Social programs predicated on the notion that their beneficiaries will change some behavior (perhaps due to improved incentives or new knowledge gained during the intervention) pose unique challenges for impact evaluation. Assuming that a randomized control trial (RCT) is appropriate, research teams often don’t know exact precisely how long it will take for the treatment group to change the target behavior and reap the benefits of the program. Consequently, it is difficult to determine when the treatment and control groups should be compared, i.e. when the program in question should be evaluated.
While delaying the evaluation of a given program increases the likelihood that participants will have had the time to benefit from the program, it also typically means denying a control group access to the treatment. As Barahona (2010) forcefully argues, researchers ought to limit, to the extent possible, both the scale and the duration of any deliberate exclusion of individuals who could benefit from a program1 . With that in mind, it is surprising that questions related to the timing of impacts have been relatively neglected in the program evaluation literature (see e.g. King and Behrman 2009).