Thursday, February 05, 2009

Somewhat stochastically dominating summer camps

Blattman refers readers to a nice summary paper by Angus Deaton on using statistics in development to understand what kinds of aid assistance works. One sentence on randomized controlled trials (RCT) caught my attention:
Essentially, the RCT gives us two marginal distributions, from which we would like to infer a joint distribution; this is impossible, but the marginal distributions limit the joint distribution in a way that can be useful, for example if the distribution among the treated stochastically dominates the distribution
among the controls.
Interestingly, the distribution of the reading test results for students who were in the two week library summer camps do indeed dominate the control groups (that is, the distribution of grades is shifted to the right). The control groups either got two free books or were given small monetary incentives to participate in weekly book discussions. The mean score goes from 66 to 74 (p<.01), about a 10% increase. The cost of the camp was on the order of $20 per child. So not a huge effect for the cost, but seems reasonable enough.

No comments: