Double Machine Learning and Automated Confounder Selection: A Cautionary Tale

Paul Hünermund; Beyers Louw; Itamar Caspi

doi:10.5465/AMBPP.2022.14311abstract

Double Machine Learning and Automated Confounder Selection: A Cautionary Tale

Paul Hünermund, Beyers Louw, Itamar Caspi

Department of Strategy and Innovation

Research output: Chapter in Book/Report/Conference proceeding › Conference abstract in proceedings › Research › peer-review

Abstract

Double machine learning (DML) is becoming an increasingly popular tool for automated model selection in high-dimensional settings. These approaches rely on the assumption of conditional independence, which may not hold in big-data settings where the covariate space is large. This paper shows that DML is very sensitive to the inclusion of even a few "bad controls" in the covariate space. The resulting bias varies with the nature of the causal model, which raises concerns about the feasibility of selecting control variables in a data-driven way.

Original language	English
Title of host publication	Proceedings of the Eighty-second Annual Meeting of the Academy of Management
Editors	Sonia Taneja
Number of pages	1
Place of Publication	Briarcliff Manor, NY
Publisher	Academy of Management
Publication date	2022
Pages	2199
DOIs	https://doi.org/10.5465/AMBPP.2022.14311abstract
Publication status	Published - 2022
Event	The Academy of Management Annual Meeting 2022: Creating a Better World Together - Seattle, United States Duration: 5 Aug 2022 → 9 Aug 2022 Conference number: 82 https://2022.aom.org/

Conference

Conference	The Academy of Management Annual Meeting 2022
Number	82
Country/Territory	United States
City	Seattle
Period	05/08/2022 → 09/08/2022
Internet address	https://2022.aom.org/

Series	Academy of Management Proceedings
ISSN	0065-0668

Access to Document

10.5465/AMBPP.2022.14311abstractLicence: Unspecified

Handle.net

Persistent link

Cite this

@inbook{efcd4feaf45246e6bc42c7a1075d6b81,

title = "Double Machine Learning and Automated Confounder Selection: A Cautionary Tale",

abstract = "Double machine learning (DML) is becoming an increasingly popular tool for automated model selection in high-dimensional settings. These approaches rely on the assumption of conditional independence, which may not hold in big-data settings where the covariate space is large. This paper shows that DML is very sensitive to the inclusion of even a few {"}bad controls{"} in the covariate space. The resulting bias varies with the nature of the causal model, which raises concerns about the feasibility of selecting control variables in a data-driven way.",

author = "Paul H{\"u}nermund and Beyers Louw and Itamar Caspi",

year = "2022",

doi = "10.5465/AMBPP.2022.14311abstract",

language = "English",

series = "Academy of Management Proceedings",

publisher = "Academy of Management",

pages = "2199",

editor = "Sonia Taneja",

booktitle = "Proceedings of the Eighty-second Annual Meeting of the Academy of Management",

address = "United States",

note = "The Academy of Management Annual Meeting 2022 : Creating a Better World Together, AoM 2022 ; Conference date: 05-08-2022 Through 09-08-2022",

url = "https://2022.aom.org/",

}

Hünermund, P, Louw, B & Caspi, I 2022, Double Machine Learning and Automated Confounder Selection: A Cautionary Tale. in S Taneja (ed.), Proceedings of the Eighty-second Annual Meeting of the Academy of Management . Academy of Management, Briarcliff Manor, NY, Academy of Management Proceedings, pp. 2199, The Academy of Management Annual Meeting 2022, Seattle, Washington, United States, 05/08/2022. https://doi.org/10.5465/AMBPP.2022.14311abstract

Double Machine Learning and Automated Confounder Selection: A Cautionary Tale. / Hünermund, Paul; Louw, Beyers; Caspi, Itamar.
Proceedings of the Eighty-second Annual Meeting of the Academy of Management . ed. / Sonia Taneja. Briarcliff Manor, NY: Academy of Management, 2022. p. 2199 (Academy of Management Proceedings).

Research output: Chapter in Book/Report/Conference proceeding › Conference abstract in proceedings › Research › peer-review

TY - ABST

T1 - Double Machine Learning and Automated Confounder Selection

T2 - The Academy of Management Annual Meeting 2022

AU - Hünermund, Paul

AU - Louw, Beyers

AU - Caspi, Itamar

N1 - Conference code: 82

PY - 2022

Y1 - 2022

N2 - Double machine learning (DML) is becoming an increasingly popular tool for automated model selection in high-dimensional settings. These approaches rely on the assumption of conditional independence, which may not hold in big-data settings where the covariate space is large. This paper shows that DML is very sensitive to the inclusion of even a few "bad controls" in the covariate space. The resulting bias varies with the nature of the causal model, which raises concerns about the feasibility of selecting control variables in a data-driven way.

AB - Double machine learning (DML) is becoming an increasingly popular tool for automated model selection in high-dimensional settings. These approaches rely on the assumption of conditional independence, which may not hold in big-data settings where the covariate space is large. This paper shows that DML is very sensitive to the inclusion of even a few "bad controls" in the covariate space. The resulting bias varies with the nature of the causal model, which raises concerns about the feasibility of selecting control variables in a data-driven way.

U2 - 10.5465/AMBPP.2022.14311abstract

DO - 10.5465/AMBPP.2022.14311abstract

M3 - Conference abstract in proceedings

T3 - Academy of Management Proceedings

SP - 2199

BT - Proceedings of the Eighty-second Annual Meeting of the Academy of Management

A2 - Taneja, Sonia

PB - Academy of Management

CY - Briarcliff Manor, NY

Y2 - 5 August 2022 through 9 August 2022

ER -