Auction Theory: Four Formats and the Revenue Equivalence Theorem

Four formats, one underlying question

An auction is a mechanism for discovering a price when the seller does not know what buyers are willing to pay. The four classic formats differ in how bids are revealed and how the winner pays, and each format changes what a rational bidder should actually do:

English   — open ascending bids, highest bidder wins, pays their final bid
Dutch     — price starts high and falls, first bidder to accept wins, pays that price
Sealed-bid (first-price)  — bids hidden, highest wins, pays their own bid
Vickrey (second-price)    — bids hidden, highest wins, pays the SECOND-highest bid

Why Vickrey auctions make honesty the best policy

In a sealed first-price auction, bidding your true value is a losing strategy — you would pay exactly what you bid, so bidding below your true valuation is the only way to leave any surplus on the table for yourself, and every rational bidder shades their bid downward. William Vickrey's insight in 1961 was that changing the payment rule, not the bidding process, fixes this: in a second-price auction, you win with your bid but pay the runner-up's bid instead, which means bidding your true value is a dominant strategy — bidding higher risks overpaying for a win you would have gotten anyway, and bidding lower only risks losing an auction you would have profitably won. No amount of strategizing about what others might bid changes that conclusion.

Revenue equivalence: format matters less than you'd think

The revenue equivalence theorem (Vickrey, later generalized by Myerson and Riley) states that under fairly general conditions — risk-neutral bidders, independent private valuations drawn from the same distribution, and a rule that the highest-value bidder always wins — all four of these auction formats yield the same expected revenue to the seller, and the same expected payment from any given bidder. This is deeply counter-intuitive on first encounter, because the formats look so different, but it follows from each bidder's optimal strategy adjusting to exactly offset the format's rules: sealed first-price bidders shade down to compensate for paying their own bid, but the second-highest valuation ends up as the expected winning payment regardless of format.

Where revenue equivalence breaks down

The theorem's assumptions are exactly where it gets interesting to violate. Risk-averse bidders bid more aggressively in a first-price sealed auction than the theorem predicts, because a slightly higher bid buys a real reduction in the risk of losing, which risk-neutral bidders would not pay extra for — so first-price and Dutch auctions tend to raise more revenue than English or Vickrey auctions once bidders are risk-averse. Correlated valuations (bidders whose private information is linked, as in most real auctions where a shared uncertain resale value exists) break the independence assumption and open the door to the winner's curse — the winning bidder is, by definition of having the highest estimate among many, statistically likely to have overestimated the item's true value, and sophisticated bidders must shade their bids to correct for it.

Dutch and English are not mirror images in practice

Even though English and Dutch auctions are provably revenue-equivalent to sealed-bid formats under the theorem's assumptions, they differ practically: an English auction reveals information as it proceeds — bidders watch the price climb and can infer something about others' valuations — while a Dutch auction reveals nothing until the moment someone accepts. This is why English auctions dominate in settings like art and antiques where bidders benefit from observing rivals, while Dutch (descending) formats are used where speed matters, such as flower and fish markets, where a single fast decline finds the clearing price with less time spent per lot.

Frequently asked questions

Try it live

Everything above runs in your browser — open Auction Dynamics and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.