Auctions Economics 383 - Auction Theory Instructor: Songzi Du Simon Fraser University October 1, 2015 ECON 383 (SFU) Auctions October 1, 2015 1 / 26 Auctions Mechanisms of transaction: bargaining, posted price, auctions Auction: take bids, allocate resource, and collect payments. Babylonian wife auction (500 BC) Auction of the Roman Empire by the Praetorian Guard (who had killed Emperor Pertinax in 193 AD). The winning bidder Didius Julianus was crowned Emperor; beheaded 9 weeks later (winner’s curse). Google AdWords auction (revenue of USD$28 billion in 2010), eBay, Amazon Financial auctions (treasury bills, settlement of credit default swap, stock exchange) ECON 383 (SFU) Auctions October 1, 2015 2 / 26 Second price auction A single, indivisible good. Second price auction: 1 2 Every bidder submits a bid, simultaneously (seal bid). The highest bidder gets the object and pays the second highest bid; everyone else does not pay. Also known as Vickrey auction. Proxy bidding in eBay: a computer program that automatically and minimally increases your bid (up to your pre-specified maximum amount) to ensure that you are the top bidder. ECON 383 (SFU) Auctions October 1, 2015 3 / 26 Second price auction A single, indivisible good. Second price auction: 1 2 Every bidder submits a bid, simultaneously (seal bid). The highest bidder gets the object and pays the second highest bid; everyone else does not pay. Also known as Vickrey auction. Proxy bidding in eBay: a computer program that automatically and minimally increases your bid (up to your pre-specified maximum amount) to ensure that you are the top bidder. Bidder i has a value vi for the good (his private information), payoff of vi − Pi if he gets it, 0 if not. ECON 383 (SFU) Auctions October 1, 2015 3 / 26 Ascending bid auction Also known as English auction. The auction is carried out interactively in real time. The auctioneer gradually raises the price, starting from some reserve price (e.g., zero), bidders drop out until finally only one bidder remains, and that bidder wins the object at this final price. Variants of ascending bid auction: bidders shout out prices, or submit them electronically. ECON 383 (SFU) Auctions October 1, 2015 4 / 26 Ascending bid auction ECON 383 (SFU) Auctions October 1, 2015 5 / 26 Strategy in Second Price Auction Strategy: a function si (vi ) that maps values to bids. ECON 383 (SFU) Auctions October 1, 2015 6 / 26 Strategy in Second Price Auction Strategy: a function si (vi ) that maps values to bids. n bidders Payoff function: vi − max(b1 , . . . , bi−1 , bi+1 , . . . , bn ) if bi > max(b1 , . . . , bi−1 , bi+1 , . . . , bn ) Ui (vi , b1 , b2 , . . . , bn ) = 0 otherwise ECON 383 (SFU) Auctions October 1, 2015 6 / 26 Strategy in Second Price Auction Strategy: a function si (vi ) that maps values to bids. n bidders Payoff function: vi − max(b1 , . . . , bi−1 , bi+1 , . . . , bn ) if bi > max(b1 , . . . , bi−1 , bi+1 , . . . , bn ) Ui (vi , b1 , b2 , . . . , bn ) = 0 otherwise Dominant strategy si (vi ) satisfies: for every (s1 ( · ), . . . , si−1 ( · ), si+1 ( · ), . . . , sn ( · )) and every (v1 , v2 , . . . , vn ), Ui (vi , s1 (v1 ), . . . , si−1 (vi−1 ), si (vi ), si+1 (vi+1 ), . . . , sn (vn )) ≥Ui (vi , s1 (v1 ), . . . , si−1 (vi−1 ), bi , si+1 (vi+1 ), . . . , sn (vn )) for every bi ∈ R. ECON 383 (SFU) Auctions October 1, 2015 6 / 26 Why second price? Why not third price? Third price auction: the highest bidder gets the good and pays the third highest bid; everyone else do not pay. Is truthful bidding the dominant strategy? ECON 383 (SFU) Auctions October 1, 2015 7 / 26 Auction of two goods Auction of two indivisible, identical goods. Each bidder i wants only one good, has a value vi if he gets a good. As before, each bidder submits a bid. Third-price auction: the top two bidders each gets a good, and each pays the third highest bid; the rest do not pay. Is truthful bidding the dominant strategy? ECON 383 (SFU) Auctions October 1, 2015 8 / 26 Facts about uniform distribution Suppose n bidders, with values vi randomly and independently drawn from the uniform distribution on [0, 1]: P(vi ≤ x) = x, P(v1 ≤ x1 , v2 ≤ x2 , v3 ≤ x3 ) = x1 · x2 · x3 , for x’s between 0 and 1. E[max(v1 , v2 , . . . , vn )] = n−1 n , E[max2 (v1 , v2 , . . . , vn )] = , n+1 n+1 n−2 1 , . . . , E[min(v1 , v2 , . . . , vn )] = , n+1 n+1 where max2 means second highest, max3 third highest, etc. E[max3 (v1 , v2 , . . . , vn )] = ECON 383 (SFU) Auctions October 1, 2015 9 / 26 Reserve price in second price auction Reserve price (r ): the minimum bid that is considered in the (second price) auction, announced before the auction. 1 2 The good is sold to the highest bidder if the highest bid is equal or above r ; otherwise, the good is not sold. The winning bidder (if any) pays the maximum of the second-place bid and the reserve price. ECON 383 (SFU) Auctions October 1, 2015 10 / 26 Reserve price in second price auction Reserve price (r ): the minimum bid that is considered in the (second price) auction, announced before the auction. 1 2 The good is sold to the highest bidder if the highest bid is equal or above r ; otherwise, the good is not sold. The winning bidder (if any) pays the maximum of the second-place bid and the reserve price. Why set reserve price? What is the role of reserve price in revenue? Suppose the seller has no value for the (single, indivisible) good that he is auctioning. There are n bidders, with values randomly and independently drawn from the uniform distribution on [0, 1]. What’s the optimal reserve price when n = 1? n = 2? What happens when the seller uses a posted price? ECON 383 (SFU) Auctions October 1, 2015 10 / 26 Second price auction with reserve price r Let Rev(r ) be the seller’s revenue given a reserve price r ∈ [0, 1]. If there is only n = 1 bidder: Rev(r ) = (1 − r ) · r . Rev0 (r ) = 1 − 2r Optimal reserve price r = 1/2 (from solving Rev0 (r ) = 0). If there are n = 2 bidders: 1−r Rev(r ) = 2r (1 − r ) · r + (1 − r ) · r + 3 2 Rev0 (r ) = 2r (1 − 2r ) Optimal reserve price r = 1/2. ECON 383 (SFU) Auctions October 1, 2015 11 / 26 Second price auction with reserve price r An observation: the seller’s revenue from the optimal reserve price and 1 bidder (1/4) is less than his revenue from zero reserve price and 2 bidders (1/3). This is a general theorem (Bulow and Klemperer). Setting the optimal reserve price is less profitable than simply attracting an additional bidder. ECON 383 (SFU) Auctions October 1, 2015 12 / 26 First price auction A single, indivisible good. First price auction: 1 2 Every bidder submits a bid, simultaneously (seal bid). The highest bidder gets the object and pays his own bid; everyone else does not pay. Bidder i has a value vi for the good (his private information), payoff of vi − Pi if he gets it, 0 if not. ECON 383 (SFU) Auctions October 1, 2015 13 / 26 Descending bid auction Also known as Dutch auction. The auction is carried out interactively in real time. The auctioneer gradually lowers the price from some high initial value until the first moment when some bidder accepts and pays the current price. Flowers have long been sold in the Netherlands using this procedure. ECON 383 (SFU) Auctions October 1, 2015 14 / 26 A Model of First Price Auction n bidders (n ≥ 2) Each bidder i (1 ≤ i ≤ n) has a private value vi for the good. 0 ≤ vi ≤ 1. The distribution of vi is the uniform distribution on [0, 1]. Identical and independent distribution for every bidder. Bidding strategy is a function si (vi ) that maps values to bids. vi is bidder i’s type. ECON 383 (SFU) Auctions October 1, 2015 15 / 26 Strategy in First Price Auction Payoff function: vi − bi if bi > max(b1 , . . . , bi−1 , bi+1 , . . . , bn ) Ui (vi , b1 , b2 , . . . , bn ) = 0 otherwise ECON 383 (SFU) Auctions October 1, 2015 16 / 26 Strategy in First Price Auction Payoff function: vi − bi if bi > max(b1 , . . . , bi−1 , bi+1 , . . . , bn ) Ui (vi , b1 , b2 , . . . , bn ) = 0 otherwise Bayesian Nash Equilibrium: strategy profile (s1 (v1 ), s2 (v2 ), . . . , sn (vn )) such that for every bidder i and every vi , E[Ui (vi , s1 (v1 ), . . . , si−1 (vi−1 ), si (vi ), si+1 (vi+1 ), . . . , sn (vn ))] ≥E[Ui (vi , s1 (v1 ), . . . , si−1 (vi−1 ), bi , si+1 (vi+1 ), . . . , sn (vn ))] for every bi ∈ R. ECON 383 (SFU) Auctions October 1, 2015 16 / 26 Solving for equilibrium (first price auction) We focus on symmetric equilibrium: s1 = s2 = · · · = sn = s. What is bidder i’s profit from bidding s(vi ), given that others also bid according to s? Ui (vi ) = (vi )n−1 · (vi − b(vi )) Bidder i of type vi maximizes (by bidding s(x)): max x n−1 (vi − s(x)) 0≤x≤1 FOC: (n − 1)x ECON 383 (SFU) n−2 vi − (n − 1)x n−2 Auctions s(x) − x s (x) n−1 0 =0 x=vi October 1, 2015 17 / 26 Solving for equilibrium (first price auction) FOC: (n − 1)(vi )n−2 vi − (n − 1)(vi )n−2 s(vi ) − (vi )n−1 s 0 (vi ) = 0 Rearrange: s(vi ) = vi − vi 0 s (vi ) n−1 Guess: s(vi ) = A(vi )k ⇒ A(vi )k = vi − Clearly k = 1. Then A = 1 − Equilibrium bidding strategy: A n−1 , s(vi ) = vi Ak(vi )k−1 . n−1 i.e., A = n−1 n . n−1 vi n s(vi ) < vi . This is called bid shading. ECON 383 (SFU) Auctions October 1, 2015 18 / 26 All Pay Auction All pay auction: the highest bidder gets the good, everyone pays his/her bid. Everything else as before (a single good, simultaneous bids, private values, etc.) Example (bribery): in 2008, Governor Rod Blagojevich of Illinois tried to sell Barack Obama’s senate seat to the highest bidder. Other examples: war of attrition, political campaign, Olympic game, etc. ECON 383 (SFU) Auctions October 1, 2015 19 / 26 Solving for equilibrium (all pay auction) We focus on symmetric equilibrium: s1 = s2 = · · · = sn = s. What is bidder i’s profit from bidding s(vi ), given that others also bid according to s? Ui (vi ) = (vi )n−1 vi − s(vi ) Bidder i of type vi maximizes (by bidding s(x)): max x n−1 vi − s(x) 0≤x≤1 FOC: (n − 1)x ECON 383 (SFU) n−2 vi − s (x) Auctions 0 =0 x=vi October 1, 2015 20 / 26 Solving for equilibrium (all pay auction) FOC: (n − 1)(vi )n−1 = s 0 (vi ). Guess: s(vi ) = A(vi )k ⇒ (n − 1)(vi )n−1 = Ak(vi )k−1 . k = n and n − 1 = Ak, i.e., A = n−1 n . Equilibrium bidding strategy: s(vi ) = ECON 383 (SFU) n−1 (vi )n n Auctions October 1, 2015 21 / 26 Average Price Auction Average Price Auction: the highest bidder gets the good, pays the average of all bids; everyone else does not pay. Everything else as before (a single good, simultaneous bids, private values, etc.) ECON 383 (SFU) Auctions October 1, 2015 22 / 26 Solving for equilibrium (average price auction, n = 2) Suppose n = 2. We focus on symmetric equilibrium: s1 = s2 = s. What is bidder i’s profit from bidding s(vi ), given that the other also bids according to s? 1 Ui (vi ) = vi · vi − (s(vi ) + E[s(vj ) | vj ≤ vi ]) , j 6= i 2 Bidder i of type vi maximizes (by bidding s(x)): 1 max x · vi − · (s(x) + E[s(vj ) | vj ≤ x]) 0≤x≤1 2 ECON 383 (SFU) Auctions October 1, 2015 23 / 26 Solving for equilibrium (average price auction, n = 2) Guess: s(vi ) = Avi Ax 1 ⇒ max x · vi − · Ax + 0≤x≤1 2 2 FOC: 3 3 = vi − Avi = 0 vi − Ax 2 2 x=vi A = 23 . Equilibrium bidding strategy: 2 s(vi ) = vi 3 ECON 383 (SFU) Auctions October 1, 2015 24 / 26 Revelation Principle Bidder i’s equilibrium strategy si (vi ) is his “agent.” Bidder i tells the “agent” his true value, the “agent” bids on his behalf. No incentive to deviate from the strategy si is equivalent to an incentive to report the true value to the “agent.” This is known as the revelation principle. Bidder i is not necessarily bidding truthfully with si (vi ) (i.e., si (vi ) needs not be vi ). ECON 383 (SFU) Auctions October 1, 2015 25 / 26 Comparing payments from auctions First price auction: sfp (vi ) = n−1 n vi . Second pay auction: ssp (vi ) = vi . All pay auction: sall (vi ) = ECON 383 (SFU) n−1 n n (vi ) . Auctions October 1, 2015 26 / 26 Comparing payments from auctions First price auction: sfp (vi ) = n−1 n vi . Second pay auction: ssp (vi ) = vi . All pay auction: sall (vi ) = n−1 n n (vi ) . In all of these auctions, the expected payment of a bidder i with value n vi is n−1 n (vi ) . Same payment, i.e., revenue equivalence! Bidders respond strategically to the change in auction rule, un-do the intended change. ECON 383 (SFU) Auctions October 1, 2015 26 / 26