Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 3 (Global Price Monotonicity) — under any condition of Lemma 3, every optimal pricing path converges monotonically to the unique steady state p**

Proved
ReferencePricing.PriceMonotone.theorem3_global_price_monotonicity

by mikedeng1 · Oct 3, 2026 · Mathlib 0df444a (Lean v4.33.1)

dynamic-pricingmonotone-convergencep2o-batch-rm-cafterp2o-gran-per-chapterp2o-plan-paperp2o-v1reference-pricesteady-state

Theorem 3 (Global Price Monotonicity), Popescu and Wu 2007, p. 420. Consider the dynamic pricing problem with reference effects. Prices lie in P=[0,pˉ]P = [0,\bar p]P=[0,pˉ​] with pˉ>0\bar p > 0pˉ​>0. The reference price evolves as rt+1=αrt+(1−α)ptr_{t+1} = \alpha r_t + (1-\alpha)p_trt+1​=αrt​+(1−α)pt​ with memory α∈[0,1)\alpha \in [0,1)α∈[0,1), and profits are discounted by β∈(0,1)\beta \in (0,1)β∈(0,1). Demand is D(p,r)=d(p)+R(r−p,r)D(p,r) = d(p) + R(r-p,r)D(p,r)=d(p)+R(r−p,r) and the short-term profit is Π(p,r)=p D(p,r)\Pi(p,r) = p\,D(p,r)Π(p,r)=pD(p,r). Let VVV be the value function, p∗(r)p^*(r)p∗(r) the largest maximizer over PPP of W(p,r)=Π(p,r)+βV(αr+(1−α)p)W(p,r) = \Pi(p,r) + \beta V(\alpha r + (1-\alpha)p)W(p,r)=Π(p,r)+βV(αr+(1−α)p), and q∗(r)=αr+(1−α)p∗(r)q^*(r) = \alpha r + (1-\alpha)p^*(r)q∗(r)=αr+(1−α)p∗(r). For r0∈Pr_0 \in Pr0​∈P the state path is r0∗=r0r^*_0 = r_0r0∗​=r0​, rt+1∗=q∗(rt∗)r^*_{t+1} = q^*(r^*_t)rt+1∗​=q∗(rt∗​), and the optimal pricing path is pt∗=p∗(rt∗)p^*_t = p^*(r^*_t)pt∗​=p∗(rt∗​).

Assume Assumptions 1–3 and that ddd is differentiable. Assume also that at least one of the alternative conditions of Lemma 3 holds: (a) α=0\alpha = 0α=0; (b) D(p,r)D(p, r)D(p,r) is convex in rrr on PPP for every p∈Pp \in Pp∈P; (c) Π\PiΠ is strongly concave on (0,pˉ)2(0,\bar p)^2(0,pˉ​)2 and, for one constant K>0K > 0K>0, Πrr≥−K\Pi_{rr} \ge -KΠrr​≥−K and Πpr≥K/2\Pi_{pr} \ge K/2Πpr​≥K/2 there.

Then there is a price p∗∗p^{**}p∗∗ with the following properties. It is the unique steady state of the problem. It is the unique solution in PPP of

π′(p)1−β=p λ(p)1−αβ,(3)\frac{\pi'(p)}{1-\beta} = \frac{p\,\lambda(p)}{1-\alpha\beta}, \tag{3}1−βπ′(p)​=1−αβpλ(p)​,(3)

where π(p)=p d(p)\pi(p) = p\,d(p)π(p)=pd(p) and λ(r)=Rx(0,r)\lambda(r) = R_x(0,r)λ(r)=Rx​(0,r). And for every initial reference price r0∈Pr_0 \in Pr0​∈P, the optimal pricing path (pt∗)t≥0(p^*_t)_{t \ge 0}(pt∗​)t≥0​ is monotone (nondecreasing or nonincreasing) and converges to p∗∗p^{**}p∗∗.

Consequently only three pricing strategies can be optimal: skimming (prices decreasing), penetration (prices increasing), and constant pricing.

Formalization Note. The hypothesis is the disjunction LemmaThreeCondition, as the paper's "alternative conditions" require. "Monotonically" means the whole pricing path is monotone from t=0t = 0t=0, not only eventually. "The unique steady-state price p∗∗p^{**}p∗∗, characterized by Theorem 1" is rendered as existence and uniqueness of the steady state, together with p∗∗p^{**}p∗∗ being the unique solution of (3) in PPP. Differentiable ℝ d is added because the paper writes π′\pi'π′ without assuming it exists.

Condition (c) is kept exactly as printed, although on P=[0,pˉ]P = [0, \bar p]P=[0,pˉ​] it can rarely hold. Strong concavity forces Πrr≤−2b<0\Pi_{rr} \le -2b < 0Πrr​≤−2b<0 on the whole open square. But Πrr=p Drr\Pi_{rr} = p\,D_{rr}Πrr​=pDrr​, which tends to 000 as p→0p \to 0p→0 whenever DrrD_{rr}Drr​ stays bounded. The paper's own use of (c) (linear relative-difference models) is on a price range with a positive lower bound. A near-vacuous disjunct only weakens the hypothesis, so it does not trivialize the theorem, whose content comes from (a) and (b). The hypotheses are jointly satisfiable: pˉ=3/4\bar p = 3/4pˉ​=3/4, d(p)=1−pd(p) = 1-pd(p)=1−p, R(x,r)=x/4R(x, r) = x/4R(x,r)=x/4 satisfies Assumptions 1–3 and condition (b) for every α∈[0,1)\alpha \in [0,1)α∈[0,1) and β∈(0,1)\beta \in (0,1)β∈(0,1).

Preamble
import Mathlib
import Definitions.Def_RobustBooking_Shared_Assumptions
import Definitions.Def_RobustBooking_Shared_optPrice
import Definitions.Def_ReferencePricing_PriceMonotone_LemmaThreeCondition
open RobustBooking.Shared

open Filter Topology
Formal statement
namespace ReferencePricing.PriceMonotone

theorem theorem3_global_price_monotonicity
    (pbar α β : ℝ) (d : ℝ → ℝ) (R : ℝ → ℝ → ℝ)
    (hpbar : 0 < pbar) (hα₀ : 0 ≤ α) (hα₁ : α < 1) (hβ₀ : 0 < β) (hβ₁ : β < 1)
    (hA1 : Assumption1 pbar d R) (hA2 : Assumption2 R) (hA3 : Assumption3 pbar d R)
    (hd : Differentiable ℝ d)
    (hL3 : LemmaThreeCondition pbar α d R) :
    ∃ p : ℝ, IsSteadyState (profit d R) α β pbar p ∧
      (∀ q : ℝ, IsSteadyState (profit d R) α β pbar q → q = p) ∧
      deriv (baseProfit d) p / (1 - β) = p * refSlope R p / (1 - α * β) ∧
      (∀ q ∈ Set.Icc (0 : ℝ) pbar,
        deriv (baseProfit d) q / (1 - β) = q * refSlope R q / (1 - α * β) → q = p) ∧
      ∀ r₀ ∈ Set.Icc (0 : ℝ) pbar,
        (Monotone (pricePath (profit d R) α β pbar r₀) ∨
          Antitone (pricePath (profit d R) α β pbar r₀)) ∧
        Tendsto (pricePath (profit d R) α β pbar r₀) atTop (𝓝 p) := by sorry

end ReferencePricing.PriceMonotone
Source
Popescu and Wu, Dynamic Pricing Strategies with Reference Effects, Operations Research 55(3), 2007, p. 420, Theorem 3 (proof: Lemma 3 with Theorem 2; Appendix A.6–A.7, pp. 426–427)
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

Read-back for theorem3_global_price_monotonicity

This theorem uses eight definitions from the imported bundle files: Assumption1, Assumption2, Assumption3, LemmaThreeCondition, profit, baseProfit, refSlope, IsSteadyState and pricePath. Their bodies are not shown here, so this read-back does not unfold them. Each is treated as an opaque named predicate or function, and every claim below holds only in terms of what those definitions actually say.

Data. The statement takes:

  • real numbers pˉ\bar ppˉ​, α\alphaα and β\betaβ;
  • a real function d:R→Rd : \mathbb{R} \to \mathbb{R}d:R→R;
  • a real function of two real variables R:R×R→RR : \mathbb{R} \times \mathbb{R} \to \mathbb{R}R:R×R→R, written R(x,y)R(x, y)R(x,y).

Hypotheses. All of the following are assumed:

  • pˉ>0\bar p > 0pˉ​>0;
  • 0≤α<10 \le \alpha < 10≤α<1;
  • 0<β<10 < \beta < 10<β<1;
  • pˉ\bar ppˉ​, ddd and RRR satisfy the predicate Assumption1;
  • RRR satisfies the predicate Assumption2;
  • pˉ\bar ppˉ​, ddd and RRR satisfy the predicate Assumption3;
  • ddd is differentiable at every real number;
  • pˉ\bar ppˉ​, α\alphaα, ddd and RRR satisfy the predicate LemmaThreeCondition.

Notation. Write:

  • Π=profit(d,R)\Pi = \mathrm{profit}(d, R)Π=profit(d,R), a function built from ddd and RRR;
  • Π0=baseProfit(d)\Pi_0 = \mathrm{baseProfit}(d)Π0​=baseProfit(d), a real function built from ddd;
  • σ(q)=refSlope(R,q)\sigma(q) = \mathrm{refSlope}(R, q)σ(q)=refSlope(R,q), a real number built from RRR and qqq;
  • SS(q)\mathrm{SS}(q)SS(q) for the predicate "qqq is a steady state for Π\PiΠ with parameters α,β,pˉ\alpha, \beta, \bar pα,β,pˉ​" (IsSteadyState);
  • (ptr0)t(p_t^{r_0})_t(ptr0​​)t​ for the path pricePath(Π,α,β,pˉ,r0)\mathrm{pricePath}(\Pi, \alpha, \beta, \bar p, r_0)pricePath(Π,α,β,pˉ​,r0​) started from r0r_0r0​. It is a real-valued function of an ordered time index. The index type is fixed by the definition and not shown here.

Conclusion. There exists a real number ppp with all five of the following properties.

  1. SS(p)\mathrm{SS}(p)SS(p) holds.
  2. Every real qqq with SS(q)\mathrm{SS}(q)SS(q) equals ppp. So ppp is the unique steady state among all real numbers, not just those in [0,pˉ][0, \bar p][0,pˉ​].
  3. ppp satisfies the equation
Π0′(p)1−β  =  p σ(p)1−αβ.\frac{\Pi_0'(p)}{1-\beta} \;=\; \frac{p\,\sigma(p)}{1-\alpha\beta}.1−βΠ0′​(p)​=1−αβpσ(p)​.

Here Π0′\Pi_0'Π0′​ is the derivative of Π0\Pi_0Π0​. If Π0\Pi_0Π0​ is not differentiable at a point, its derivative there is taken to be 000. 4. Every q∈[0,pˉ]q \in [0, \bar p]q∈[0,pˉ​] that satisfies the same equation, Π0′(q)1−β=q σ(q)1−αβ\frac{\Pi_0'(q)}{1-\beta} = \frac{q\,\sigma(q)}{1-\alpha\beta}1−βΠ0′​(q)​=1−αβqσ(q)​, equals ppp. The statement does not itself say that p∈[0,pˉ]p \in [0, \bar p]p∈[0,pˉ​]. It only says that no point of [0,pˉ][0, \bar p][0,pˉ​] other than ppp solves the equation. 5. For every starting value r0∈[0,pˉ]r_0 \in [0, \bar p]r0​∈[0,pˉ​], both of the following hold:

  • the path t↦ptr0t \mapsto p_t^{r_0}t↦ptr0​​ is monotone non-decreasing or monotone non-increasing (weak monotonicity, so a constant path counts);
  • ptr0→pp_t^{r_0} \to pptr0​​→p as t→∞t \to \inftyt→∞.

Degenerate and edge cases.

  • Division by zero cannot occur. The hypotheses give 1−β∈(0,1)1-\beta \in (0,1)1−β∈(0,1) and αβ∈[0,1)\alpha\beta \in [0,1)αβ∈[0,1), so 1−αβ∈(0,1]1-\alpha\beta \in (0,1]1−αβ∈(0,1] and both denominators are strictly positive.
  • α=0\alpha = 0α=0 is allowed. The right-hand denominator is then 111 and the equation reads Π0′(p)/(1−β)=p σ(p)\Pi_0'(p)/(1-\beta) = p\,\sigma(p)Π0′​(p)/(1−β)=pσ(p).
  • The interval is never trivial. Because pˉ>0\bar p > 0pˉ​>0, [0,pˉ][0, \bar p][0,pˉ​] has more than one point, so the claims about all r0r_0r0​ and qqq in it are never vacuous for lack of points.
  • Derivatives default to zero. The hypotheses make ddd differentiable, but not necessarily Π0\Pi_0Π0​. At any point where Π0\Pi_0Π0​ fails to be differentiable, the equation is read with Π0′=0\Pi_0' = 0Π0′​=0 there. The equation then becomes 0=q σ(q)/(1−αβ)0 = q\,\sigma(q)/(1-\alpha\beta)0=qσ(q)/(1−αβ), which q=0q = 0q=0 satisfies automatically.
  • Unconstrained real inputs. Items 2 and 4 quantify over real numbers, so the definitions of Π\PiΠ, Π0\Pi_0Π0​, σ\sigmaσ, SS\mathrm{SS}SS and the path are applied to arbitrary reals. That includes negative values and values above pˉ\bar ppˉ​, where their values are whatever the definitions assign.
  • Possible vacuity. Whether Assumption1–Assumption3 and LemmaThreeCondition can all hold together cannot be determined from this code. If they cannot, the theorem is vacuously true.
Human review
  • Endorsed by Shuze Chen · Oct 21, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 21, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me