EconCSLib.MechanismDesign.Auction.BayesianSingleItem #
Single-item auctions in incomplete-information settings, modeled as direct Bayesian mechanisms with transfers.
This file specializes the general interface from MechanismDesign/Auction/MechBayesian to the
standard direct-revelation single-item auction environment:
- each agent reports a scalar type in
ℝ - the type profile is drawn from a common prior measure
- each agent receives a winning probability
Q_i(t) ∈ [0,1] - payments are determined by the mechanism's payment rule
The continuous private-value data are recorded separately via
ContinuousTypeProfile, so Myerson-style regularity assumptions can be attached
without forcing them into the core mechanism fields.
Interim objects are defined against an explicit family of opponent-type priors
μᵢ on profiles t₋ᵢ. This keeps the generic measure-based definitions
separate from any density-based formulas one may later derive under
independence.
Structure hierarchy #
MechanismWithTransfers I (fun _ => ℝ) (I → ℝ) ℝ
└─ SingleParameterMechanism I ℝ -- allocation probabilities + payments
└─ BayesianSingleItemAuction I -- prior + continuous type data
├─ toDirectBayesianMechanismWithTransfers
└─ toSingleParameterMechanism
A scalar type distribution on the interval [0, ω], recorded by its CDF.
Upper bound of the support interval is nonnegative.
Cumulative distribution function.
- monotoneOn_cdf : MonotoneOn self.cdf (Set.Icc 0 ω)
Monotonicity of the CDF on the support interval.
Normalization at the lower endpoint.
Normalization at the upper endpoint.
- differentiableOn_cdf : DifferentiableOn ℝ self.cdf (Set.Ioo 0 ω)
Smoothness assumption used in continuous-type auction theory.
Instances For
Opponent type profiles for real-valued auction types.
Equations
- OpponentTypeProfile I i = OpponentProfile I ℝ i
Instances For
A direct Bayesian single-item auction with scalar private types and probabilistic allocation.
The allocation rule returns a function Q : I → ℝ, where Q i is agent i's
probability of receiving the item under the reported type profile. This matches
the allocation/payment shape of SingleParameterMechanism I ℝ.
The mechanism is direct: the message space equals the type space, which is taken
to be ℝ for each agent, and payments are real-valued. Continuous-type data
are stored separately in typeData, while the Bayesian prior is recorded by
extra fields.
- allocationRule : (I → ℝ) → I → ℝ
- paymentRule : (I → ℝ) → I → ℝ
- prior : MeasureTheory.Measure (I → ℝ)
Common prior probability measure over true type profiles.
- prob_prior : MeasureTheory.IsProbabilityMeasure self.prior
The prior is a probability measure.
- opponentPrior (i : I) : MeasureTheory.Measure (OpponentTypeProfile I i)
- prob_opponentPrior (i : I) : MeasureTheory.IsProbabilityMeasure (self.opponentPrior i)
Each opponent-type prior is a probability measure.
- typeData : ContinuousTypeProfile I
Continuous private-value support and CDF data for each agent.
Instances For
Insert one coordinate into an opponent profile.
Equations
- BayesianSingleItemAuction.profileInsert i x t j = if h : j = i then x else t ⟨j, h⟩
Instances For
The inserted coordinate is the inserted value.
Other coordinates come from the opponent profile.
View an incomplete-information single-item auction as the corresponding direct Bayesian mechanism with transfers.
Equations
- A.toDirectBayesianMechanismWithTransfers = { prior := A.prior, prob_prior := ⋯, allocationRule := A.allocationRule, paymentRule := A.paymentRule }
Instances For
The allocation rule respects the single-item probability budget.
Equations
- A.RespectsSingleItemCapacity = ∀ (t : I → ℝ), ∑ i : I, A.allocationRule t i ≤ 1
Instances For
Feasibility for a probabilistic single-item auction:
- each winning probability lies in
[0,1] - the total allocation probability is at most
1
Equations
Instances For
Seller revenue at a report profile.
Equations
- A.sellerRevenue t = ∑ i : I, A.paymentRule t i
Instances For
Ex-ante seller revenue.
Equations
- A.expectedSellerRevenue = ∫ (t : I → ℝ), A.sellerRevenue t ∂A.prior
Instances For
Agent i's one-dimensional density, defined as the derivative of the
stored CDF Fᵢ.
Instances For
Type measure generated by fᵢ on [0, ωᵢ].
Equations
- A.typeMeasure i = (MeasureTheory.volume.restrict (Set.Ioc 0 (A.typeData.omega i))).withDensity fun (v : ℝ) => ENNReal.ofReal (A.typeDensity i v)
Instances For
The joint density f(t) of the type profile under independence.
If the agents' types are independent with agent-specific one-dimensional
densities fᵢ, then the profile density is the product ∏ᵢ fᵢ(tᵢ).
Equations
- A.jointDensity t = ∏ i : I, A.typeDensity i (t i)
Instances For
Product type prior.
Equations
- A.productPrior = MeasureTheory.Measure.pi fun (i : I) => A.typeMeasure i
Instances For
Product prior over i's opponents.
Equations
- A.opponentProductPrior i = MeasureTheory.Measure.pi fun (j : { j : I // j ≠ i }) => A.typeMeasure ↑j
Instances For
Stored priors are the product priors.
Equations
- A.HasIndependentTypePriors = (A.prior = A.productPrior ∧ ∀ (i : I), A.opponentPrior i = A.opponentProductPrior i)
Instances For
Auctions over the same Bayesian selling environment.
Equations
- A.HasSameSellingEnvironment B = (B.prior = A.prior ∧ B.opponentPrior = A.opponentPrior ∧ B.typeData = A.typeData)
Instances For
Integral against typeMeasure as an interval integral.
Real-valued form of integral_typeMeasure_eq_intervalIntegral_smul.
The reported type profile obtained by combining zᵢ with an opponent
profile t₋ᵢ.
Equations
Instances For
The inserted coordinate is zᵢ.
Other coordinates come from the opponent profile.
Updating coordinate i replaces the inserted report.
Split a profile into coordinate i and its opponents.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Fubini decomposition under the product prior.
Interim expectation of a full-profile observable after report zᵢ.
Equations
- A.interimExpectation i z_i f = ∫ (t : OpponentTypeProfile I i), f (BayesianSingleItemAuction.reportProfile i z_i t) ∂A.opponentPrior i
Instances For
Allocation integrand for interim probability.
Equations
- A.interimAllocationIntegrand i z_i t = A.allocationRule (BayesianSingleItemAuction.reportProfile i z_i t) i
Instances For
Payment integrand for interim payment.
Equations
- A.interimPaymentIntegrand i z_i t = A.paymentRule (BayesianSingleItemAuction.reportProfile i z_i t) i
Instances For
Quasi-linear utility integrand.
Equations
- A.interimQuasiLinearUtilityIntegrand i t_i z_i t = t_i * A.interimAllocationIntegrand i z_i t - A.interimPaymentIntegrand i z_i t
Instances For
Allocation integrands are integrable.
Equations
- A.HasIntegrableInterimAllocation = ∀ (i : I) (z_i : ℝ), MeasureTheory.Integrable (fun (t : OpponentTypeProfile I i) => A.interimAllocationIntegrand i z_i t) (A.opponentPrior i)
Instances For
Payment integrands are integrable.
Equations
- A.HasIntegrableInterimPayment = ∀ (i : I) (z_i : ℝ), MeasureTheory.Integrable (fun (t : OpponentTypeProfile I i) => A.interimPaymentIntegrand i z_i t) (A.opponentPrior i)
Instances For
Allocation and payment integrability.
Equations
Instances For
Bounded a.e. strongly measurable allocation integrands are integrable.
Bounded a.e. strongly measurable payment integrands are integrable.
A.e. strong measurability and bounds imply interim integrability.
Feasibility and a.e. strong measurability imply allocation integrability.
Interim allocation probability q_i(z_i).
Equations
- A.interimAllocProb i z_i = A.interimExpectation i z_i fun (r : I → ℝ) => A.allocationRule r i
Instances For
Interim expected payment m_i(z_i).
Equations
- A.interimExpectedPayment i z_i = A.interimExpectation i z_i fun (r : I → ℝ) => A.paymentRule r i
Instances For
The generic interim expectation specializes to the allocation integrand.
The generic interim expectation specializes to the payment integrand.
Product-side integrability gives interval integrability of m_i f_i.
Feasible auctions have nonnegative interim allocation probabilities.
Feasible auctions have interim allocation probabilities at most 1.
Feasible interim allocation probabilities lie in [0, 1].
Payment revenue through interim payments #
Expected total payment revenue under the base prior.
Instances For
Expected seller revenue of B, evaluated in environment A.
Equations
Instances For
Ex-ante revenue through interim expected payments.
Equations
- A.expectedInterimPaymentRevenue B = ∑ i : I, ∫ (v : ℝ) in 0..A.typeData.omega i, B.interimExpectedPayment i v * A.typeDensity i v
Instances For
Ex-ante revenue agrees with the interim-payment expression.
Equations
Instances For
Payment-side Fubini hypotheses connecting ex-ante and interim expressions.
- payment_integrable (i : I) : MeasureTheory.Integrable (fun (t : I → ℝ) => B.paymentRule t i) A.prior
Instances For
Build payment-side interim Fubini hypotheses from type-measure Fubini.
Independent priors and product-side payment integrability imply the payment-side interim Fubini package.
Supportwise interval integrability of q_i.
Equations
- A.HasIntervalIntegrableInterimAllocationOnSupport = ∀ (i : I), IntervalIntegrable (A.interimAllocProb i) MeasureTheory.volume 0 (A.typeData.omega i)
Instances For
Global interval integrability of q_i.
Equations
- A.HasIntervalIntegrableInterimAllocation = ∀ (i : I) (a b : ℝ), IntervalIntegrable (A.interimAllocProb i) MeasureTheory.volume a b
Instances For
Interim utility q_i(z_i) * t_i - m_i(z_i).
Equations
- A.interimQuasiLinearUtility i t_i z_i = A.interimAllocProb i z_i * t_i - A.interimExpectedPayment i z_i
Instances For
Truthful interim payoff U_i(t_i).
Equations
- A.equilibriumPayoff i t_i = A.interimQuasiLinearUtility i t_i t_i
Instances For
Interim incentive compatibility.
Equations
- A.IsIncentiveCompatible = ∀ (i : I) (t_i z_i : ℝ), A.interimQuasiLinearUtility i t_i z_i ≤ A.equilibriumPayoff i t_i
Instances For
Misreport utility in terms of truthful payoff at the report.
Utility integrability from allocation and payment integrability.
The utility integrand integrates to interim utility.
Pointwise DSIC inequality for the interim utility integrand.
DSIC implies interim IC under interim integrability.
[MSZ 12.48] IC iff the one-dimensional payoff inequality.
IC implies monotonicity of q_i.
IC gives global interval integrability of q_i.
IC makes truthful payoff convex.
Interim individual rationality.
Equations
- A.IsIndividuallyRational = ∀ (i : I) (t_i : ℝ), 0 ≤ A.equilibriumPayoff i t_i
Instances For
Interim individual rationality on [0, ωᵢ].
Equations
- A.IsIndividuallyRationalOnSupport = ∀ (i : I) (t_i : ℝ), 0 ≤ t_i → t_i ≤ A.typeData.omega i → 0 ≤ A.equilibriumPayoff i t_i
Instances For
[MSZ 12.49] Payoff envelope.
Equations
- A.HasInterimEnvelopeFormula = ∀ (i : I) (v_i : ℝ), A.equilibriumPayoff i v_i = A.equilibriumPayoff i 0 + ∫ (z : ℝ) in 0..v_i, A.interimAllocProb i z
Instances For
[MSZ 12.49] Payment identity.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Envelope derivative condition.
Equations
- A.HasInterimEnvelopeDerivative = ∀ (i : I) (v_i : ℝ), HasDerivAt (A.equilibriumPayoff i) (A.interimAllocProb i v_i) v_i
Instances For
IC and differentiability give the envelope derivative.
The envelope derivative gives the payoff envelope.
IC plus the envelope derivative gives the payoff envelope.
[MSZ 12.49] Payoff envelope from IC.
[MSZ 12.146] Payoff identity.
The payoff envelope implies the payment identity.
The envelope derivative implies the payment identity.
[MSZ 12.147] Payment identity from IC.
Envelope and payment identities from IC and the derivative condition.
[MSZ 12.49] Envelope and payment identities from IC.
Same q_i and zero-type payment give the same truthful payoff.
Same q_i and zero-type payment give the same interim payment.
[MSZ 12.50] Same q_i and zero-type payment give the same interim objects.
Nonnegative envelope increments on [0, ωᵢ].
Equations
Instances For
Feasibility gives nonnegative envelope increments on the support.
[MSZ 12.52] Supportwise IR iff zero-type payment is nonpositive.
Supportwise IR makes zero-type expected payment nonpositive.
IC and IR bound interim expected payment by the envelope term.