Does the theorem require identical bids across formats?

No. Different equilibrium strategies are central to the standard comparison.

Can reserve prices affect expected revenue?

Yes. A reserve can change allocation and participation conditions, so equivalence with an unrestricted auction does not follow automatically.

Is it relevant to AI?

Yes. Auction mechanisms and strategic agents appear in resource allocation and multi-agent systems, where benchmark assumptions must be explicit.

AI Glossary / Mathematics and optimization / Advanced

Revenue equivalence theorem

Definition

The revenue equivalence theorem states that, under specified auction-model assumptions, mechanisms with the same allocation probabilities and the same expected utility for each bidder's lowest type generate the same expected payments. A familiar consequence is equal expected seller revenue across several standard single-item auction formats.

What it means

A first-price auction makes the winner pay their own bid. A second-price auction makes the winner pay the second-highest bid. Their payment rules differ, but bidders also change their strategies. Under a standard benchmark, those strategic changes balance out in expected revenue.

The benchmark commonly assumes a fixed number of risk-neutral bidders whose private values are independent draws from the same continuous distribution. Standard equilibrium implementations award the item to the highest-value bidder and give the lowest possible valuation type zero expected surplus.

The conclusion concerns an expectation over the modeled uncertainty. It does not say that the seller receives the same amount in every individual auction, that bids are identical, or that real auction platforms are interchangeable.

How it works

Let q(v) be a bidder's probability of receiving the item when their private value is v. Let p(v) be their expected payment and U(v)=vq(v)−p(v) their expected utility under risk-neutral, quasilinear preferences.

In the appropriate incentive-compatible private-value setting, an envelope argument gives U(v)=U(v₀)+∫ from v₀ to v of q(t)dt, where v₀ is the lowest value. Consequently, q and the boundary utility determine p. Mechanisms that share both cannot freely choose different expected payments while preserving those incentive conditions.

The standard first-price and second-price comparison applies this idea through their equilibrium allocation rules. Bidders in the first-price format shade bids below their values, while truthful bidding is weakly dominant in the ordinary second-price private-value auction.

Changing reserve prices can change allocation probabilities. Entry rules can change the bidder population. Risk preferences, correlated information, budgets and non-equilibrium behavior can also invalidate a particular benchmark comparison. The theorem must be matched to a complete model rather than applied from an auction's name alone.

Three practical examples

A marketplace research team uses revenue equivalence as a benchmark before investigating why two simulated auction formats produce different average revenues.
A multi-agent systems class compares strategic bidding policies in first-price and second-price environments.
An auction designer checks whether a proposed reserve changes who receives the item before claiming equivalence with a no-reserve mechanism.

Think of it this way

Think of two games that award the same prize under the same underlying chances, but use different payment instructions. Players adjust how they participate, so comparing the instructions without comparing their strategies misses the result.

The analogy is limited: auction equivalence requires precise preference, information and incentive assumptions. It is not a general claim that different payment systems always cancel out.

Worked example

Two uniform private values

Assume two risk-neutral bidders independently draw values uniformly from 0 to 1, with no reserve price. In the symmetric first-price equilibrium, each bids half their value. The winner pays half the larger value.

The maximum of the two values has cumulative distribution t² on [0,1], so its density is 2t. Its expected value is ∫₀¹t·2t dt=2/3. Expected first-price revenue is therefore (1/2)(2/3)=1/3.

In the second-price auction, truthful bids make the payment the smaller value. The minimum has density 2(1−t), giving expected revenue ∫₀¹t·2(1−t)dt=1/3.

The expectations match, but individual outcomes need not. With values 0.8 and 0.6, the first-price equilibrium bids are 0.4 and 0.3, producing revenue 0.4. Truthful second-price bidding produces revenue 0.6. This example demonstrates equal averages under the stated model, not identical payments for a fixed pair of values.

Illustrative case study

A fictional compute-allocation research platform compares two single-slot auction rules using simulated agents. An initial test reuses truthful bids in both formats and reports different average revenue as a contradiction of the theorem.

The researchers identify the mistake: truthful bidding is not the stipulated first-price equilibrium strategy in their benchmark. They rerun the conceptual comparison with the appropriate strategy and separately document experiments with learning agents that may not reach equilibrium.

The team then adds budgets and variable participation as new experimental conditions. It treats those as changes to the model requiring fresh analysis, rather than assuming the benchmark theorem covers them automatically.

Common misconceptions and mistakes

Equal expected revenue means equal revenue in each auction. The worked example shows otherwise.
First-price bidders should bid truthfully because second-price bidders do. The strategic incentives differ.
Any two mechanisms are revenue equivalent. Allocation and boundary-utility conditions matter.
The theorem proves all real formats perform equally. Actual information, constraints and behavior may differ from the benchmark.
Revenue equivalence means revenue optimality. Equivalent mechanisms need not maximize revenue among all possible designs.

Business use

The theorem is useful for framing questions about auction mechanisms used in digital marketplaces and autonomous-agent research. It helps separate differences caused by payment labels from differences caused by participation, information or strategic behavior.

For an operational comparison, specify the allocation rule, bidder population, value model, reserve policy and behavioral assumptions. Measure absolute outcomes and variability as well as averages. A simulation that uses a particular bidding policy establishes results for that policy; it does not establish an equilibrium claim merely by calling the agents rational.

Check your understanding

In the worked example, one format earns 0.4 and the other 0.6 for a particular value pair. Has revenue equivalence failed?

Answer: No. The theorem compares expected revenue across the value distribution under its assumptions, not every individual realization.

Related terms

Mechanism design
Expected utility
Auction
Bayesian Nash equilibrium
private value
incentive compatibility

Learn next

Read first-price auction and second-price auction. Recalculate the two realized payments for values 0.9 and 0.2, then explain why that single comparison cannot determine which format has higher expected revenue.

Sources

Cornell Mathematics, Revenue Equivalence: https://pi.math.cornell.edu/~mec/Winter2009/Spulido/Auctions/revenue.html. University teaching treatment of the standard benchmark.

Brunnermeier, chapter excerpt, pages 185–186: https://www.princeton.edu/~markus/research/papers/book_chapter6_test.pdf. Allocation, boundary utility and expected-payment argument reviewed. Calculations and fictional platform here are teaching illustrations.

Previous: Revelation principle Next: Reverse data management