Assessing the Case for Social Experiments
分析了社会实验方法的假设条件，指出实验在识别项目平均影响时有效，但无法识别收益分布，除非附加假设。
This paper analyzes the method of social experiments. The assumptions that justify the experimental method are exposited. Parameters of interest in evaluating social programs are discussed. The authors show how experiments sometimes serve as instrumental variables to identify program impacts. The most favorable case for experiments ignores variability across persons in response to treatments received and assumes that mean impacts of a program are the main object of interest in conducting an evaluation. Experiments do not identify the distribution of program gains unless additional assumptions are maintained. Evidence on the validity of the assumptions used to justify social experiments is presented.