pith. sign in

arxiv: 2604.06643 · v1 · submitted 2026-04-08 · 💰 econ.EM

Testing for Monotone Equilibrium Strategies in Games of Incomplete Information

Yu-Chin Hsu , Tong Li , Chu-An Liu , Hidenori Takahashi This is my paper

Pith reviewed 2026-05-10 17:57 UTC · model grok-4.3

classification 💰 econ.EM
keywords monotone strategiesBayesian Nash equilibriumincomplete information gamesmoment inequalitiesquasi-inverseprocurement auctionscartel detectioneconometric testing
0
0 comments X

The pith

Monotonicity of differentiable equilibrium strategies is equivalent to monotonicity of a quasi-inverse strategy identified from observed actions.

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

This paper develops a framework to test whether players employ monotone strategies in Bayesian Nash equilibrium when their private information is unobserved. It establishes that, assuming symmetric independent private types and differentiable strategies, the monotonicity condition is equivalent to monotonicity of a quasi-inverse function that can be recovered directly from the distribution of observed actions. This equivalence allows the test to be cast as a set of moment inequalities that can be checked using standard econometric methods. The approach is illustrated with simulations and an application to detecting cartels in procurement auctions.

Core claim

Under symmetric independent private types, monotonicity of differentiable equilibrium strategies is equivalent to monotonicity of a quasi-inverse strategy identified from observed actions. This equivalence permits reformulation of the testing problem as checking a countable collection of moment inequalities based on unconditional expectations, for which the authors develop a Cramer-von Mises type statistic with bootstrap critical values that accommodates covariates and game heterogeneity.

What carries the argument

Quasi-inverse strategy identified from observed actions, which carries the monotonicity property equivalently to the true equilibrium strategy and enables the moment inequality representation.

If this is right

  • The testing problem is reduced to verifying a countable set of moment inequalities.
  • A Cramer-von Mises-type statistic with bootstrap critical values provides a practical test.
  • The framework handles covariates and heterogeneous games.
  • It can be used to detect deviations such as cartels in auction data.

Where Pith is reading between the lines

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

  • If the test rejects monotonicity, it may indicate collusion or other non-standard behavior in the game.
  • The method could potentially be adapted to test related properties in other incomplete-information models.
  • Researchers might combine this with other identification strategies to relax the symmetry assumption in future applications.

Load-bearing premise

Players have symmetric independent private types and equilibrium strategies are differentiable.

What would settle it

Generate data from a game with a non-monotone equilibrium strategy under symmetric independent private types and check whether the quasi-inverse is also non-monotone and the test detects the violation.

Figures

Figures reproduced from arXiv: 2604.06643 by Chu-An Liu, Hidenori Takahashi, Tong Li, Yu-Chin Hsu.

Figure 1
Figure 1. Figure 1: Quasi-inverse equilibrium strategy, ξ(b) the test performance much [PITH_FULL_IMAGE:figures/full_fig_p033_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Quasi-inverse equilibrium strategy for N = 2, 3, or 4 [PITH_FULL_IMAGE:figures/full_fig_p066_2.png] view at source ↗
read the original abstract

This paper develops a unified framework for testing monotonicity of Bayesian Nash equilibrium strategies in unobserved types in games of incomplete information. We show that, under symmetric independent private types, monotonicity of differentiable equilibrium strategies is equivalent to monotonicity of a quasi-inverse strategy identified from observed actions. This allows the problem to be reformulated as testing a countable set of moment inequalities involving unconditional expectations. We propose a Cramer-von Mises-type statistic with bootstrap critical values. The method accommodates covariates and game heterogeneity. Monte Carlo simulations demonstrate finite-sample performance, and an application to procurement auctions illustrates cartel detection.

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 / 3 minor

Summary. The paper develops a unified framework for testing monotonicity of Bayesian Nash equilibrium strategies in games of incomplete information with unobserved types. Under symmetric independent private types and differentiability of equilibrium strategies, it establishes equivalence between monotonicity of the strategy and monotonicity of a quasi-inverse strategy identified from the observed action distribution. This permits reformulation of the hypothesis as testing a countable set of unconditional moment inequalities, for which the authors propose a Cramer-von Mises statistic with bootstrap critical values. The procedure accommodates covariates and game heterogeneity. Finite-sample performance is examined via Monte Carlo simulations, and the method is applied to procurement auctions to detect potential cartels.

Significance. If the central equivalence holds, the paper makes a useful contribution to empirical game theory and auction econometrics by supplying a practical, distribution-free test for a core equilibrium property that does not require direct observation of private types. The reduction to moment inequalities and the bootstrap implementation are standard yet well-adapted to the setting, and the Monte Carlo evidence together with the cartel-detection application provide concrete illustrations of applicability. The framework could serve as a diagnostic tool in structural estimation of auctions and other incomplete-information games.

major comments (2)
  1. [Main equivalence result (likely §2 or §3)] The equivalence result between monotonicity of differentiable equilibrium strategies and monotonicity of the quasi-inverse (under SIPV) is load-bearing for the entire testing procedure. The manuscript should present the full derivation, including the precise role of differentiability and symmetry, in a dedicated theorem with proof to permit verification that no additional restrictions are implicitly imposed.
  2. [Bootstrap critical values section] The bootstrap procedure for the CvM statistic must be shown to control size when the null involves a countable collection of moment inequalities; the current description leaves open whether recentering or other adjustments are applied to handle the continuum of inequalities induced by the quasi-inverse.
minor comments (3)
  1. Notation for the quasi-inverse strategy and the identified action distribution should be introduced once and used consistently; occasional shifts between symbols for the same object reduce readability.
  2. [Monte Carlo section] The Monte Carlo design could usefully report power against specific non-monotone alternatives (e.g., non-monotonic jumps) rather than only size under the null, to better illustrate the test's ability to detect violations.
  3. [Empirical application] In the auction application, clarify how game heterogeneity (different numbers of bidders or reserve prices) is incorporated into the moment inequalities without altering the testable implications.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and insightful comments, which will help strengthen the clarity and rigor of our manuscript. We address each major comment point by point below, indicating the revisions we will make.

read point-by-point responses

  1. Referee: The equivalence result between monotonicity of differentiable equilibrium strategies and monotonicity of the quasi-inverse (under SIPV) is load-bearing for the entire testing procedure. The manuscript should present the full derivation, including the precise role of differentiability and symmetry, in a dedicated theorem with proof to permit verification that no additional restrictions are implicitly imposed.

    Authors: We agree that presenting the central equivalence as a dedicated theorem with a complete proof will enhance transparency and allow readers to verify the assumptions. In the revised manuscript, we will add Theorem 2.1 in Section 2, which formally states the equivalence: under symmetric independent private types (SIPV) and differentiability of the equilibrium strategy, monotonicity of the strategy is equivalent to monotonicity of the quasi-inverse identified from the observed action distribution. The full proof will be provided in the appendix, explicitly detailing the role of symmetry in ensuring the quasi-inverse is identified from the action distribution and the role of differentiability in establishing the monotonicity equivalence, without imposing any additional restrictions beyond those stated in the paper. revision: yes

  2. Referee: The bootstrap procedure for the CvM statistic must be shown to control size when the null involves a countable collection of moment inequalities; the current description leaves open whether recentering or other adjustments are applied to handle the continuum of inequalities induced by the quasi-inverse.

    Authors: We thank the referee for highlighting the need to clarify size control of the bootstrap. The current bootstrap for the CvM statistic applies recentering to handle the countable set of moment inequalities under the null, following standard approaches for such tests. To address concerns about the continuum induced by the quasi-inverse, we will expand the bootstrap section to explicitly describe the recentering procedure, how the inequalities are discretized into a countable collection in practice, and the asymptotic justification for size control. We will also add a reference to relevant results on bootstrap validity for moment inequality tests. This revision will make the implementation fully transparent. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the derivation chain

full rationale

The central derivation establishes a mathematical equivalence between monotonicity of differentiable equilibrium strategies and monotonicity of the quasi-inverse strategy recovered from the observed action distribution, under the maintained SIPV symmetry, independence, and differentiability assumptions. This equivalence is invoked to rewrite the hypothesis as a countable collection of unconditional moment inequalities; the moment inequalities are defined directly from the identified quasi-inverse without any fitted parameter being relabeled as a prediction. No self-citation chain, ansatz smuggling, or uniqueness theorem imported from the authors' prior work is required to close the argument. The CvM statistic and bootstrap are constructed from these inequalities in the standard way, and the Monte Carlo and auction application serve as external checks rather than internal validation loops. The derivation is therefore self-contained against the stated assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard domain assumptions from game theory; no free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption Symmetric independent private types
    Required for the equivalence between monotonicity of equilibrium strategies and monotonicity of the quasi-inverse strategy.
  • domain assumption Differentiability of equilibrium strategies
    Needed to establish the equivalence result stated in the abstract.

pith-pipeline@v0.9.0 · 5391 in / 1195 out tokens · 54353 ms · 2026-05-10T17:57:21.236500+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

  1. [1]

    Estimateν(b 1, b2, q) bybν(b1, b2, q) = cM(b 2, q)cW(b 1, q)− cM(b 1, q)cW(b 2, q) according to (A.1)

    work page

  2. [2]

    Computebσ2 ν,ϵ(b1, b2, q) according to (A.2)

    work page

  3. [3]

    Compute the test statistic bTS,au according to (A.3)

    work page

  4. [4]

    Compute the GMS functionψ S(b1, b2, q) according to (A.5)

    work page

  5. [5]

    , Kbootstrap sample, computebν ∗,k(b1, b2, q) and define Φ∗,k t (·) = √ S(bν∗,k(b1, b2, q)− bν(b1, b2, q))

    Fork= 1, . . . , Kbootstrap sample, computebν ∗,k(b1, b2, q) and define Φ∗,k t (·) = √ S(bν∗,k(b1, b2, q)− bν(b1, b2, q)). 3

    work page

  6. [6]

    , K o plusη

    For a significance levelα <1/2, compute the bootstrapped critical value ˆc η,au as the (1−α+η)-th quantile of n X (b1,b2,q)∈L max nΦ∗,k(b1, b2, q) bσν,ϵ(b1, b2, q) +ψ S(b1, b2, q),0 o2 Q(b1, b2, q) :k= 1, . . . , K o plusη

    work page

  7. [7]

    Proof of Theorem A.1:Given that n Bi,ℓ1 b≤B i,ℓ ≤b+ a q + 1 N−1 1 Bi,ℓ ≤b+ a q b+ a q −B i,ℓ −1(B i,ℓ ≤b)(b−B i,ℓ) :b∈[b ,¯b], q= 2,

    Reject the null if bTS,au >ˆcη,au. Proof of Theorem A.1:Given that n Bi,ℓ1 b≤B i,ℓ ≤b+ a q + 1 N−1 1 Bi,ℓ ≤b+ a q b+ a q −B i,ℓ −1(B i,ℓ ≤b)(b−B i,ℓ) :b∈[b ,¯b], q= 2, . . . o , n 1 b≤B i,ℓ ≤b+ a q :b∈[b ,¯b], q= 2, . . . o , are both Vapnik-Chervonenkis (VC) classes of functions, we have √ S(cM(·,·)−M(·,·)) and √ S(cW(·,·)−W(·,·)) weakly converge to Gaus...

    work page 2019

  8. [8]

    Estimateθby bθwhich is given as bθ= 1 S X i,ℓ X ′ ℓXℓ −1 1 S X i,ℓ X ′ ℓ log(Bi,ℓ) , and get bBu i,ℓ = exp(−Xℓbθ)Bi,ℓ

    work page

  9. [9]

    Estimateν(b 1, b2, q, θ) bybν(b1, b2, q,bθ) = cM(b 2, q,bθ)cW(b 1, q,bθ)− cM(b 1, q,bθ)cW(b 2, q,bθ) in which cM(b, q,bθ) = 1 S X i,ℓ bBu i,ℓ1 b≤ bBu i,ℓ ≤b+ a q + 1 N−1 1 bBu i,ℓ ≤b+ a q b+ a q − bBu i,ℓ −1( bBu i,ℓ ≤b)(b− bBu i,ℓ) , cW(b, q,bθ) = 1 S X i,ℓ 1 b≤ bBu i,ℓ ≤b+ a q

    work page

  10. [10]

    , Kbootstrap sample, compute bθ∗,k, the bootstrap estimator forθ, and obtainbu∗,k i,ℓ = exp(−X ∗,k i,ℓ bθ∗,k)B∗,k i,ℓ andbν∗,k(b1, b2, q,bθ∗,k)

    Fork= 1, . . . , Kbootstrap sample, compute bθ∗,k, the bootstrap estimator forθ, and obtainbu∗,k i,ℓ = exp(−X ∗,k i,ℓ bθ∗,k)B∗,k i,ℓ andbν∗,k(b1, b2, q,bθ∗,k)

    work page

  11. [11]

    Compute the bootstrap standard error estimate forσ ν(b1, b2, q,bθ) as bσ∗ ν(b1, b2, q,bθ) = 1 K KX k=1 S(bν∗,k(b1, b2, q,bθ∗,k)−¯ν∗(b1, b2, q,bθ))2 1/2 ,where ¯ν∗(b1, b2, q,bθ) = 1 K KX k=1 bν∗,k(b1, b2, q,bθ∗,k), and definebσ∗ ν,ϵ(b1, b2, q,bθ) = max{bσ∗ ν(b1, b2, q,bθ), √ϵ·bσ∗ ν(b,(b + b)/2,2,bθ)}

    work page

  12. [12]

    Compute the test statistic as bTS,se = X (b1,b2,q)∈L max n√ S bν(b1, b2, q,bθ) bσ∗ ν,ϵ(b1, b2, q,bθ) ,0 o2 Q(b1, b2, q). 9

    work page

  13. [13]

    Compute the GMS functionψ S(t, b1, b2, q) as ψS(t, b1, b2, q) =−β S ·1 √ Sbν(b1, b2, q,bθ) bσ∗ ν,ϵ(b1, b2, q,bθ) <−κ S

    work page

  14. [14]

    Let Φ ∗,k(b1, b2, q) = √ S(bν∗,k(b1, b2, q,bθ∗,k)−bν(b1, b2, q,bθ))

    work page

  15. [15]

    , K o and define the bootstrapped critical value ˆcη,se = ˆc+η

    For a significance levelα <1/2, compute ˆcas the (1−α+η)-th quantile of n X (b1,b2,q)∈L max n Φ∗,k(b1, b2, q) bσ∗ ν,ϵ(b1, b2, q,bθ) +ψ S(t, b1, b2, q),0 o2 Q(b1, b2, q) :k= 1, . . . , K o and define the bootstrapped critical value ˆcη,se = ˆc+η

    work page

  16. [16]

    Implementation Procedure for Semipara- metric Test under Observed Auction Heterogeneity

    Reject the null if bTS,se >ˆcη,se. We summarize the asymptotic properties of the semiparametric test in the Appendix. Note that the bootstrap estimator for the asymptotic standard error in Step 5 can be applied to other cases as well. However, it is more time-consuming to compute bootstrap standard error estimator than to compute the standard error estima...

    work page 1996