Proposition 1 — uninformed customers: the unique pure PSE is for , for (corrected: ; exception set is a union)
ProvedQueueSignal.proposition_1Consider a single-server queue whose server is of high quality (reward ) with prior probability and of low quality (reward ) otherwise, with exponential service at rate , Poisson arrivals at potential rate and waiting cost per unit time, where , and . The server knows its type and chooses to reveal () or conceal () its queue; all customers are uninformed of the quality and update their belief from the server's action. Let and be the effective arrival rates (1) and (3) under belief . Then:
- the functions and on have a unique crossing point ;
- for every with
and every pure signaling profile :
- if , then admits beliefs that make it a pure perfect sequential equilibrium if and only if (pooling on concealment);
- if , then admits beliefs that make it a pure perfect sequential equilibrium if and only if (pooling on revelation).
Hence, with only uninformed customers, the queue-disclosure action carries no information about quality: away from finitely many exceptional arrival rates, the pure perfect sequential equilibrium pools, on concealment for small markets and on revelation for large ones.
Formalization Note "The unique pure-strategy perfect sequential equilibrium is " is the equivalence, over the four pure profiles, between "some beliefs make a pure perfect sequential equilibrium" and "". Perfect sequential equilibrium is QueueSignal.IsPurePSE (Grossman–Perry: no credible off-path belief makes a deviation strictly profitable).
Differences from the source:
- correction: "otherwise" (λ ≥ λ̂), it is pooling on R → λ > λ̂; reason: at λ = λ̂ both (C,C) and (R,R) (each with off-path belief δ) are pure perfect sequential equilibria, for every admissible parameter tuple; with V_H = 5/2, V_L = 2, µ = 1, θ = 1/2, δ = 1/2 (Example 1, p.554) λ̂ ≈ 0.947 lies outside both exception sets, so the printed statement fails there
- correction: "except at those potential arrival rates under which λ^O(1) = λ^U(0) and λ^O(0) = λ^U(1)" (read as a conjunction) → λ is excluded when either equality holds (the union of the two sets); reason: with the parameters above at λ = 1: λ^O(0) = 4/5 = λ^U(1) but λ^O(1) = 5/6 ≠ 3/4 = λ^U(0), and the separating (C,R) with δ^R = 0, δ^C = 1 is a pure perfect sequential equilibrium, though it is neither pooling profile; if "and" lists the two sets, this is the reading of the page
- reading: perfect sequential equilibrium via the credible updating rule (Definition 2), with possibly many credible beliefs (footnote 1) → a pure sequential equilibrium such that no credible off-path belief makes some type strictly better off by deviating (Grossman–Perry); reason: the test the paragraph after Definition 2 applies; under the other reading (the equilibrium's own off-path belief must be credible) the proposition is false: with the parameters above at λ = 9/10, (R,R) with δ^C ≈ 0.0951 (both types indifferent) would also qualify
- encoding: "the unique crossing point of λ^O(δ) and λ^U(δ)" named λ̂ on p.549 → clause 1 asserts that a positive crossing point exists and is unique, and binds λ̂ to it; reason: the object the page names; it is not a free parameter
- encoding: server and customers play the full signaling game; sequential rationality of the customers (Naor's threshold n(δ^R) in the revealed queue, the joining probability p(δ^C) of (2) in the concealed queue) → the server's payoff after R under belief d is λ^O(d) of (1), after C it is λ^U(d) of (3), the same for both types; reason: the paper's own reduction on p.548 (sequential rationality is equivalent to the customers playing (n(δ^R), p(δ^C)) and the server maximizing its effective arrival rate; π_{i,H} = π_{i,L})
- encoding: beliefs δ^R, δ^C → real numbers in [0, 1], one per signal (a function Sig → ℝ); reason: Lean convention; same instances
- encoding: "the unique pure-strategy perfect sequential equilibrium is pooling on C (resp. R)" → for every pure profile σ: (∃ beliefs making σ a pure perfect sequential equilibrium) ↔ σ = (C,C) (resp. (R,R)); reason: equilibria with the same profile differ only in their off-path beliefs; the paper counts signaling outcomes
import Mathlib import Definitions.Def_QueueSignal_SignalingGame import Definitions.Def_QueueSignal_QueueModel
namespace QueueSignal
/-- **Proposition 1** (Guo, Haviv, Luo, Wang, *Signaling Service Quality Through Queue Disclosure*, MSOM 25(2) (2023),
doi:10.1287/msom.2022.1170, §4.2, p.549, PDF p.8), corrected at `λ = λ̂` and in the
exception set.
Consider the baseline game in which all customers are uninformed of service quality, with
`μ, θ > 0`, `V_H > V_L > θ/μ` and `0 < δ < 1`. Then
1. `λ^O(δ)` and `λ^U(δ)`, as functions of the potential arrival rate `λ > 0`, have a unique
crossing point `λ̂ > 0`;
2. for every `λ > 0` with `λ^O(1) ≠ λ^U(0)` and `λ^O(0) ≠ λ^U(1)`, and every pure profile `σ`:
* if `λ < λ̂`, then `σ` (with suitable beliefs) is a pure perfect sequential equilibrium iff
`σ = (C, C)`;
* if `λ > λ̂`, then `σ` (with suitable beliefs) is a pure perfect sequential equilibrium iff
`σ = (R, R)`.
**Formalization Note.** "The unique pure-strategy perfect sequential equilibrium is `X`" is the
iff over the four pure profiles, with the beliefs existentially quantified (equilibria differ
in their off-path beliefs). Perfect sequential equilibrium is `IsPurePSE` (no credible
off-path belief makes a deviation strictly profitable). The printed "otherwise" (`λ ≥ λ̂`) is
corrected to `λ > λ̂`, and the printed exception set "`λ^O(1) = λ^U(0)` and `λ^O(0) = λ^U(1)`" is
read as the union of the two equality sets. -/
theorem proposition_1 (P : Params) (hP : P.Valid) :
∃ lhat : ℝ, (0 < lhat ∧ lamO P lhat P.δ = lamU P lhat P.δ) ∧
(∀ lam : ℝ, 0 < lam → lamO P lam P.δ = lamU P lam P.δ → lam = lhat) ∧
∀ lam : ℝ, 0 < lam → lamO P lam 1 ≠ lamU P lam 0 → lamO P lam 0 ≠ lamU P lam 1 →
∀ σ : Profile,
(lam < lhat →
((∃ b : Sig → ℝ, IsPurePSE P.δ (pay P lam) σ b) ↔ σ = (.C, .C))) ∧
(lhat < lam →
((∃ b : Sig → ℝ, IsPurePSE P.δ (pay P lam) σ b) ↔ σ = (.R, .R))) := by sorry
end QueueSignal
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.