Adaptive experimental design using the propensity score

Jinyong Hahn, Keisuke Hirano, Dean Karlan

Research output: Contribution to journalArticle

8 Citations (Scopus)

Abstract

Many social experiments are run in multiple waves or replicate earlier social experiments. In principle, the sampling design can be modified in later stages or replications to allow for more efficient estimation of causal effects. We consider the design of a two-stage experiment for estimating an average treatment effect when covariate information is available for experimental subjects. We use data from the first stage to choose a conditional treatment assignment rule for units in the second stage of the experiment. This amounts to choosing the propensity score, the conditional probability of treatment given covariates. We propose to select the propensity score to minimize the asymptotic variance bound for estimating the average treatment effect. Our procedure can be implemented simply using standard statistical software and has attractive large-sample properties.

Original languageEnglish (US)
Pages (from-to)96-108
Number of pages13
JournalJournal of Business and Economic Statistics
Volume29
Issue number1
DOIs
StatePublished - Jan 1 2011

Fingerprint

Propensity Score
Experimental design
Average Treatment Effect
experiment
Experiment
Covariates
Variance Bounds
Causal Effect
Statistical Software
Sampling Design
Efficient Estimation
Asymptotic Variance
Conditional probability
available information
Replication
Assignment
Choose
Minimise
Unit
Average treatment effect

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Social Sciences (miscellaneous)
  • Economics and Econometrics
  • Statistics, Probability and Uncertainty

Cite this

@article{d323d17ad4d34d8882e79e3f69ecab33,
title = "Adaptive experimental design using the propensity score",
abstract = "Many social experiments are run in multiple waves or replicate earlier social experiments. In principle, the sampling design can be modified in later stages or replications to allow for more efficient estimation of causal effects. We consider the design of a two-stage experiment for estimating an average treatment effect when covariate information is available for experimental subjects. We use data from the first stage to choose a conditional treatment assignment rule for units in the second stage of the experiment. This amounts to choosing the propensity score, the conditional probability of treatment given covariates. We propose to select the propensity score to minimize the asymptotic variance bound for estimating the average treatment effect. Our procedure can be implemented simply using standard statistical software and has attractive large-sample properties.",
author = "Jinyong Hahn and Keisuke Hirano and Dean Karlan",
year = "2011",
month = "1",
day = "1",
doi = "10.1198/jbes.2009.08161",
language = "English (US)",
volume = "29",
pages = "96--108",
journal = "Journal of Business and Economic Statistics",
issn = "0735-0015",
publisher = "American Statistical Association",
number = "1",

}

Adaptive experimental design using the propensity score. / Hahn, Jinyong; Hirano, Keisuke; Karlan, Dean.

In: Journal of Business and Economic Statistics, Vol. 29, No. 1, 01.01.2011, p. 96-108.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Adaptive experimental design using the propensity score

AU - Hahn, Jinyong

AU - Hirano, Keisuke

AU - Karlan, Dean

PY - 2011/1/1

Y1 - 2011/1/1

N2 - Many social experiments are run in multiple waves or replicate earlier social experiments. In principle, the sampling design can be modified in later stages or replications to allow for more efficient estimation of causal effects. We consider the design of a two-stage experiment for estimating an average treatment effect when covariate information is available for experimental subjects. We use data from the first stage to choose a conditional treatment assignment rule for units in the second stage of the experiment. This amounts to choosing the propensity score, the conditional probability of treatment given covariates. We propose to select the propensity score to minimize the asymptotic variance bound for estimating the average treatment effect. Our procedure can be implemented simply using standard statistical software and has attractive large-sample properties.

AB - Many social experiments are run in multiple waves or replicate earlier social experiments. In principle, the sampling design can be modified in later stages or replications to allow for more efficient estimation of causal effects. We consider the design of a two-stage experiment for estimating an average treatment effect when covariate information is available for experimental subjects. We use data from the first stage to choose a conditional treatment assignment rule for units in the second stage of the experiment. This amounts to choosing the propensity score, the conditional probability of treatment given covariates. We propose to select the propensity score to minimize the asymptotic variance bound for estimating the average treatment effect. Our procedure can be implemented simply using standard statistical software and has attractive large-sample properties.

UR - http://www.scopus.com/inward/record.url?scp=78449246035&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78449246035&partnerID=8YFLogxK

U2 - 10.1198/jbes.2009.08161

DO - 10.1198/jbes.2009.08161

M3 - Article

VL - 29

SP - 96

EP - 108

JO - Journal of Business and Economic Statistics

JF - Journal of Business and Economic Statistics

SN - 0735-0015

IS - 1

ER -