Myerson–Satterthwaite Theorem

Myerson–Satterthwaite theorem is a foundational impossibility theorem in Bayesian mechanism design and business and economics theory, proved by American economists Roger Myerson and Mark Satterthwaite in their 1983 Journal of Economic Theory article "Efficient Mechanisms for Bilateral Trading." Addressing the classic problem of bilateral bargaining over an indivisible good when a seller's reservation cost and a buyer's reservation valuation are independent private information whose probability distributions overlap, the theorem establishes that no trading institution, bargaining protocol, or auction mechanism can simultaneously satisfy four desirable axioms: (1) Bayesian Incentive Compatibility (truthful reporting of private valuations forms a Bayesian Nash equilibrium), (2) Interim Individual Rationality (voluntary participation by both buyer and seller after learning their own private valuation), (3) Ex-Post Pareto Efficiency (trade occurs if and only if the buyer's valuation exceeds the seller's cost), and (4) Budget Balance (no external third-party subsidy is injected into the trade). Consequently, whenever bilateral private information renders gains from trade uncertain, private bargaining incurs positive transaction costs in the form of deadweight loss from missed mutually beneficial trades—providing the definitive information-theoretic boundary to the frictionless Coase theorem.

Private Valuations and Quasi-Linear Payoffs

Consider a bilateral trading environment in which a single seller (Player 1) owns one indivisible unit of an asset and negotiates with a single potential buyer (Player 2):

• The seller's private reservation cost is a random variable v₁ ∈ [a₁, b₁] drawn independently from a cumulative distribution function F₁(v₁) with continuous positive probability density f₁(v₁) > 0 on [a₁, b₁].
• The buyer's private reservation valuation is a random variable v₂ ∈ [a₂, b₂] drawn independently from a cumulative distribution function F₂(v₂) with continuous positive density f₂(v₂) > 0 on [a₂, b₂].
• The distributions F₁ and F₂ are common knowledge, and their supports have a non-empty interior intersection (a₁, b₁) ∩ (a₂, b₂) ≠ ∅, so that Pr(v₂ > v₁) > 0 and Pr(v₂ < v₁) > 0.

Both agents have risk-neutral, quasi-linear utility in money. If the asset is transferred with probability p ∈ [0, 1], the buyer pays monetary transfer t₂ ∈ ℝ, and the seller receives monetary transfer t₁ ∈ ℝ, their realized payoffs are u₁ = t₁ − v₁p and u₂ = v₂p − t₂.

Direct Revelation Mechanisms and the Four Axioms

By Roger Myerson's (1979) Revelation Principle, any Bayesian Nash equilibrium outcome of any multi-stage bargaining game can be replicated without loss of generality by a direct revelation mechanism ⟨p(v₁, v₂), t₁(v₁, v₂), t₂(v₁, v₂)⟩ in which both parties simultaneously report their types (v̂₁, v̂₂) and truthful reporting is an equilibrium. Define the interim expected trading probabilities P₁(v̂₁) = E_v2[p(v̂₁, v₂)] and P₂(v̂₂) = E_v1[p(v₁, v̂₂)], interim expected transfers T₁(v̂₁) and T₂(v̂₂), and equilibrium interim expected utilities U₁(v₁) = T₁(v₁) − v₁P₁(v₁) and U₂(v₂) = v₂P₂(v₂) − T₂(v₂). The four institutional axioms are:

1. Bayesian Incentive Compatibility (BIC):

1. Interim Individual Rationality (IIR / Voluntary Participation):

1. Ex-Post Pareto Efficiency (EPE):

1. Weak Budget Balance (BB):

Step 1: Envelope Characterization and Expected Budget Surplus

Just as in the Revenue Equivalence Theorem, Bayesian incentive compatibility implies monotonicity of interim allocation rules (P₁(v₁) is non-increasing and P₂(v₂) is non-decreasing) together with the Envelope Theorem for equilibrium information rents:

Taking expectations over F₁(v₁) and F₂(v₂) and integrating by parts yields the Myerson–Satterthwaite feasibility identity: a monotonic trading rule p(v₁, v₂) can be implemented by a BIC, IIR, and budget-balanced mechanism (where U₁(b₁) ≥ 0, U₂(a₂) ≥ 0, and E[t₂ − t₁] ≥ 0) if and only if:

where ψ₂(v₂) = v₂ − (1 − F₂(v₂)) / f₂(v₂) is the buyer's virtual valuation and ψ₁(v₁) = v₁ + F₁(v₁) / f₁(v₁) is the seller's virtual cost.

Step 2: Evaluating the Ex-Post Efficient Rule p^*(v₁, v₂) = 𝕀(v₂ ≥ v₁)

Substituting the ex-post Pareto efficient allocation rule p^*(v₁, v₂) = 1 whenever v₂ ≥ v₁ into the integral and integrating by parts over the overlap interval [a, b] = [a₁, b₁] ∩ [a₂, b₂] yields a negative definite integral:

Because F₁(x) > 0 and 1 − F₂(x) > 0 on (a, b), the right-hand side is negative. Consequently, if individual rationality holds (U₁(b₁) ≥ 0 and U₂(a₂) ≥ 0), any ex-post efficient mechanism must run an expected budget deficit E[t₂ − t₁] ≤ −∫_a^b F₁(x)(1 − F₂(x))*dx* < 0, which equals the expected deficit of the two-sided Vickrey–Clarke–Groves mechanism.

When v₁ and v₂ are drawn independently from the uniform distribution U[0, 1], first-best expected gains from trade are ∫₀¹ ∫_v1¹ (v₂ − v₁)*dv*₂dv₁ = 1/6 ≈ 0.1667, and the VCG subsidy required for first-best efficiency is also ∫₀¹ x(1 − x)*dx* = 1/6 ≈ 0.1667.

Myerson and Satterthwaite (1983) solved for the second-best mechanism that maximizes expected gains from trade subject to BIC, IIR, and budget balance, finding that trade should occur if and only if v₂ − v₁ ≥ 1/4 = 0.25. Kalyan Chatterjee and William Samuelson (1983) proved that this second-best allocation is implemented by the split-the-difference double auction (P = (b + s) / 2 whenever b ≥ s), in which agents shade their bids and asks via the linear equilibrium strategies b^*(v₂) = (2/3)v₂ + 1/12 and s^*(v₁) = (2/3)v₁ + 1/4, achieving expected social welfare of 9/64 ≈ 0.1406 (an unavoidable deadweight loss of 15.625% of first-best surplus).

Relaxing any of the structural assumptions of the Myerson–Satterthwaite theorem reveals institutional escape routes:

1. Non-Overlapping Supports: If the minimum buyer valuation exceeds the maximum seller cost (a₂ ≥ b₁), any fixed price P ∈ [b₁, a₂] achieves 100% efficiency, budget balance, and dominant-strategy incentive compatibility.
2. Ex-Ante Participation Constraint: If parties contract before learning their private valuations, the d'Aspremont–Gérard-Varet mechanism achieves first-best efficiency and budget balance via ex-ante lump-sum transfers.
3. Symmetric Partnership Dissolution (Cramton, Gibbons, and Klemperer, 1987): Whereas 100%–0% ownership places worst-off types at the boundaries v₁ = b₁ and v₂ = a₂, if two partners initially share 50%–50% ownership of a joint venture, each partner's worst-off intermediate type has a lower reservation utility, allowing a partnership dissolution auction to achieve ex-post Pareto efficiency, BIC, IIR, and budget balance simultaneously.
4. Asymptotic Efficiency in Large Double Auctions (Rustichini, Satterthwaite, and Williams, 1994): With M buyers and M sellers, strategic bid-shading shrinks at rate O(1/M) and inefficiency shrinks at rate O(1/M²), converging to Arrow–Debreu model competitive efficiency.

How does the Myerson–Satterthwaite theorem qualify the Coase theorem?

The Coase theorem (1960) states that if property rights are well-defined and transaction costs are zero, private bargaining will reach a Pareto-efficient allocation regardless of who initially holds the property right. The Myerson–Satterthwaite theorem proves two qualifications under asymmetric information: first, private information is an endogenous transaction cost that prevents 100% efficient bargaining whenever valuation supports overlap; second, the initial allocation of ownership rights matters for efficiency, as Cramton, Gibbons, and Klemperer (1987) proved that 50–50 shared ownership can achieve full efficiency where 100–0 ownership cannot.

Why can't buyer and seller simply use a Vickrey auction to trade efficiently?

In a two-sided Vickrey–Clarke–Groves mechanism, truthful reporting requires that when v₂ ≥ v₁, the buyer pays the seller's reported cost v₁ while the seller receives the buyer's reported valuation v₂. Because v₂ > v₁ whenever trade occurs, the mediator must pay out v₂ to the seller while collecting only v₁ from the buyer, losing v₂ − v₁ > 0 on every trade and violating budget balance.

Why do traders shade their bids and asks in a double auction?

In a budget-balanced double auction where the transaction price lies between the buyer's bid and the seller's ask (such as P = (b + s) / 2), each trader faces a trade-off between the probability of trade and the price of trade. By bidding slightly below their true valuation v₂, a buyer reduces the price paid on all trades that still go through, losing only marginal trades where the seller's ask was nearly equal to v₂. Because this marginal price benefit is first-order while the lost-trade cost is second-order at b = v₂, every buyer with v₂ > a₂ shades their bid downward and every seller with v₁ < b₁ marks their ask upward.

1. Myerson, R. B., & Satterthwaite, M. A. (1983). "Efficient Mechanisms for Bilateral Trading." Journal of Economic Theory, 29(2), 265–281. https://doi.org/10.1016/0022-0531(83)90048-090048-0)
2. Chatterjee, K., & Samuelson, W. (1983). "Bargaining Under Incomplete Information." Operations Research, 31(5), 835–851. https://doi.org/10.1287/opre.31.5.835
3. Cramton, P., Gibbons, R., & Klemperer, P. (1987). "Dissolving a Partnership Efficiently." Econometrica, 55(3), 615–632. https://doi.org/10.2307/1913602
4. Rustichini, A., Satterthwaite, M. A., & Williams, S. R. (1994). "Convergence to Efficiency in a Simple Market with Incomplete Information." Econometrica, 62(5), 1041–1063. https://doi.org/10.2307/2951506
5. Myerson, R. B. (1979). "Incentive Compatibility and the Bargaining Problem." Econometrica, 47(1), 61–73. https://doi.org/10.2307/1912346