Published on

Notes on Roughgarden AGT Lecture - 2

Lecture link

Simple single-item sealed-bid first/second-price auctions

  1. There is a single item for sale.

  2. Each bidder ii has a private valuation viv_i for the item.

    • The valuation is the maximum amount of money that ii is willing to pay for the item.
  3. Each bidder has a quasilinear utility function:

    • If ii wins, ui=vipiu_i = v_i - p_i;pip_i is the price that bidder ii pays.
    • If ii does not win, ui=0u_i = 0.
  4. The bids are submitted privately to the auctioneer, who then determines the winner and the price.

  5. The auction can be first-price or second-price.

In first-price auctions, one does not want to bids its true valuation, as it always results in a zero utility. Instead, the bid amount depends on others' bids. On the other hand, in the second-price auction, the (weakly) dominant strategy for a players is to simply bid its true valuation.

Claim 1. In a seal-bid second-price auction, bidding the valuation is a dominant strategy.

Proof sketch. Fix a player ii and bib_{-i}; let bb^* be the maximum of bib_{-i}. If vi<bv_i < b^*, then bidding viv_i results in a zero utility (as ii loses). Nevertheless, any other bid that results in ii winning implies a payment of b>vib^* > v_i, and thus a negative utility.

If vibv_i \geq b^*, then bidding viv_i results in a utility of vibv_i - b^* which is at least as good as not winning.

One can also observe that truth telling in a second-price auction never gives a negative utility, regardless of the bids of others.

Overall, second-price auctions have the following desirable properties:

  1. dominant-strategy incentive-compatible(DSIC) - telling the truth is a dominant strategy for all players (and the utility is always non-negative).

  2. Can be implemented efficiently.

  3. If bidders are truthful, then the auction maximizes the valuation of the outcome (social surplus).

Remark

Note that in a standard second-price auction, the second-highest bid equals to the highest bid if there is a tie. However, if we change the rule to paying the second-highest bid that is strictly less than the highest bid (for now, lets assume there are always at least two distinct bids), then bidding own valuation is no longer a dominant strategy, as overbidding sometimes helps.

For instance, fix ii, and suppose viv_i lies between bb^* and bb', where bb' is the second-highest bid in bib_{-i} that is strictly less than bb^*. Then, bidding bb^* gives a non-zero expected utility for ii (randomness comes from tie breaking).

See page 55 of the lecture note for an example of ads auctions.

Step 1: Assume, without justification, that bidders bid truthfully. Then, how should we assign bidders to slots so that the above properties (2) and (3) hold? Step 2: Given our answer to Step 1, how should we set selling prices so that the above property (1) holds?