Center for Effective Global Action

Development interventions are commonly piloted by organizations with strong community ties. Reminding beneficiaries that a pilot is being evaluated may prompt them to take costly actions that reflect favorably on the implementer. We test for this form of desirability bias in an evaluation of an unsuccessful agricultural extension pilot that ultimately drove treated farmers away from the target crops. Making the evaluation salient during endline data collection led participants to neutralize this negative treatment effect by altering input purchases and cultivation patterns. Participants’ desire to support implementers can help explain why promising pilot results frequently fail to replicate at scale.