[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120838-en":3,"doc-seo-120838-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},120838,16904993612988,"Olivia Brown","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","MACHINE LEARNING FOR STAGGERED DIFFERENCE-IN-DIFFERENCES AND DYNAMIC TREATMENT EFFECT HETEROGENEITY","We combine two recently proposed nonparametric difference-in-differences approaches and extend them to assess treatment effect heterogeneity under staggered adoption using machine learning. The resulting method, machine learning difference-in-differences (MLDID), estimates time-varying conditional average treatment effects on the treated, enabling detailed inference on drivers of heterogeneity. Simulation results show accurate recovery of true predictors. An application to Brazil’s Family Health Program links faster impacts for subgroups in poverty and in urban locations to dynamic heterogeneous effects on infant mortality.","arXiv :2310 . 11962v1 [ econ .EM] 18 Oct 2023  \nMACHINE LEARNING FOR STAGGERED  \nDIFFERENCE-IN-DIFFERENCES AND DYNAMIC TREATMENT  \nEFFECT HETEROGENEITY  \nJULIA HATAMYAR, NOEMI KREIF, RUDI ROCHA, AND MARTIN HUBER  \nAbstract. We combine two recently proposed nonparametric difference-in-differences methods, extending them to enable the examination of treatment effect heterogeneity in the staggered adoption setting using machine learning. The proposed method, machine learning difference-in-differences (MLDID), allows for estimation of time-varying conditional average treatment effects on the treated, which can be used to conduct detailed inference on drivers of treatment effect heterogeneity. We perform simulations to evaluate the performance of MLDID and find that it accurately identifies the true predictors of treatment effect heterogeneity. We then use MLDID to evaluate the heterogeneous impacts of Brazil’s Family Health Program on infant mortality, and find those in poverty and urban locations experienced the impact of the policy more quickly than other subgroups.  \n1. Introduction  \nDifference-in-differences (DID) with staggered treatment adoption, a popular econometric approach to estimating dynamic treatment effects, is a powerful tool for estimating the causal effects of policies or programs in real-world settings. Staggered implementation of treatment can provide a useful quasi-experimental design to estimate the causal effect of a policy by allowing the researcher to control for confounders that change over time. Estimators that rely on the staggered DID design 1 require the critical assumption that the joint distribution of treatment and covariates and the timing of treatment implementation are independent (Callaway and Sant’Anna 2021) . In other words, the timing of the treatment adoption should not be related to other factors that are also affecting the outcome being examined.  \nDate: October 19, 2023 .  \nJulia Hatamyar, Corresponding Author [e-mail: julia.hatamyar@york.ac.uk. All](e-mail: julia.hatamyar@york.ac.uk. All) code used in this paper is  \navailable at [https://github.com/jhatamyar/MLDID](https://github.com/jhatamyar/MLDID).  \nThis work was funded by the UK Medical Research Council (Grant \\#: MR/T04487X/1) .  \n1See A. C. Baker, Larcker, and Wang (2022) for an overview of various estimators for use with staggered DID  \n2 MACHINE LEARNING & STAGGERED DID  \nThis assumption may not hold in many important observational settings, for example when disadvantaged areas are targeted for implementation first.  \nIn this paper, we create a novel and useful extension of existing methods for estimating dynamic treatment effects by combining the nonparametric machine learning-based (ML) DID estimator proposed by Lu, Nie, and Wager (2019), with a staggered adoption framework as in Callaway and Sant’Anna (2021) . We refer to our extension as MLDID. One key strength of our method is inherited from Lu, Nie, and Wager (2019), and allows for treatment and timing of treatment to be correlated.2 This is done by combining ML estimates of various nuisance models (e.g. the outcome model, the propensity score for entering a treatment group, among others unique to the DID setting) in an appropriate way. Our proposed method has another major advantage: due to the incorporation of ML, it allows for the investigation of treatment effect heterogeneity in a way that allows for more variables and is more flexible than the traditional approach of adding interaction terms in the analysis or subsetting the data by particular covariate values. By combining ML estimates of nuisance models, and then aggregating them appropriately, we can obtain predictions of dynamic Conditional Average Treatment Effects on the Treated (CATTs) for each unit of observation,(i.e., CATTs for each time period post-treatment) . We then propose for these predictions tobe used for inference on dynamic treatment effect heterogeneity in a data-driven manner, for example reg","cbCaia4yyr6VPqoX","https://ap.wps.com/l/cbCaia4yyr6VPqoX","pdf",688313,1,43,"English","en",105,"# Introduction\n## Difference-in-differences with staggered adoption\n## Method extension via MLDID\n# Applications and evidence","[{\"question\":\"What problem does MLDID address in staggered difference-in-differences designs?\",\"answer\":\"MLDID extends staggered DID to enable estimation of dynamic treatment effect heterogeneity using machine learning, rather than relying only on traditional interaction terms or subgrouping.\"},{\"question\":\"How does MLDID obtain time-varying treatment effects?\",\"answer\":\"It combines machine-learning estimates of multiple nuisance models and aggregates them to produce predictions of dynamic conditional average treatment effects on the treated (CATTs) for each post-treatment time period.\"},{\"question\":\"What do the simulations and the Brazil Family Health Program application show?\",\"answer\":\"Simulations indicate small estimation error and correct identification of predictors of heterogeneity. In the Brazil application, subgroups in poverty and urban areas experience the policy impact more quickly than other groups.\"}]","MACHINE LEARNING FOR STAGGERED DIFFERENCE-IN-DIFFERENCES AND DYNAMIC TREATMENT EFFECT HETEROGENEITY | PDF",1785732284,108,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"machine-learning-for-staggered-difference-in-differences-and-dynamic-treatment-effect-heterogeneity","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/machine-learning-for-staggered-difference-in-differences-and-dynamic-treatment-effect-heterogeneity/120838/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does MLDID address in staggered difference-in-differences designs?","Question",{"text":75,"@type":76},"MLDID extends staggered DID to enable estimation of dynamic treatment effect heterogeneity using machine learning, rather than relying only on traditional interaction terms or subgrouping.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does MLDID obtain time-varying treatment effects?",{"text":80,"@type":76},"It combines machine-learning estimates of multiple nuisance models and aggregates them to produce predictions of dynamic conditional average treatment effects on the treated (CATTs) for each post-treatment time period.",{"name":82,"@type":73,"acceptedAnswer":83},"What do the simulations and the Brazil Family Health Program application show?",{"text":84,"@type":76},"Simulations indicate small estimation error and correct identification of predictors of heterogeneity. In the Brazil application, subgroups in poverty and urban areas experience the policy impact more quickly than other groups.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]