{"title":"synth2: Synthetic control method with placebo tests, robustness test, and visualization","authors":"Guanpeng Yan, Qiang Chen","doi":"10.1177/1536867x231195278","DOIUrl":null,"url":null,"abstract":"The synthetic control method (Abadie and Gardeazabal, 2003, American Economic Review 93: 113–132, Abadie, Diamond, and Hainmueller, 2010, Journal of the American Statistical Association 105: 493–505) is a popular method for causal inference in panel data with one treated unit that often uses placebo tests for statistical inference. While the synthetic control method can be implemented by the excellent command synth (Abadie, Diamond, and Hainmueller, 2011, Statistical Software Components S457334, Department of Economics, Boston College), it is still inconvenient for users to conduct placebo tests. As a wrapper program for synth, our proposed synth2 command provides convenient utilities to automate both in-space and in-time placebo tests, as well as the leave-one-out robustness test. Moreover, synth2 produces a complete set of graphs to visualize covariate or unit weights, covariate balance, actual or predicted outcomes, treatment effects, placebo tests, ratio of posttreatment mean squared prediction error to pretreatment mean squared prediction error, pointwise p-values (two-sided, right-sided, and left-sided), and the leave-one-out robustness test. We illustrate the use of the synth2 command by revisiting the classic example of California’s tobacco control program (Abadie, Diamond, and Hainmueller 2010).","PeriodicalId":51171,"journal":{"name":"Stata Journal","volume":null,"pages":null},"PeriodicalIF":3.2000,"publicationDate":"2023-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Stata Journal","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1177/1536867x231195278","RegionNum":2,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"SOCIAL SCIENCES, MATHEMATICAL METHODS","Score":null,"Total":0}
引用次数: 1
Abstract
The synthetic control method (Abadie and Gardeazabal, 2003, American Economic Review 93: 113–132, Abadie, Diamond, and Hainmueller, 2010, Journal of the American Statistical Association 105: 493–505) is a popular method for causal inference in panel data with one treated unit that often uses placebo tests for statistical inference. While the synthetic control method can be implemented by the excellent command synth (Abadie, Diamond, and Hainmueller, 2011, Statistical Software Components S457334, Department of Economics, Boston College), it is still inconvenient for users to conduct placebo tests. As a wrapper program for synth, our proposed synth2 command provides convenient utilities to automate both in-space and in-time placebo tests, as well as the leave-one-out robustness test. Moreover, synth2 produces a complete set of graphs to visualize covariate or unit weights, covariate balance, actual or predicted outcomes, treatment effects, placebo tests, ratio of posttreatment mean squared prediction error to pretreatment mean squared prediction error, pointwise p-values (two-sided, right-sided, and left-sided), and the leave-one-out robustness test. We illustrate the use of the synth2 command by revisiting the classic example of California’s tobacco control program (Abadie, Diamond, and Hainmueller 2010).
综合控制法(Abadie and Gardeazabal, 2003,《美国经济评论》93:113-132;Abadie, Diamond, and Hainmueller, 2010,《美国统计协会杂志》105:493-505)是一种常用的面板数据因果推断方法,其中一个处理单元通常使用安慰剂检验进行统计推断。虽然合成控制方法可以通过优秀的命令synth来实现(Abadie, Diamond, and Hainmueller, 2011, Statistical Software Components S457334, Department of Economics, Boston College),但对于用户进行安慰剂测试仍然不方便。作为synth的包装程序,我们提出的synth2命令提供了方便的实用程序,可以自动执行空间内和时间内的安慰剂测试,以及略去一个稳健性测试。此外,synth2还生成了一套完整的图表,用于可视化协变量或单位权重、协变量平衡、实际或预测结果、治疗效果、安慰剂检验、治疗后均方预测误差与预处理均方预测误差的比值、点向p值(双侧、右侧和左侧)以及留一鲁棒性检验。我们通过回顾加州烟草控制项目的经典示例来说明synth2命令的使用(Abadie, Diamond, and Hainmueller 2010)。
期刊介绍:
The Stata Journal is a quarterly publication containing articles about statistics, data analysis, teaching methods, and effective use of Stata''s language. The Stata Journal publishes reviewed papers together with shorter notes and comments, regular columns, book reviews, and other material of interest to researchers applying statistics in a variety of disciplines.