pith. sign in

arxiv: 2507.09415 · v2 · pith:5DTTKGH2new · submitted 2025-07-12 · 💰 econ.TH · math.OC

Contracting a crowd of heterogeneous agents

Guillermo Alonso Alvarez , Erhan Bayraktar , Ibrahim Ekren This is my paper

Pith reviewed 2026-05-21 23:43 UTC · model grok-4.3

classification 💰 econ.TH math.OC
keywords optimal contract designheterogeneous agentsnetwork spilloverscontinuum approximationlinear quadratic frameworkprincipal agent problemfinite population
0
0 comments X

The pith

In linear-quadratic settings with network spillovers, the continuum contract approximates the finite-agent optimum to order 1/N when evaluated on large samples.

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

This paper establishes a method for designing optimal contracts for large groups of heterogeneous agents whose actions influence each other through a network. By solving the problem first in a continuum limit where agents are treated as a continuous distribution, then applying that solution to a finite but large sample, one obtains contracts that perform nearly as well as the exact finite-agent optimum. The approximation error shrinks like 1/N, making the approach practical for populations too big for direct computation. A sympathetic reader would care because direct optimization over many interacting agents quickly becomes intractable, while this limit-based method scales without losing much value. The work also shows how to adjust incentives based on each agent's position in the network to maximize the principal's payoff.

Core claim

The paper solves the finite-agent principal-agent problem with heterogeneous agents and network interactions under linear-quadratic payoffs to obtain explicit optimal contracts and equilibrium efforts. It then derives the corresponding continuum limit and proves that sampling the continuum contract on N agents yields admissible contracts whose value for the principal differs from the true finite optimum by at most order 1/N. Stability of the solution under small changes to the interaction function is established, along with comparative statics illustrating the effect of network centrality on effort levels and incentive design.

What carries the argument

The continuum limit of the linear-quadratic principal-agent problem with network spillovers, which provides explicit solutions that transfer to finite samples with controlled error.

If this is right

  • The optimal contract targets higher incentives to agents generating larger spillovers.
  • Small perturbations in the interaction function lead to stable changes in the optimal contract.
  • Network position determines individual effort and the overall value extracted by the principal.
  • Explicit formulas allow computation of comparative statics without resolving the full optimization.

Where Pith is reading between the lines

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

  • The 1/N convergence suggests that similar limit arguments might work in other contract design problems with large populations.
  • This framework could inform policy for regulating incentives in interconnected markets or organizations.
  • One could test the approximation numerically by solving small-N cases exactly and checking the error decay.

Load-bearing premise

The analysis relies on linear-quadratic utilities and costs to derive closed-form solutions and the specific 1/N error rate.

What would settle it

Computing the exact optimal contract value for a specific finite N and interaction function, then comparing it to the value from the sampled continuum contract, would show whether the difference is indeed bounded by a constant times 1/N.

Figures

Figures reproduced from arXiv: 2507.09415 by Erhan Bayraktar, Guillermo Alonso Alvarez, Ibrahim Ekren.

Figure 1
Figure 1. Figure 1: Solution to the contracting problem with [PITH_FULL_IMAGE:figures/full_fig_p019_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Solution to the contracting problem with [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Solution to the contracting problem with [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Solution to the contracting problem with [PITH_FULL_IMAGE:figures/full_fig_p020_4.png] view at source ↗
read the original abstract

We study optimal contract design for large populations of heterogeneous agents whose actions generate network spillovers represented by an interaction function. In a linear-quadratic framework, we solve the finite-agent problem and its continuum limit, obtaining explicit optimal contracts and equilibrium efforts. We show that the continuum contract can be evaluated on a large finite sample of agents to obtain admissible contracts that achieve the finite-agent principal's value up to an error of order 1/N. This provides a scalable approximation for settings with many interacting agents. We also prove stability with respect to perturbations of the interaction function and provide comparative statics and numerical examples showing how network position affects effort, incentives, and the principal's value. The results identify how optimal incentives should be targeted toward agents whose actions generate larger spillovers.

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.

Referee Report

2 major / 2 minor

Summary. The paper studies optimal contract design for large populations of heterogeneous agents whose actions generate network spillovers via an interaction function. In a linear-quadratic framework, it derives explicit optimal contracts and equilibrium efforts for both the finite-N problem and its continuum limit. It shows that evaluating the continuum contract on a realized finite sample yields admissible contracts whose value approximates the finite-agent principal's optimum at rate O(1/N). The paper further establishes stability of the solution with respect to perturbations of the interaction function, provides comparative statics on how network position affects effort and incentives, and includes numerical illustrations.

Significance. If the explicit derivations and approximation hold, the work supplies a scalable, closed-form approach to incentive design in large networked populations, which is valuable for mechanism design applications involving spillovers (e.g., crowdsourcing or social networks). The reduction to a linear system in the finite case, the O(1/N) guarantee via empirical-measure concentration on the quadratic objective, and the stability result are concrete strengths that distinguish the contribution from purely asymptotic mean-field analyses.

major comments (2)
  1. [§4] §4 (Continuum limit and approximation): the argument that the value difference is exactly quadratic in the measure deviation (hence inherits the 1/N rate from standard concentration) is load-bearing for the central approximation claim; the manuscript should state the precise moment conditions on the interaction kernel that justify interchanging the quadratic form with the empirical deviation without additional uniformity arguments.
  2. [§3.2] §3.2 (Finite-N linear system): the claim of unique solvability for the principal's problem for every N relies on the interaction matrix being invertible; while the LQ structure makes this plausible, an explicit bound on the smallest eigenvalue (or a uniform invertibility condition) would strengthen the result for arbitrary network realizations.
minor comments (2)
  1. [Numerical examples] The numerical examples would be clearer if the specific network topologies (e.g., degree distributions or adjacency matrices) and parameter values were reported in a table or appendix so that the comparative statics on network position can be reproduced.
  2. [Model setup] Notation for the interaction function and the type distribution could be introduced earlier and used consistently to improve readability across the finite and continuum sections.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive suggestions. The comments point to useful clarifications that will strengthen the technical presentation of the approximation and solvability results. We respond to each major comment below.

read point-by-point responses

  1. Referee: [§4] §4 (Continuum limit and approximation): the argument that the value difference is exactly quadratic in the measure deviation (hence inherits the 1/N rate from standard concentration) is load-bearing for the central approximation claim; the manuscript should state the precise moment conditions on the interaction kernel that justify interchanging the quadratic form with the empirical deviation without additional uniformity arguments.

    Authors: We agree that an explicit statement of the moment conditions improves rigor. The quadratic structure of the principal's objective means the value difference is precisely a quadratic form in the empirical-measure deviation; under the maintained assumption that the interaction kernel is bounded and continuous, this quadratic form is continuous with respect to weak convergence, so the O(1/N) rate follows directly from standard empirical-measure concentration without further uniformity arguments. In the revision we will add a short remark in §4 stating these conditions (boundedness and continuity of the kernel) and confirming that they suffice for the interchange. revision: yes

  2. Referee: [§3.2] §3.2 (Finite-N linear system): the claim of unique solvability for the principal's problem for every N relies on the interaction matrix being invertible; while the LQ structure makes this plausible, an explicit bound on the smallest eigenvalue (or a uniform invertibility condition) would strengthen the result for arbitrary network realizations.

    Authors: We appreciate the suggestion. The finite-N first-order condition yields a linear system whose matrix is I minus a scaled interaction matrix whose entries are given by the realized interaction function. Our maintained boundedness assumption on the interaction function ensures that the spectral radius is strictly less than one for any finite network, guaranteeing invertibility. To address the referee's point, we will include in the revision an explicit lower bound on the smallest eigenvalue expressed in terms of the supremum norm of the interaction function and the model parameters; this bound holds uniformly over all networks satisfying the paper's assumptions. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained from primitives

full rationale

The paper derives explicit optimal contracts and equilibrium efforts directly from the linear-quadratic primitives for both the finite-N and continuum problems. The central approximation result—that evaluating the continuum contract on a finite sample yields admissible contracts with value error of order 1/N—follows from standard empirical-measure concentration applied to the quadratic objective and interaction kernel, without reducing the target quantity to a fitted parameter or self-referential definition. No load-bearing step invokes a self-citation chain, uniqueness theorem from prior author work, or ansatz smuggled via citation; the 1/N bound is obtained from the explicit quadratic structure and concentration inequalities rather than by construction. The model assumptions (linear-quadratic utilities and costs) are stated upfront and required for closed forms, but this is an explicit modeling choice rather than circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on the choice of a linear-quadratic framework to obtain closed forms; no free parameters or invented entities are mentioned in the abstract.

axioms (1)
  • domain assumption Utilities and costs are linear-quadratic.
    Explicitly invoked in the abstract as the setting that permits closed-form solutions and the 1/N error bound.

pith-pipeline@v0.9.0 · 5652 in / 1297 out tokens · 56497 ms · 2026-05-21T23:43:41.780973+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages · 1 internal anchor

  1. [1]

    A Rank-Based Reward between a Principal and a Field of Agents: Application to Energy Savings

    Clemence Alasseur, Erhan Bayraktar, Roxana Dumitriescu, and Quentin Jacquet. A Rank-Based Reward between a Principal and a Field of Agents: Application to Energy Savings . SSRN 4253686 , 2022

    work page 2022

  2. [2]

    Stochastic graphon games: II

    Alexander Aurell, René Carmona, and Mathieu Laurière. Stochastic graphon games: II . T he linear-quadratic case. Appl Math Optim , 85(39), 2022

    work page 2022

  3. [3]

    Randomized dynamic programming principle and Feynman-Kac representation for optimal control of McKean-Vlasov dynamics

    Erhan Bayraktar, Andrea Cosso, and Huy ê n Pham. Randomized dynamic programming principle and Feynman-Kac representation for optimal control of McKean-Vlasov dynamics . Transactions of the American Mathematical Society , 370(3):2115--2160, 2018

    work page 2018

  4. [4]

    Graphon mean field systems

    Erhan Bayraktar, Suman Chakraborty, and Ruoyu Wu. Graphon mean field systems. The Annals of Applied Probability , 33(5):3587--3619, 2023

    work page 2023

  5. [5]

    Large Tournament Games

    Erhan Bayraktar, Jak s a Cvitani \'c , and Yuchong Zhang. Large Tournament Games . The Annals of Applied Probability , 29(6):3695--3744, 2019

    work page 2019

  6. [6]

    Graphon particle systems with common noise

    Erhan Bayraktar, Xihao He, and Donghan Kim. Graphon particle system with common noise. arXiv preprint arXiv:2507.03265 , 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  7. [7]

    Concentration of measure for graphon particle system

    Erhan Bayraktar and Donghan Kim. Concentration of measure for graphon particle system . Advances in Applied Probability , 56:1279--1306, 2024

    work page 2024

  8. [8]

    Stationarity and uniform in time convergence for the graphon particle system

    Erhan Bayraktar and Ruoyu Wu. Stationarity and uniform in time convergence for the graphon particle system . Stochastic Processes and Their Applications , 150:532--568, 2022

    work page 2022

  9. [9]

    Propagation of chaos of Forward–Backward Stochastic Differential Equations with Graphon interactions

    Erhan Bayraktar, Ruoyu Wu, and Xin Zhang. Propagation of chaos of Forward–Backward Stochastic Differential Equations with Graphon interactions . Appl Math Optim , 88(25), 2023

    work page 2023

  10. [10]

    Terminal Ranking Games

    Erhan Bayraktar and Yuchong Zhang. Terminal Ranking Games . Mathematics of Operations Research , 46(4):1349--1365, 2021

    work page 2021

  11. [11]

    Non-parametric estimates for graphon mean-field particle systems

    Erhan Bayraktar and Hongyi Zhou. Non-parametric estimates for graphon mean-field particle systems . Bernoulli , 31(4):2940--2961, 2025

    work page 2025

  12. [12]

    Nonlinear graphon mean-field systems

    Fabio Coppini, Anna De Crescenzo, and Huyen Pham. Nonlinear graphon mean-field systems. arXiv preprint arXiv:2402.08628 , 2024

    work page arXiv 2024

  13. [13]

    P robabilistic T heory of M ean F ield G ames with A pplications I - II

    René Carmona and François Delarue. P robabilistic T heory of M ean F ield G ames with A pplications I - II . Springer Nature, 2018

    work page 2018

  14. [14]

    Linear-quadratic optimal control for non-exchangeable mean-field SDEs and applications to systemic risk

    Anna De Crescenzo, Filippo de Feo, and Huy ê n Pham. Linear-quadratic optimal control for non-exchangeable mean-field SDEs and applications to systemic risk. arXiv preprint arXiv:2503.03318 , 2025

    work page arXiv 2025

  15. [15]

    Caines and Minyi Huang

    Peter E. Caines and Minyi Huang. Graphon mean field games and their equations. SIAM Journal on Control and Optimization , 59:4373 -- 4399, 2021

    work page 2021

  16. [16]

    A probabilistic weak formulation of mean field games and applications

    Ren é Carmona and Daniel Lacker. A probabilistic weak formulation of mean field games and applications. Annals of Applied Probability , 25(3):1189–1231, 2015

    work page 2015

  17. [17]

    Dynamic programming approach to principal--agent problems

    Jak s a Cvitani \'c , Dylan Possama \" , and Nizar Touzi. Dynamic programming approach to principal--agent problems. Finance and Stochastics , 22:1--37, 2018

    work page 2018

  18. [18]

    Stackelberg Mean Field Games: convergence and existence results to the problem of Principal with multiple Agents in competition

    Mao Fabrice Djete. Stackelberg Mean Field Games: convergence and existence results to the problem of Principal with multiple Agents in competition . arXiv preprint arXiv:2309.00640 , 2023

    work page arXiv 2023

  19. [19]

    A Machine Learning Method for Stackelberg Mean Field Games

    G ö kçe Dayanıklı and Mathieu Lauri è re. A Machine Learning Method for Stackelberg Mean Field Games . Mathematics of Operations Research , 2023

    work page 2023

  20. [20]

    Mean–field moral hazard for optimal energy demand response management

    Romuald Elie, Emma Hubert, Thibaut Mastrolia, and Dylan Possamaï. Mean–field moral hazard for optimal energy demand response management. Mathematical Finance , 31(1):399--473, 2020

    work page 2020

  21. [21]

    A tale of a principal and many, many agents

    Romuald \'E lie, Thibaut Mastrolia, and Dylan Possamaï. A tale of a principal and many, many agents. Mathematics of Operations Research , 44(2):377--766, 2019

    work page 2019

  22. [22]

    Contracting theory with multiple competitive agents

    Romuald \'E lie and Dylan Possama \" . Contracting theory with multiple competitive agents. SIAM Journal of Control and Optimization , 57(2):1157--1188, 2019

    work page 2019

  23. [23]

    Caines, and Roland P

    Minyi Huang, Peter E. Caines, and Roland P. Malham é . The N ash certainty equivalence principle and M ckeanvlasov systems: A n invariance principle and entry adaptation. 46th IEEE Conference on Decision and Control , page 121–126, 2007

    work page 2007

  24. [24]

    Aggregation and linearity in the provision of intertemporal incentives

    Bengt Holmstrom and Paul Milgrom. Aggregation and linearity in the provision of intertemporal incentives. Econometrica: Journal of the Econometric Society , pages 303--328, 1987

    work page 1987

  25. [25]

    Malhamé, and Peter E

    Minyi Huang, Roland P. Malhamé, and Peter E. Caines. Large population stochastic dynamic games: closed-loop McKean-Vlasov systems and the Nash certainty equivalence principle. Communications in Information and Systems , 6(3):221--252, 2006

    work page 2006

  26. [26]

    Pierre L. Lions. Théorie des jeux de champ moyen et applications. , Cours du Collège de France, 2006-20122

    work page 2006

  27. [27]

    Lasry and P.L

    J.-M. Lasry and P.L. Lions. Jeux à champ moyen. i. le cas stationnaire. C. R. M ath. A cad. S ci. P aris , 343(9):619--625, 2006

    work page 2006

  28. [28]

    Lasry and P.L

    J.-M. Lasry and P.L. Lions. Jeux à champ moyen. ii. horizon fini et contr ô le optimal. C. R. M ath. A cad. S ci. P aris , 343(10):679–684, 2006

    work page 2006

  29. [29]

    Lasry and P.-L

    J.-M. Lasry and P.-L. Lions. Mean field games. Japanese Journal of Mathematics , 2:229--260, 2007

    work page 2007

  30. [30]

    A label-state formulation of stochastic graphon games and approximate equilibria on large networks

    Daniel Lacker and Agathe Soret. A label-state formulation of stochastic graphon games and approximate equilibria on large networks. Mathematics of Operations Research , 48(4):1987--2018, 2022

    work page 1987

  31. [31]

    Agency problem and mean field system of agents with moral hazard

    Thibaut Mastrolia and Jiacheng Zhang. Agency problem and mean field system of agents with moral hazard. Journal of Optimization Theory and Applications , 205(3):1--32, 2025

    work page 2025

  32. [32]

    Graphon Games

    Francesca Parise and Asuman Ozdaglar. Graphon Games . Proceedings of the 2019 ACM Conference on Economics and Computation , pages 457--458, 2019

    work page 2019

  33. [33]

    Graphon games: A statistical framework for network games and interventions

    Francesca Parise and Asuman Ozdaglar. Graphon games: A statistical framework for network games and interventions . Econometrica , 91(1):191--225, 2023

    work page 2023

  34. [34]

    Bellman equation and viscosity solutions for mean-field stochastic control problem

    Huy ê n Pham and Xiaoli Wei. Bellman equation and viscosity solutions for mean-field stochastic control problem. ESAIM: Control, Optimization and Calculus of Variations , 24(1), 2018

    work page 2018

  35. [35]

    A continuous-time version of the principal-agent problem

    Yuliy Sannikov. A continuous-time version of the principal-agent problem. The Review of Economic Studies , 75(3), 2008

    work page 2008