Suboptimal Provision of Privacy and Statistical Accuracy When They are Public Goods

Ian M. Schmutte; John M. Abowd; Lars Vilhuber; William Sexton

arxiv: 1906.09353 · v1 · pith:UDM27JLGnew · submitted 2019-06-21 · 💰 econ.TH · cs.CR· cs.DB

Suboptimal Provision of Privacy and Statistical Accuracy When They are Public Goods

John M. Abowd , Ian M. Schmutte , William Sexton , Lars Vilhuber This is my paper

Pith reviewed 2026-05-25 17:56 UTC · model grok-4.3

classification 💰 econ.TH cs.CRcs.DB

keywords differential privacypublic goodsstatistical accuracyprivacy protectionprivate provisiondata qualitypopulation statisticsincentives

0 comments

The pith

Private firms that release population statistics under differential privacy produce inefficiently low data quality.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Private technology companies hold large databases and can release population statistics that compete with those from public agencies. Privacy protection and statistical accuracy both act as public goods, so private firms face profit incentives that diverge from social goals. The paper builds a model of a firm that publishes statistics while satisfying a differential privacy guarantee. Under profit maximization the firm selects a privacy-accuracy tradeoff that delivers data quality below the social optimum. This result matters because widespread private statistics could leave society with systematically less accurate information than would be chosen if the full social value were internalized.

Core claim

In the model a private firm chooses the level of differential privacy and the resulting statistical accuracy to maximize its own profit; because the firm does not capture the external benefits that higher accuracy confers on all users, the equilibrium accuracy lies strictly below the socially efficient level.

What carries the argument

The firm's profit-maximizing choice of the differential privacy parameter, which determines the accuracy of the published statistics without internalizing their public-good value.

If this is right

Private provision results in statistical accuracy that is too low relative to the social optimum.
At least one of the two public goods—privacy or accuracy—will be supplied suboptimally.
Public statistical agencies may continue to hold an efficiency advantage in delivering data quality.
Policy may need to correct the private firm's incentive misalignment to reach efficient outcomes.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Minimum accuracy requirements could be imposed on private statistical products to offset the underprovision.
The same incentive logic may apply to other markets in which firms supply information goods that carry privacy externalities.
Direct measurement of accuracy in private versus public releases on identical populations would test the predicted gap.

Load-bearing premise

The private firm maximizes only its own profit and does not internalize the full social value of privacy protection and statistical accuracy.

What would settle it

An empirical comparison showing that the accuracy of statistics released by private firms equals or exceeds the level that would be chosen by a social planner facing the same differential privacy constraint.

read the original abstract

With vast databases at their disposal, private tech companies can compete with public statistical agencies to provide population statistics. However, private companies face different incentives to provide high-quality statistics and to protect the privacy of the people whose data are used. When both privacy protection and statistical accuracy are public goods, private providers tend to produce at least one suboptimally, but it is not clear which. We model a firm that publishes statistics under a guarantee of differential privacy. We prove that provision by the private firm results in inefficiently low data quality in this framework.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Private firm underprovides accuracy when privacy and accuracy are public goods in this DP model.

read the letter

The core result is that when privacy and statistical accuracy are public goods, a private firm will choose differential privacy parameters that deliver less accuracy than the social optimum. The authors build a model where the firm maximizes profit by selecting how much privacy to provide and what accuracy to deliver. They then compare the firm's choice to what a social planner would do and prove the firm underprovides accuracy. This is new in the sense that it brings the public-goods framework to the specific mechanics of differential privacy parameter choice. The abstract presents it as a proof within their framework, and the stress-test finds no internal inconsistency. The paper does a solid job of keeping the model simple enough that the inefficiency comes through clearly from the incentive misalignment. The proof appears to follow without circularity or hidden steps, and it treats the public goods nature directly. The soft spots are that the result depends on the firm having no incentive to internalize the public benefits, which is the central assumption but could be sensitive to how the profit function is specified or if there are repeated interactions. Since this is a theoretical paper, the magnitude isn't quantified. The assumption that the private firm faces profit-maximizing incentives that do not internalize the full social value is standard for this type of analysis but might not capture all real-world factors like brand reputation. This is useful for economists studying the division of labor between public statistical agencies and private data providers. A reader focused on privacy economics or information policy would get the most from it. The work is coherent enough that it should go to peer review rather than be desk rejected. It shows clear thinking on the topic.

Referee Report

0 major / 2 minor

Summary. The paper constructs a model of a profit-maximizing firm that publishes population statistics subject to a differential privacy guarantee. It proves that the firm's incentive-compatible choice of privacy parameters and resulting statistical accuracy is strictly below the social planner's optimum when both privacy protection and statistical accuracy are non-rival, non-excludable public goods.

Significance. If the central proof is correct, the result supplies a clean theoretical mechanism for why private firms under-supply data quality relative to a social optimum in settings where privacy and accuracy are public goods. This is directly relevant to policy debates on private versus public statistical agencies and to the design of privacy regulations that affect data publication incentives. The use of differential privacy as the formal privacy notion is a strength, as it allows the model to be stated in terms of a well-defined privacy-accuracy tradeoff.

minor comments (2)

The abstract states the main result but does not indicate the functional forms or the precise definition of 'data quality' used in the proof; adding a one-sentence preview of the key functional forms would improve readability.
The manuscript would benefit from an explicit statement of the planner's problem (e.g., as a numbered equation) immediately before the proof that the firm's solution is strictly inferior.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the supportive summary of the paper and for recommending minor revision. The assessment correctly captures the model's focus on a profit-maximizing firm choosing differential privacy parameters and the resulting statistical accuracy when both are public goods.

Circularity Check

0 steps flagged

Theoretical model of public-goods underprovision; derivation self-contained

full rationale

The paper constructs an explicit economic model in which a profit-maximizing firm chooses differential-privacy parameters while both privacy protection and statistical accuracy are treated as public goods. It then compares the firm's equilibrium to the social planner's optimum and proves underprovision of data quality. This comparison follows directly from the stated incentive assumptions (firm does not internalize full social value) without any fitted parameters, self-referential definitions, or load-bearing self-citations. The result is a standard welfare comparison inside the model and does not reduce to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no identifiable free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5626 in / 890 out tokens · 32241 ms · 2026-05-25T17:56:41.097542+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages

[1]

Abowd, J. M. and Schmutte, I. M. (2019). An economic analysis of privacy protection and statistical accuracy as social choices, American Economic Review 109(1): 171–202. Acquisti, A., Taylor, C. and Wagman, L. (2016). The economic s of privacy ,Journal of Economic Literature 54(2): 442–492. URL: http://www.aeaweb.org/articles?id=10.1257/jel.54.2.442 Anten...

work page doi:10.1257/jel.54.2.442 2019
[2]

URL: http://doi.acm.org/10.1145/773153.773173 Dwork, C. (2006). Differential privacy , Proceedings of the 33rd International Collo- quium on Automata, Languages and Programming pp. 1–12. URL: N/A Dwork, C., McSherry , F., Nissim, K. and Smith, A. (2006). Cal ibrating Noise to Sensitivity in Private Data Analysis, Proceedings of the Third conference on The...

work page doi:10.1145/773153.773173 2006
[3]

and Smith, A

URL: https://link.springer.com/chapter/10.1007/11681878 14 Dwork, C., McSherry , F., Nissim, K. and Smith, A. (2017). Cal ibrating noise to sensitivity in private data analysis, Journal of Privacy and Conﬁdentiality 7(3): 17–

work page doi:10.1007/11681878 2017
[4]

and Roth, A

URL: https://doi.org/10.29012/jpc.v7i3.405 Dwork, C. and Roth, A. (2014). The algorithmic foundations o f differential privacy ,Foundations and T rends in Theoretical Computer Science 9(3-4): 211–407. URL: http://www.nowpublishers.com/articles/foundations-and-trends-in-theoretical- computer-science/TCS-042 Erlingsson, ´U., Pihur, V . and Korolova, A. (201...

work page doi:10.29012/jpc.v7i3.405 2014
[5]

and Roth, A

URL: http://dl.acm.org/citation.cfm?id=2660267.2660348 23 Ghosh, A. and Roth, A. (2015). Selling privacy at auction, Games and Economic Behavior 91: 334–346. URL: https://www.sciencedirect.com/science/article/pii/S0899825613000961 Goldfarb, A., Greenstein, S. M. and Tucker, C. E. (2015). Economic Analysis of the Digital Economy. URL: https://books.google....

work page doi:10.1007/978-3-642-28914-9 2015
[6]

Samuelson, P . A. (1954). The pure theory of public expenditu re, Review of Eco- nomics and Statistics 37: 387–389. Spence, A. M. (1975). Monopoly , quality , and regulation, The Bell Journal of Eco- nomics 6(2): 417–429. URL: http://www.jstor.org/stable/3003237 Wasserman, L. and Zhou, S. (2010). A Statistical Framework f or Differential Pri- vacy ,Journa...

work page arXiv 1954
[7]

bounded” differential privacy . However, this is not the standard deﬁ- nition of ε-differential privacy , which they take from Dwork et al. (20 06), because a “worst-case

So, D and D′ are adjacent databases and x and y are the adjacent histogram representations of D and D′, respectively . Some caution is required when reviewing the related literature because deﬁnitions m ay be stated in terms of adjacent databases or adjacent histograms. A.2 T ranslation of the Ghosh-Roth Model to Our Notation In this appendix we show that...

work page 2015

[1] [1]

Abowd, J. M. and Schmutte, I. M. (2019). An economic analysis of privacy protection and statistical accuracy as social choices, American Economic Review 109(1): 171–202. Acquisti, A., Taylor, C. and Wagman, L. (2016). The economic s of privacy ,Journal of Economic Literature 54(2): 442–492. URL: http://www.aeaweb.org/articles?id=10.1257/jel.54.2.442 Anten...

work page doi:10.1257/jel.54.2.442 2019

[2] [2]

URL: http://doi.acm.org/10.1145/773153.773173 Dwork, C. (2006). Differential privacy , Proceedings of the 33rd International Collo- quium on Automata, Languages and Programming pp. 1–12. URL: N/A Dwork, C., McSherry , F., Nissim, K. and Smith, A. (2006). Cal ibrating Noise to Sensitivity in Private Data Analysis, Proceedings of the Third conference on The...

work page doi:10.1145/773153.773173 2006

[3] [3]

and Smith, A

URL: https://link.springer.com/chapter/10.1007/11681878 14 Dwork, C., McSherry , F., Nissim, K. and Smith, A. (2017). Cal ibrating noise to sensitivity in private data analysis, Journal of Privacy and Conﬁdentiality 7(3): 17–

work page doi:10.1007/11681878 2017

[4] [4]

and Roth, A

URL: https://doi.org/10.29012/jpc.v7i3.405 Dwork, C. and Roth, A. (2014). The algorithmic foundations o f differential privacy ,Foundations and T rends in Theoretical Computer Science 9(3-4): 211–407. URL: http://www.nowpublishers.com/articles/foundations-and-trends-in-theoretical- computer-science/TCS-042 Erlingsson, ´U., Pihur, V . and Korolova, A. (201...

work page doi:10.29012/jpc.v7i3.405 2014

[5] [5]

and Roth, A

URL: http://dl.acm.org/citation.cfm?id=2660267.2660348 23 Ghosh, A. and Roth, A. (2015). Selling privacy at auction, Games and Economic Behavior 91: 334–346. URL: https://www.sciencedirect.com/science/article/pii/S0899825613000961 Goldfarb, A., Greenstein, S. M. and Tucker, C. E. (2015). Economic Analysis of the Digital Economy. URL: https://books.google....

work page doi:10.1007/978-3-642-28914-9 2015

[6] [6]

Samuelson, P . A. (1954). The pure theory of public expenditu re, Review of Eco- nomics and Statistics 37: 387–389. Spence, A. M. (1975). Monopoly , quality , and regulation, The Bell Journal of Eco- nomics 6(2): 417–429. URL: http://www.jstor.org/stable/3003237 Wasserman, L. and Zhou, S. (2010). A Statistical Framework f or Differential Pri- vacy ,Journa...

work page arXiv 1954

[7] [7]

bounded” differential privacy . However, this is not the standard deﬁ- nition of ε-differential privacy , which they take from Dwork et al. (20 06), because a “worst-case

So, D and D′ are adjacent databases and x and y are the adjacent histogram representations of D and D′, respectively . Some caution is required when reviewing the related literature because deﬁnitions m ay be stated in terms of adjacent databases or adjacent histograms. A.2 T ranslation of the Ghosh-Roth Model to Our Notation In this appendix we show that...

work page 2015