Distributional robustness of K-class estimators and the PULSE

Research output: Contribution to journalJournal articlepeer-review

Standard

Distributional robustness of K-class estimators and the PULSE. / Jakobsen, Martin Emil; Peters, Jonas.

In: The Econometrics Journal, Vol. 25, No. 2, 2022, p. 404–432.

Research output: Contribution to journalJournal articlepeer-review

Harvard

Jakobsen, ME & Peters, J 2022, 'Distributional robustness of K-class estimators and the PULSE', The Econometrics Journal, vol. 25, no. 2, pp. 404–432. https://doi.org/10.1093/ectj/utab031

APA

Jakobsen, M. E., & Peters, J. (2022). Distributional robustness of K-class estimators and the PULSE. The Econometrics Journal, 25(2), 404–432. https://doi.org/10.1093/ectj/utab031

Vancouver

Jakobsen ME, Peters J. Distributional robustness of K-class estimators and the PULSE. The Econometrics Journal. 2022;25(2):404–432. https://doi.org/10.1093/ectj/utab031

Author

Jakobsen, Martin Emil ; Peters, Jonas. / Distributional robustness of K-class estimators and the PULSE. In: The Econometrics Journal. 2022 ; Vol. 25, No. 2. pp. 404–432.

Bibtex

@article{503249bf091e4f3f83b1d91d7b98841e,
title = "Distributional robustness of K-class estimators and the PULSE",
abstract = "While causal models are robust in that they are prediction optimal under arbitrarily strong interventions, they may not be optimal when the interventions are bounded. We prove that the classical K-class estimator satisfies such optimality by establishing a connection between K-class estimators and anchor regression. This connection further motivates a novel estimator in instrumental variable settings that minimizes the mean squared prediction error subject to the constraint that the estimator lies in an asymptotically valid confidence region of the causal coefficient. We call this estimator PULSE (p-uncorrelated least squares estimator), relate it to work on invariance, show that it can be computed efficiently, as a data-driven K-class estimator, even though the underlying optimization problem is nonconvex, and prove consistency. We evaluate the estimators on real data and perform simulation experiments illustrating that PULSE suffers from less variability. There are several settings, including weak instrument settings, where it outperforms other estimators.",
author = "Jakobsen, {Martin Emil} and Jonas Peters",
year = "2022",
doi = "10.1093/ectj/utab031",
language = "English",
volume = "25",
pages = "404–432",
journal = "Econometrics Journal",
issn = "1368-4221",
publisher = "Wiley",
number = "2",

}

RIS

TY - JOUR

T1 - Distributional robustness of K-class estimators and the PULSE

AU - Jakobsen, Martin Emil

AU - Peters, Jonas

PY - 2022

Y1 - 2022

N2 - While causal models are robust in that they are prediction optimal under arbitrarily strong interventions, they may not be optimal when the interventions are bounded. We prove that the classical K-class estimator satisfies such optimality by establishing a connection between K-class estimators and anchor regression. This connection further motivates a novel estimator in instrumental variable settings that minimizes the mean squared prediction error subject to the constraint that the estimator lies in an asymptotically valid confidence region of the causal coefficient. We call this estimator PULSE (p-uncorrelated least squares estimator), relate it to work on invariance, show that it can be computed efficiently, as a data-driven K-class estimator, even though the underlying optimization problem is nonconvex, and prove consistency. We evaluate the estimators on real data and perform simulation experiments illustrating that PULSE suffers from less variability. There are several settings, including weak instrument settings, where it outperforms other estimators.

AB - While causal models are robust in that they are prediction optimal under arbitrarily strong interventions, they may not be optimal when the interventions are bounded. We prove that the classical K-class estimator satisfies such optimality by establishing a connection between K-class estimators and anchor regression. This connection further motivates a novel estimator in instrumental variable settings that minimizes the mean squared prediction error subject to the constraint that the estimator lies in an asymptotically valid confidence region of the causal coefficient. We call this estimator PULSE (p-uncorrelated least squares estimator), relate it to work on invariance, show that it can be computed efficiently, as a data-driven K-class estimator, even though the underlying optimization problem is nonconvex, and prove consistency. We evaluate the estimators on real data and perform simulation experiments illustrating that PULSE suffers from less variability. There are several settings, including weak instrument settings, where it outperforms other estimators.

U2 - 10.1093/ectj/utab031

DO - 10.1093/ectj/utab031

M3 - Journal article

VL - 25

SP - 404

EP - 432

JO - Econometrics Journal

JF - Econometrics Journal

SN - 1368-4221

IS - 2

ER -

ID: 304486364