摘要：Real-world ex-ante impact assessments are far from the ideal experimental design, where the eligible population is
supposed to be randomly assigned to treatment and control groups. Often, many surveys in developing contexts do not
even collect data from a comparison group. We propose a methodology that recovers the counterfactual for ex-ante
impact assessments of policy interventions under the conditions of distance decay in the exposure to continuous
treatments and lack of control groups. We test this approach on data from a large-scale irrigation project in Ethiopia.
