Skip to content
Tutorials Center
My learning

Topics

Lesson 28 of 37 ยท Mechanism Design and Auctions

The Vickrey Second-Price Auction

Prove and test that bidding your true value is a dominant strategy in a second-price auction, with Python checks, edge cases, collusion issues and sponsored-search use.

The Vickrey auction, named after the economist William Vickrey, has a lovely property: telling the truth is the best strategy no matter what anyone else does. You do not need to guess the opponents' values, their strategies or even how many there are. This makes it a model case of a strategy-proof mechanism, and it is the starting point for the general VCG mechanism later in this chapter.

The rules

Each bidder submits one sealed bid. The highest bidder wins the item and pays the second-highest bid. Losers pay nothing. A bidder with value v who wins and pays p gets v - p; a loser gets 0. Ties are broken at random.

The payment does not depend on your own bid, except through whether you win. Your bid only decides whether you get the item, not what you pay. That is what removes the temptation to shade.

Why truth-telling is dominant

Fix any bid by the others, and let m be the highest of their bids. Compare bidding your value v with bidding some other number b.

  • If v > m: truthful bidding wins and pays m, earning v - m > 0. Any bid above m wins at the same price (nothing changes). A bid below m loses and earns 0, which is worse.
  • If v < m: truthful bidding loses and earns 0. A bid above m would win and pay m, earning v - m < 0, which is worse. Other bids still lose.
  • If v = m: the payoff is zero either way.

So bidding v is never worse, and sometimes strictly better, than any alternative, whatever the others do. That is the definition of a weakly dominant strategy.

Why truth wins

Price is the rival's bid You never pay your own bid
Overbidding Risks winning at a loss
Underbidding Risks losing a profitable win
Bid = value Never worse, whatever others do
1 / 2

A brute-force check

Instead of trusting the argument, test it. The program below fixes a value, then for every possible bid by you and every possible highest rival bid on a grid, compares your utility from bidding the truth with the utility from deviating.

Python Truthful bidding is never beaten

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 count of violations is zero. Now try the same test for a first-price auction (you pay your own bid) and you will find many violations, because shading beats truthful bidding there.

Python The same test fails for first-price auctions

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.

Other equilibria and edge cases

The truthful strategy is dominant, but a second-price auction also has other Nash equilibria. For instance, one bidder bids a huge amount while all the others bid zero: the high bidder wins at a price of zero and nobody can profit by changing. Those equilibria use non-dominant strategies and look strange, but they matter when a seller worries about coordination among bidders.

Other things to keep in mind:

  • Private values. The proof requires that your value does not depend on what others know. With common values, the dominance result fails.
  • Collusion. A group can agree that one member bids high and the others bid zero, and split the savings. The Vickrey auction is vulnerable to such rings.
  • Seller cheating. The seller sees the bids. If bids are not verifiable, the seller could insert a fake bid just below the winner's to raise the price. The auction is therefore sometimes hard to use in practice, and sealed-bid commitments (cryptographic) help.
  • Low revenue. With few bidders the second-highest bid can be very low. A reserve price r (no sale below r; the winner pays max(second bid, r)) helps, as discussed in the next lesson.

Where the second-price idea appears

Online advertising uses a descendant: the generalized second-price (GSP) auction for several ad slots. Each advertiser pays the bid of the one just below them. Unlike the single-item Vickrey auction, GSP is not truthful. Advertisers shade their bids, though the system still has good equilibria. The VCG mechanism, covered in mechanism design, restores truthfulness for multiple items by charging each winner the harm they cause to others.

A simulation of strategies in repeated auctions

To see how much you would lose from shading in a second-price auction, run many auctions in which your rivals bid truthfully and you apply different shading factors.

Python Profit as a function of shading

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 profit peaks at factor 1.0: both underbidding and overbidding cost you money on average.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. In a Vickrey auction, what does the winner pay?

    The winner pays the second-highest bid, so the price does not depend on the winner's own bid.
  2. 2. Why does underbidding not help in a second-price auction?

    Your bid affects whether you win, not the price you pay, so shading down only risks lost profit.
  3. 3. Which is a weakness of the Vickrey auction?

    Colluding bidders and low second-highest bids are known drawbacks, even though truth-telling is dominant.
  4. 4. How does the generalized second-price (GSP) auction for ads differ from the Vickrey auction?

    With several slots and position-based clicks, bidding the true value is not dominant in GSP.

Summary

  • Vickrey: highest bid wins, pays the second-highest bid.
  • Truthful bidding is weakly dominant, whatever the others bid.
  • A brute-force check confirms no profitable deviation; the same check fails for first-price auctions.
  • There are other equilibria, collusion risk and low-revenue risk.
  • Descendants (GSP for ads, VCG for general settings) build on the idea of charging the externality.

Next: Revenue Equivalence and Optimal Auctions