IDEAS home Printed from https://ideas.repec.org/p/arx/papers/2609.19418.html

Stable Policy Learning

Author

Listed:
  • Harvey Barnhard
  • Giacomo Opocher
  • Rahul Singh

Abstract

In evidence-based policymaking, typically one experimental sample is observed, then a learned policy recommendation is implemented at scale. Policies learned from the experimental data can perform well in expected welfare, yet random sampling in the experiment can produce recommendations with poor welfare outcomes. In this paper, we ask: how should policy learning algorithms balance expected welfare against sampling risk? Our main contribution is to show that algorithmic stability plays a central role in characterizing and navigating the tradeoff. Intuitively, if a policy learning algorithm's recommendation remains stable when one experimental unit is replaced, then that algorithm has limited sampling risk. We propose a method for policy learning called policy-vote bagging, which learns treatment decisions on many subsamples then averages their votes into treatment probabilities. Relative to using one subsample, averaging across subsamples preserves expected welfare and improves expected utility for a risk-averse researcher. We derive sharp bounds linking estimation accuracy, subsample size, and welfare variation, including an exact guarantee under CARA utility.

Suggested Citation

  • Harvey Barnhard & Giacomo Opocher & Rahul Singh, 2026. "Stable Policy Learning," Papers 2609.19418, arXiv.org.
  • Handle: RePEc:arx:papers:2609.19418
    as

    Download full text from publisher

    File URL: https://arxiv.org/pdf/2609.19418
    File Function: Latest version
    Download Restriction: no
    ---><---

    References listed on IDEAS

    as
    1. Tomasz Strzalecki, 2011. "Axiomatic Foundations of Multiplier Preferences," Econometrica, Econometric Society, vol. 79(1), pages 47-73, January.
    2. Stefan Wager & Susan Athey, 2018. "Estimation and Inference of Heterogeneous Treatment Effects using Random Forests," Journal of the American Statistical Association, Taylor & Francis Journals, vol. 113(523), pages 1228-1242, July.
    3. Fabio Maccheroni & Massimo Marinacci & Aldo Rustichini, 2006. "Ambiguity Aversion, Robustness, and the Variational Representation of Preferences," Econometrica, Econometric Society, vol. 74(6), pages 1447-1498, November.
    4. Andrews, Donald W K, 1986. "Stability Comparisons of Estimators," Econometrica, Econometric Society, vol. 54(5), pages 1207-1235, September.
    5. Davide Viviano & Jelena Bradic, 2024. "Fair Policy Targeting," Journal of the American Statistical Association, Taylor & Francis Journals, vol. 119(545), pages 730-743, January.
    6. Toru Kitagawa & Aleksey Tetenov, 2021. "Equality-Minded Treatment Choice," Journal of Business & Economic Statistics, Taylor & Francis Journals, vol. 39(2), pages 561-574, March.
    7. Eric Mbakop & Max Tabord‐Meehan, 2021. "Model Selection for Treatment Choice: Penalized Welfare Maximization," Econometrica, Econometric Society, vol. 89(2), pages 825-848, March.
    8. Toru Kitagawa & Aleksey Tetenov, 2018. "Who Should Be Treated? Empirical Welfare Maximization Methods for Treatment Choice," Econometrica, Econometric Society, vol. 86(2), pages 591-616, March.
    9. Giacomo Opocher, 2026. "Producing Policy Recommendations: from Statistical Decision Theory to Empirical Practice," Papers 2607.29281, arXiv.org, revised Aug 2026.
    10. Fabio Maccheroni & Massimo Marinacci & Doriana Ruffino, 2013. "Alpha as Ambiguity: Robust Mean‐Variance Portfolio Analysis," Econometrica, Econometric Society, vol. 81(3), pages 1075-1113, May.
    11. Victor Chernozhukov & Sokbae Lee & Adam M. Rosen & Liyang Sun, 2025. "Policy Learning with Confidence," Papers 2502.10653, arXiv.org, revised Oct 2026.
    12. Paolo Ghirardato & Fabio Maccheroni & Massimo Marinacci & Marciano Siniscalchi, 2003. "A Subjective Spin on Roulette Wheels," Econometrica, Econometric Society, vol. 71(6), pages 1897-1908, November.
    13. Victor Chernozhukov & Whitney K. Newey & Rahul Singh & Vasilis Syrgkanis, 2026. "Adversarial Estimation of Riesz Representers," Journal of the American Statistical Association, Taylor & Francis Journals, vol. 121(554), pages 1398-1409, April.
    14. Sun, Liyang, 2026. "Empirical welfare maximization with constraints," Journal of Econometrics, Elsevier, vol. 253(C).
    15. Susan Athey & Stefan Wager, 2021. "Policy Learning With Observational Data," Econometrica, Econometric Society, vol. 89(1), pages 133-161, January.
    16. Peter Klibanoff & Massimo Marinacci & Sujoy Mukerji, 2005. "A Smooth Model of Decision Making under Ambiguity," Econometrica, Econometric Society, vol. 73(6), pages 1849-1892, November.
    17. Charles F. Manski, 2004. "Statistical Treatment Rules for Heterogeneous Populations," Econometrica, Econometric Society, vol. 72(4), pages 1221-1246, July.
    18. Tommaso Denti & Luciano Pomatto, 2022. "Model and Predictive Uncertainty: A Foundation for Smooth Ambiguity Preferences," Econometrica, Econometric Society, vol. 90(2), pages 551-584, March.
    Full references (including those not matched with items on IDEAS)

    Most related items

    These are the items that most often cite the same works as this one and are cited by the same works as this one.
    1. Toru Kitagawa & Sokbae Lee & Chen Qiu, 2022. "Treatment Choice with Nonlinear Regret," Papers 2205.08586, arXiv.org, revised Oct 2024.
    2. Cheaheon Lim & Yechan Park, 2026. "Dynamically Consistent Statistical Decisions," Papers 2607.10519, arXiv.org.
    3. Sarah Moon, 2025. "Optimal Policy Choices Under Uncertainty," Papers 2503.03910, arXiv.org, revised Aug 2026.
    4. Nan Liu & Yanbo Liu & Yuya Sasaki & Yuanyuan Wan, 2025. "Nonparametric Uniform Inference in Binary Classification and Policy Values," Papers 2511.14700, arXiv.org, revised Dec 2025.
    5. Emily Breza & Arun G. Chandrasekhar & Davide Viviano, 2025. "Evidence aggregation with ignorance in mind: learning what we do (not) know for archetypes discovery," Papers 2501.13355, arXiv.org, revised Jun 2026.
    6. Raphael Langevin, 2026. "Policy Learning with Observational Data: The Case of Hepatitis C Treatment for HIV/HCV Co-Infected Patients," Papers 2605.16593, arXiv.org, revised Oct 2026.
    7. Hirano, Keisuke & Porter, Jack R., 2020. "Asymptotic analysis of statistical decision rules in econometrics," Handbook of Econometrics, in: Steven N. Durlauf & Lars Peter Hansen & James J. Heckman & Rosa L. Matzkin (ed.), Handbook of Econometrics, edition 1, volume 7, chapter 0, pages 283-354, Elsevier.
    8. Feng, Kai & Hong, Han & Nekipelov, Denis, 2026. "Statistical inference of optimal allocations I: Regularities and their implications," Journal of Econometrics, Elsevier, vol. 254(PB).
    9. Ai, Chunrong & Fang, Yue & Xie, Haitian, 2026. "Data-driven policy learning for continuous treatments," Journal of Econometrics, Elsevier, vol. 253(C).
    10. Toru Kitagawa & Weining Wang & Mengshan Xu, 2022. "Policy Choice in Time Series by Empirical Welfare Maximization," Papers 2205.03970, arXiv.org, revised Jun 2026.
    11. André, Eric, 2016. "Crisp monetary acts in multiple-priors models of decision under ambiguity," Journal of Mathematical Economics, Elsevier, vol. 67(C), pages 153-161.
    12. Alejandro Sanchez-Becerra, 2023. "Robust inference for the treatment effect variance in experiments using machine learning," Papers 2306.03363, arXiv.org.
    13. Toru Kitagawa & Weining Wang & Mengshan Xu, 2024. "Policy choice in time series by empirical welfare maximization," CeMMAP working papers 27/24, Institute for Fiscal Studies.
    14. Hugo Bodory & Federica Mascolo & Michael Lechner, 2024. "Enabling Decision-Making with the Modified Causal Forest: Policy Trees for Treatment Assignment," Papers 2406.02241, arXiv.org.
    15. Timothy B. Armstrong & Martin Weidner & Andrei Zeleneev, 2022. "Robust Estimation and Inference in Panels with Interactive Fixed Effects," Papers 2210.06639, arXiv.org, revised May 2025.
    16. Toru Kitagawa & Sokbae Lee & Chen Qiu, 2023. "Treatment choice, mean square regret and partial identification," The Japanese Economic Review, Springer, vol. 74(4), pages 573-602, October.
    17. Sun, Liyang, 2026. "Empirical welfare maximization with constraints," Journal of Econometrics, Elsevier, vol. 253(C).
    18. Davide Viviano, 2019. "Policy Targeting under Network Interference," Papers 1906.10258, arXiv.org, revised Apr 2024.
    19. Daniel F. Pellatt, 2022. "PAC-Bayesian Treatment Allocation Under Budget Constraints," Papers 2212.09007, arXiv.org, revised Jun 2023.
    20. Kock, Anders Bredahl & Preinerstorfer, David, 2026. "Regularizing fairness in optimal policy learning with distributional targets," Journal of Econometrics, Elsevier, vol. 254(PB).

    More about this item

    NEP fields

    This paper has been announced in the following NEP Reports:

    Statistics

    Access and download statistics

    Corrections

    All material on this site has been provided by the respective publishers and authors. You can help correct errors and omissions. When requesting a correction, please mention this item's handle: RePEc:arx:papers:2609.19418. See general information about how to correct material in RePEc.

    If you have authored this item and are not yet registered with RePEc, we encourage you to do it here. This allows to link your profile to this item. It also allows you to accept potential citations to this item that we are uncertain about.

    If CitEc recognized a bibliographic reference but did not link an item in RePEc to it, you can help with this form .

    If you know of missing items citing this one, you can help us creating those links by adding the relevant references in the same way as above, for each refering item. If you are a registered author of this item, you may also want to check the "citations" tab in your RePEc Author Service profile, as there may be some citations waiting for confirmation.

    For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: arXiv administrators (email available below). General contact details of provider: https://arxiv.org/ .

    Please note that corrections may take a couple of weeks to filter through the various RePEc services.

    IDEAS is a RePEc service. RePEc uses bibliographic data supplied by the respective publishers.