Myerson-Satterthwaite Theorem

• Myerson–Satterthwaite Theorem is a foundational result in mechanism design, proving that no mechanism can meet Bayesian incentive compatibility, individual rationality, weak budget balance, and ex-post efficiency in bilateral trade.
• It employs rigorous strategies such as direct revelation mechanisms, envelope techniques, and virtual values to analyze the trade-offs between truthful reporting and efficient allocation under private information.
• Extensions of the theorem highlight approximation mechanisms and large-market designs that recover a constant fraction of first-best gains from trade despite the inherent impossibility.

The Myerson–Satterthwaite theorem is a foundational impossibility result in mechanism design for bilateral trade. In its canonical form, there is one buyer with private value $v$ and one seller with private cost $c$ for a single indivisible good, with independent private values drawn from continuous distributions whose supports overlap. The theorem states that no mechanism can simultaneously be Bayesian incentive compatible, individually rational, weakly budget balanced, and ex-post efficient, where efficiency means trading whenever $v \ge c$. Equivalently, no such mechanism can always attain the first-best gains from trade, $G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$ (Deng et al., 2021, Rubbini, 2023).

1. Formal environment and theorem statement

The standard environment consists of a direct revelation mechanism with allocation rule $q(\hat v,\hat c)\in[0,1]$, interpreted as the probability of trade, and transfer rules to the buyer and seller. Under truthful reporting, quasilinear utilities are

$$u_b(v)=\mathbb{E}_c[v\cdot q(v,c)-p_b(v,c)],\qquad u_s(c)=\mathbb{E}_v[p_s(v,c)-c\cdot q(v,c)].$$

The efficient allocation trades whenever the buyer’s value exceeds the seller’s cost:

$q^*(v,c)=\mathbf{1}\{v\ge c\}.$

The corresponding first-best gains from trade are

$$G^*(F,G)=\mathbb{E}_{v\sim F,\;c\sim G}\big[(v-c)\mathbf{1}\{v\ge c\}\big].$$

For a mechanism $M$ with allocation rule $q_M$, achieved gains from trade are

$c$0

Under independent private values with continuous distributions and overlapping supports, Myerson–Satterthwaite rules out any mechanism satisfying BIC, IR, and weak budget balance that implements $c$1 almost everywhere (Deng et al., 2021).

A parallel formulation uses a binary allocation $c$2 and transfers $c$3, with ex-post efficiency given by $c$4 iff $c$5. In that notation, the theorem states that there does not exist $c$6 such that ex-post efficiency, BIC, interim IR for both agents, and budget balance all hold together (Rubbini, 2023).

2. Constraint system and benchmark interpretation

The theorem is defined by the conjunction of four requirements. Incentive compatibility can be posed either as DSIC or BIC; the classical formulation is typically BIC, with truthful reporting maximizing expected utility against the prior over the opponent’s type. Individual rationality requires nonnegative utility from truthful participation, either ex-post or interim/ex-ante depending on the formulation. Budget balance excludes outside subsidies: weak budget balance requires the mechanism not to run a deficit, while strong budget balance requires that buyer payments equal seller receipts pointwise (Deng et al., 2021).

The central benchmark is not merely allocative efficiency but the full first-best welfare generated by the efficient trade rule. In bilateral trade, this benchmark is especially transparent because the realized surplus from a trade is exactly $c$7. The theorem therefore has two equivalent readings. One is implementation-theoretic: the efficient allocation rule cannot be truthfully and budget-balancingly implemented. The other is welfare-theoretic: no BIC, IR, weakly budget balanced mechanism can guarantee

$c$8

for all independent continuous distributions with overlapping support (Deng et al., 2021).

A recurring distinction in later work is between the first-best benchmark and the “second best,” defined as the maximum expected gains from trade attainable subject to IC, IR, and budget balance. This distinction matters because much of the subsequent literature measures mechanisms either against first-best gains from trade or against the second-best feasible optimum.

3. Economic logic and proof architecture

The economic intuition is the tension between information rents and budget balance. Efficiency requires trade whenever $c$9, including states where buyer and seller types are arbitrarily close. But truthful revelation under private information requires informational rents for both sides, while budget balance forbids external subsidies. In a bilateral environment, there is only one realized unit of surplus, $v \ge c$0, per trade. When $v \ge c$1 and $v \ge c$2 are close, the mechanism cannot simultaneously preserve truthful revelation, keep both agents willing to participate, and finance the induced rents without sometimes suppressing trade (Deng et al., 2021).

A sharper analytical version of this argument uses virtual values. Under regularity and BIC, envelope arguments and payment identities imply that, up to IR boundary terms, expected total transfers equal expected virtual surplus induced by the allocation rule. Defining buyer virtual value and seller virtual cost by

$v \ge c$3

one obtains that under minimal IR the expected total transfer equals expected virtual surplus. Ex-post efficiency requires trading on $v \ge c$4, but overlapping supports imply that there is a non-negligible subset of that efficient region on which $v \ge c$5. Trading there makes expected virtual surplus negative, which is incompatible with budget balance together with IR (Rubbini, 2023).

This proof architecture identifies the theorem as more than a peculiarity of a specific payment formula. It is a structural incompatibility between ex-post efficient trade and truthful, budget-balanced implementation in the independent-private-values bilateral setting.

4. Approximation after impossibility

Once full efficiency is ruled out, the natural question becomes approximation to first-best gains from trade. Deng, Mao, Sivan, and Wang show that simple posted-price delegation mechanisms can guarantee a universal constant fraction of first-best gains from trade under the standard bilateral assumptions. Let SellerP denote the gains from trade achieved when the seller, after observing cost $v \ge c$6, posts

$v \ge c$7

and BuyerP denote the symmetric buyer-posted-price mechanism with

$v \ge c$8

Their main bound is

$v \ge c$9

so choosing the better of SellerP and BuyerP yields at least a $G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$0 fraction of first-best. The analysis is tightened to

$G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$1

giving an approximation factor $G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$2. These mechanisms are BIC, IR, and ex-post strongly budget balanced; the guarantees are universal over all independent continuous $G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$3 on $G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$4 with overlapping supports, and do not require regularity assumptions such as MHR (Deng et al., 2021).

Subsequent work improves the bilateral constant substantially. The random-offerer mechanism flips a fair coin between seller-pricing and buyer-pricing, and it is shown to satisfy

$G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$5

Equivalently,

$G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$6

The same paper also determines the exact worst-case approximation guarantee of seller-pricing under a monotone hazard rate assumption on the buyer’s distribution:

$G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$7

and this bound is tight (Fei, 2022).

These results reframe the theorem’s significance. Myerson–Satterthwaite excludes exact efficiency under the standard constraints, but it does not preclude constant-factor welfare guarantees from simple, transparent mechanisms.

5. Robustness beyond rational-expectations equilibrium

A later line of work asks whether the impossibility depends on rational expectations. The answer, in a broad sense, is negative. In a generalized model of implementation without rational expectations, a solution concept $G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$8 specifies permitted expectation profiles and response correspondences without requiring that agents’ beliefs about strategies be correct. The key condition is Weak Solution Consistency (WSC): for every mechanism and every type, there exists some solution in which that type weakly prefers its own-type outcome to mimicking any other type (Rubbini, 2023).

Under WSC, full implementation of a social choice function still requires Bayesian incentive compatibility. Therefore, efficient bilateral trade cannot be fully implemented even under a broad family of non-equilibrium or boundedly rational solution concepts, including level-$G^*(F,G)=\mathbb{E}[(v-c)\mathbf{1}\{v\ge c\}]$9 reasoning and interim correlated rationalizability. The paper states a corollary for the bilateral trade environment: with independent private values and overlapping supports, there is no mechanism that fully implements the ex-post efficient social choice function while satisfying BIC, interim IR, and budget balance, whether budget balance is ex-ante or ex-post (Rubbini, 2023).

This extends the theorem’s domain of relevance. A plausible implication is that the impossibility is not merely an artifact of Bayesian Nash equilibrium or of fully correct strategic beliefs; it persists under much weaker behavioral foundations when full implementation is demanded.

6. Boundary cases, non-equivalences, and altered environments

The theorem is specific to one buyer, one seller, one indivisible good, private information, overlapping support, and budget balance. Changing those ingredients can change the conclusion, but such changes do not contradict the classical statement.

One important boundary case is fixed exogenous prices. In the single-unit case, if a price $q(\hat v,\hat c)\in[0,1]$0 is fixed independently of reports, the mechanism “trade iff $q(\hat v,\hat c)\in[0,1]$1” is DSIC, IR, and strongly budget balanced, and it achieves efficiency subject to IR at that price. However, this escape disappears once the environment becomes multi-dimensional. With $q(\hat v,\hat c)\in[0,1]$2 units of the same good or $q(\hat v,\hat c)\in[0,1]$3 kinds of goods, every DSIC mechanism has competitive ratio at most $q(\hat v,\hat c)\in[0,1]$4, and in the single-good multi-unit case with submodular valuations the upper bound improves to $q(\hat v,\hat c)\in[0,1]$5, where $q(\hat v,\hat c)\in[0,1]$6 is the $q(\hat v,\hat c)\in[0,1]$7-th harmonic number; both bounds are tight (Segal-Halevi et al., 2017).

A second alteration is market size. In double auctions and matching markets, mechanisms can be ex-post IR, BIC, and ex-post weakly budget balanced while achieving both an ex-ante constant fraction of second-best gains from trade and an ex-post fraction of first-best that converges to $q(\hat v,\hat c)\in[0,1]$8 as markets grow. In the double-auction setting, the trade-reduction branch achieves an ex-post factor $q(\hat v,\hat c)\in[0,1]$9 of first-best when the efficient number of trades is $$u_b(v)=\mathbb{E}_c[v\cdot q(v,c)-p_b(v,c)],\qquad u_s(c)=\mathbb{E}_v[p_s(v,c)-c\cdot q(v,c)].$$0, and in matching markets an analogous factor is $$u_b(v)=\mathbb{E}_c[v\cdot q(v,c)-p_b(v,c)],\qquad u_s(c)=\mathbb{E}_v[p_s(v,c)-c\cdot q(v,c)].$$1 based on class sizes and compatibility degrees (Babaioff et al., 2018).

These cases clarify a common misconception. The theorem does not say that truthfulness and budget balance always preclude high efficiency in all exchange environments; rather, it identifies a sharp impossibility for the bilateral, single-item, private-values setting. Fixed-price single-unit trade changes the target, while large-market double auctions change the environment.

7. Information requirements and prior dependence

A different response to the theorem concerns informational assumptions. In bilateral trade and richer two-sided markets, abandoning priors altogether is too costly: no prior-independent mechanism with no prior information can be IC, IR, budget balanced, and guarantee any constant-factor approximation to optimal social welfare. This is presented as a generalized impossibility result that strengthens the practical force of Myerson–Satterthwaite by showing that, without any prior information, even approximate efficiency is impossible (Dütting et al., 2020).

At the same time, extremely limited prior information can suffice for approximation in broader two-sided settings. For subadditive buyers and additive sellers, an adjusted-objective VCG mechanism using a single sample from each seller distribution is IR, DSIC, ex-post budget balanced, and achieves a $$u_b(v)=\mathbb{E}_c[v\cdot q(v,c)-p_b(v,c)],\qquad u_s(c)=\mathbb{E}_v[p_s(v,c)-c\cdot q(v,c)].$$2-approximation to optimal welfare in expectation. The same work gives a matching lower bound: every deterministic IC, IR, budget-balanced bilateral-trade mechanism that receives as sole prior information a single seller sample has approximation ratio at least $$u_b(v)=\mathbb{E}_c[v\cdot q(v,c)-p_b(v,c)],\qquad u_s(c)=\mathbb{E}_v[p_s(v,c)-c\cdot q(v,c)].$$3 (Dütting et al., 2020).

This suggests a broader interpretation of the theorem’s legacy. The impossibility is not only a boundary on exact efficiency; it also organizes the modern study of which relaxations—approximation, larger markets, weaker implementation goals, or minimal statistical information—are sufficient to recover meaningful welfare guarantees without violating incentive compatibility, individual rationality, and budget balance.