Cost-effectiveness measurement in development : accounting for local costs and noisy impacts

As evidence from rigorous impact evaluations grows in development, there have been more calls to complement impact evaluation analysis with cost analysis, so that policy makers can make investment decisions based on costs as well as impacts. This paper discusses important considerations for implementing cost-effectiveness analysis in the policy making process. The analysis is applied in the context of education interventions, although the findings generalize to other areas. First, the paper demonstrates a systematic method for characterizing the sensitivity of impact estimates. Second, the concept of context-specificity is applied to cost measurement: program costs vary greatly across contexts -- both within and across countries -- and with program complexity. The paper shows how adapting a single cost ingredient across settings dramatically shifts cost-effectiveness measures. Third, the paper provides evidence that interventions with fewer beneficiaries tend to have higher per-beneficiary costs, resulting in potential cost overestimates when extrapolating to large-scale applications. At the same time, recall bias may result in cost underestimates. The paper also discusses other challenges in measuring and extrapolating cost-effectiveness measures. For cost-effectiveness analysis to be useful, policy makers will require detailed, comparable, and timely cost reporting, as well as significant effort to ensure costs are relevant to the local environment.