YI-TING CHENHsu, Y.-C.Y.-C.HsuWang, H.-J.H.-J.Wang2020-12-162020-12-16202007350015https://www.scopus.com/inward/record.url?eid=2-s2.0-85056105995&partnerID=40&md5=24786fd0fd1bdfd1425a95b9a12df289https://scholars.lib.ntu.edu.tw/handle/123456789/526227Government policies are frequently used to promote productivity. Some policies are designed to enhance production technology, while others are meant to improve production efficiency. An important issue to consider when designing and evaluating policies is whether a mediator is required or effective in achieving the desired final outcome. To better understand and evaluate the policies, we propose a new stochastic frontier model with a treatment status and a mediator, both of which are allowed to be endogenous. The model allows us to decompose the total program (treatment) effect into technology and efficiency components, and to investigate whether the effect is derived directly from the program or indirectly through a particular mediator. Supplementary materials for this article are available online. © 2018, © 2018 American Statistical Association.Mediation analysis; Stochastic frontier models; Treatment effect[SDGs]SDG3journal article10.1080/07350015.2018.14975042-s2.0-85056105995