處理效果文獻在過去30年間有不少新的理論發展, 這篇文章將回 顧在Rubin 因果模型中, 當條件獨立假設成立時, 如何以機率倒數 加權法得到平均處理效果之認定及估計結果。我們也將說明在此假 設不成立的內生性架構之下,該如何利用工具變數來對局部平均處 理效果進行機率倒數加權之認定與估計。最後, 我們介紹一個對條 件獨立假設的統計檢定, 並總結近年來處理效果文獻的發展, 以及 提出幾個未來可能的研究方向。

Over the last three decades, much progress has been made in the treatment effect literature. In this article, we review the identification and estimation results of average treatment effects based on inverse probability weighting under the unconfoundedness assumption. We also discuss similar identification and estimation results of local average treatment effects when treatment assignment is endogenous but a binary instrumental variable is available. We then introduce a test for the unconfoundedness assumption and summarize some recent developments in the treatment effect literature. Finally, we point out some future research directions.