafted, well-managed interventions. So the average treatment effect from the intervention will be the ate of a precise, well-crafted, well-managed intervention. The actual treatment effect when the intervention is eventually implemented as policy might be quite different!