Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping Support

When estimating treatment effects, the gold standard is to conduct a randomized experiment and then contrast outcomes associated with the treatment group and the control group. However, in many cases, randomized experiments are either conducted with a much smaller scale compared to the size of the t...

Full description

Saved in:
Bibliographic Details
Main Authors: Kevin Han, Han Wu, Linjia Wu, Yu Shi, Canyao Liu
Format: Article
Language:English
Published: MDPI AG 2024-09-01
Series:Econometrics
Subjects:
Online Access:https://www.mdpi.com/2225-1146/12/3/26
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850261167123988480
author Kevin Han
Han Wu
Linjia Wu
Yu Shi
Canyao Liu
author_facet Kevin Han
Han Wu
Linjia Wu
Yu Shi
Canyao Liu
author_sort Kevin Han
collection DOAJ
description When estimating treatment effects, the gold standard is to conduct a randomized experiment and then contrast outcomes associated with the treatment group and the control group. However, in many cases, randomized experiments are either conducted with a much smaller scale compared to the size of the target population or accompanied with certain ethical issues and thus hard to implement. Therefore, researchers usually rely on observational data to study causal connections. The downside is that the unconfoundedness assumption, which is the key to validating the use of observational data, is untestable and almost always violated. Hence, any conclusion drawn from observational data should be further analyzed with great care. Given the richness of observational data and usefulness of experimental data, researchers hope to develop credible methods to combine the strength of the two. In this paper, we consider a setting where the observational data contain the outcome of interest as well as a surrogate outcome, while the experimental data contain only the surrogate outcome. We propose an easy-to-implement estimator to estimate the average treatment effect of interest using both the observational data and the experimental data.
format Article
id doaj-art-ca3d75a40d214d03a05af662d077919b
institution OA Journals
issn 2225-1146
language English
publishDate 2024-09-01
publisher MDPI AG
record_format Article
series Econometrics
spelling doaj-art-ca3d75a40d214d03a05af662d077919b2025-08-20T01:55:30ZengMDPI AGEconometrics2225-11462024-09-011232610.3390/econometrics12030026Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping SupportKevin Han0Han Wu1Linjia Wu2Yu Shi3Canyao Liu4Department of Statistics, Stanford University, Stanford, CA 94305, USADepartment of Statistics, Stanford University, Stanford, CA 94305, USADepartment of Management Science and Engineering, Stanford University, Stanford, CA 94305, USAYale School of Management, Yale University, New Haven, CT 06511, USAYale School of Management, Yale University, New Haven, CT 06511, USAWhen estimating treatment effects, the gold standard is to conduct a randomized experiment and then contrast outcomes associated with the treatment group and the control group. However, in many cases, randomized experiments are either conducted with a much smaller scale compared to the size of the target population or accompanied with certain ethical issues and thus hard to implement. Therefore, researchers usually rely on observational data to study causal connections. The downside is that the unconfoundedness assumption, which is the key to validating the use of observational data, is untestable and almost always violated. Hence, any conclusion drawn from observational data should be further analyzed with great care. Given the richness of observational data and usefulness of experimental data, researchers hope to develop credible methods to combine the strength of the two. In this paper, we consider a setting where the observational data contain the outcome of interest as well as a surrogate outcome, while the experimental data contain only the surrogate outcome. We propose an easy-to-implement estimator to estimate the average treatment effect of interest using both the observational data and the experimental data.https://www.mdpi.com/2225-1146/12/3/26causal inferencetreatment effectsobservational studiessurrogate outcomesunconfoundedness
spellingShingle Kevin Han
Han Wu
Linjia Wu
Yu Shi
Canyao Liu
Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping Support
Econometrics
causal inference
treatment effects
observational studies
surrogate outcomes
unconfoundedness
title Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping Support
title_full Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping Support
title_fullStr Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping Support
title_full_unstemmed Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping Support
title_short Estimating Treatment Effects Using Observational Data and Experimental Data with Non-Overlapping Support
title_sort estimating treatment effects using observational data and experimental data with non overlapping support
topic causal inference
treatment effects
observational studies
surrogate outcomes
unconfoundedness
url https://www.mdpi.com/2225-1146/12/3/26
work_keys_str_mv AT kevinhan estimatingtreatmenteffectsusingobservationaldataandexperimentaldatawithnonoverlappingsupport
AT hanwu estimatingtreatmenteffectsusingobservationaldataandexperimentaldatawithnonoverlappingsupport
AT linjiawu estimatingtreatmenteffectsusingobservationaldataandexperimentaldatawithnonoverlappingsupport
AT yushi estimatingtreatmenteffectsusingobservationaldataandexperimentaldatawithnonoverlappingsupport
AT canyaoliu estimatingtreatmenteffectsusingobservationaldataandexperimentaldatawithnonoverlappingsupport