Skip to content
Tutorials Center
My learning

Topics

Lesson 29 of 37 ยท Mechanism Design and Auctions

Revenue Equivalence and Optimal Auctions

Why first-price, second-price and all-pay auctions raise the same revenue, and how a reserve price from Myerson's optimal auction theory raises it. Python simulations included.

A seller has two questions: who gets the item and how much will I earn? The previous lessons showed that first-price and second-price auctions give the same average revenue with uniform values. That is not a coincidence. It is a theorem, and understanding it tells a seller exactly which design choices matter. Spoiler: the format of the auction usually does not, but the reserve price does.

The revenue equivalence theorem

Suppose:

  • the bidders are risk-neutral,
  • their values are independent draws from the same continuous distribution,
  • the auction gives the item to the bidder with the highest value (it is efficient), and
  • a bidder with the lowest possible value expects zero profit.

Then every such auction, whatever its format (first-price, second-price, English, Dutch, even all-pay auctions where everyone pays their bid), gives the seller the same expected revenue. The same holds for each bidder's expected payment as a function of their value.

The reason is that the equilibrium payment of a bidder with value v is determined by how likely they are to win, and the winning probability is fixed by "highest value wins". Only the way the payment is collected differs.

Revenue equivalence

Highest value wins Same allocation rule
Same expected payment For each value type
Different formats First price, second price, all-pay
Same revenue On average

Rules change who pays when, not the average.

1 / 2

Testing it, including all-pay auctions

In an all-pay auction every bidder pays their bid, and the highest bid wins (like a lobbying contest). With n bidders and values uniform on [0, 1], the symmetric equilibrium bid is b(v) = (n - 1) / n * v to the power n. Compare the three formats.

Python Three auctions, one revenue

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

All three land at about 0.6. The all-pay auction looks very different (losers pay too!), but the average seller revenue matches.

When equivalence fails

The theorem needs its assumptions. Revenue differs when:

  • bidders are risk-averse (first-price then raises more: shading is riskier),
  • bidders are asymmetric (different distributions),
  • values are correlated or common (English beats sealed-bid),
  • there are budget limits or entry costs,
  • the seller can set a reserve price that is high.

Optimal auctions: the reserve price

Roger Myerson showed how to design the revenue-maximizing auction. The seller turns each bidder's value v into a virtual value

phi(v) = v - (1 - F(v)) / f(v)

where F is the distribution of values and f its density. The virtual value is the value minus an "information rent" the seller has to leave to higher types. The optimal auction gives the item to the bidder with the highest virtual value, provided it is non-negative. When values are uniform on [0, 1], F(v) = v and f = 1, so phi(v) = 2v - 1. This is non-negative exactly when v >= 1/2, so the optimal reserve price is 0.5, regardless of the number of bidders. Because phi is increasing and identical across bidders, the optimal allocation is still "highest value wins", but only if that value is at least 0.5.

Why is it worthwhile to sometimes keep the item? A reserve of 0.5 gives up sales to buyers with values below 0.5 (a small loss), but forces the others to pay at least 0.5 and pushes up the second-highest bid.

Python Finding the optimal reserve price

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

With two bidders, a reserve of 0.5 lifts expected revenue from 0.333 to 0.417, a 25 percent gain, for free. This is why eBay sellers set a minimum price and why governments set reserve prices in licence auctions.

The reserve also fits other formats

A first-price auction with the same reserve also yields 5/12 if bidders use the right equilibrium bid. For uniform values and a reserve r, the bid is b(v) = v - (v^n - r^n) / (n v^(n-1)) for v >= r.

Python First price with a reserve matches second price with a reserve

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The numbers match (both near 0.4167), again as revenue equivalence predicts, since both mechanisms allocate to the highest value above 0.5.

Lessons for designers

  • Spend your effort on the reserve price and on attracting bidders, not on picking between first and second price.
  • One additional bidder is typically worth more than any fine-tuned rule: in the second-price case with uniform values, revenue is (n - 1) / (n + 1), so it rises from 1/3 (n = 2) to 1/2 (n = 3).
  • The optimal auction is sometimes inefficient: the seller refuses to sell to some buyers who value the item more than the seller does. Revenue and efficiency are different goals.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. What does revenue equivalence say?

    The theorem has assumptions: risk neutrality, independent private values from the same distribution, and a highest-value-wins allocation.
  2. 2. For values uniform on [0, 1], what is the optimal reserve price?

    The virtual value 2v - 1 is zero at v = 0.5, so the seller should not sell below 0.5.
  3. 3. Which assumption, when violated, can break revenue equivalence?

    Risk-averse bidders shade less in first-price auctions, so first-price raises more than second-price.
  4. 4. Why might an optimal auction be inefficient?

    A reserve price sacrifices some gains from trade to raise the seller's expected revenue.

Summary

  • Revenue equivalence: same allocation rule plus a zero-profit lowest type implies the same expected revenue across auction formats.
  • First-price, second-price and all-pay auctions yield (n - 1) / (n + 1) for uniform values.
  • The theorem fails for risk-averse, asymmetric or correlated bidders.
  • Myerson's optimal auction uses virtual values; for U[0, 1] the optimal reserve is 0.5.
  • Reserve prices and bidder participation matter more than the format.

Next: VCG and Mechanism Design