Online Auctions
Online Auctions
1 Introduction
• How to run an auction on the Internet?
− Bidders logged at the same time and follow in real-time the auction.
− Bidder with $100 wins, pays only $1: a disaster for the seller.
• eBay auction format: a mixture of dynamic auction and second-price auction.
− The price can last several days or weeks until the end time.
− The winner is the highest bidder and pays the price displayed when the auction
ends.
2 eBay: Proxy Bidding
• When placing a bid, eBay asks the maximum bid, and eBay will bid on the
bidder’s behalf, using increments that work like ticking prices:
− The previous highest price ($10.00) is in fact now the 2nd highest bid
− The new price is then the 2nd highest bid (+ the $0.50 increment).
− eBay will automatically compute the outcome of the “bidding war” between me
and the other bidder.
− Up to $30 I continue to bid. At that price I stop (i.e., eBay stops on my behalf).
− The last price is $31, i.e., my final bid of $30 + the increment of $1.
• Since everybody uses the proxy bidding, the auction is solved “immediately”:
− $31 = the lowest bid that the new bidder (with the $40 bid) needs to win.
[...] after an auction has closed we receive emails from bidders who claim
they were attempting to place a bid and were unable to get into eBay.
There is nothing we can do to help bidders who were “locked out” while
trying to place a “last minute” bid. (Axis Mundi, a seller in eBay)
− If I bid early, then the “incremental bidder” (inexperienced) will overbid and I
may end up with a high price.
− If I bid late, the bid war doesn’t last long ⇒ lower final price
− Last minute bidding: The incremental bidder doesn’t have time to respond. Can
win the auction at the incremental bidder’s low initial bid.
• Ockenfels and Roth (2006) provide a model of eBay auctions and analyze strategic
reasons for late bidding.
• The “current price” (or “high bid”) in an auction with at least two bidders is
equal to the minimum increment over the second highest bid.
− A bidder has time to react before the end of the auction to another bidder’s bid
at time t′ < 1.
− At t = 1, everyone knows the bid history prior to t and has time to make exactly
one more bid. The probability that a bid is successfully transmitted is p < 1.
• Private value: Each bidder i has value vi, distributed according to some distribu-
tion F .
− Suppose that bidder i bids m at t = 0 and then does not bid as long as he
remains the high bidder, but bids B(> vj + s) whenever he is not the high
bidder. Bidder j’s best response is not to bid at t < 1 and to bid vj at t = 1.
The payoff to j from this strategy is p(vj − m − s) > 0.
− Suppose that bidder i’s strategy is not to bid at any time. Then, the strategy
for bidder j to bid at t = 1 will give payoff p(vj − m) < vj − m, where the RHS
is the payoff from bidding vj at t = 0.
4.1 Best response to incremental bidding
• Let n = 2 and both bidders have the same value v, where v > m + 2s.
− Each bidder i does not bid at any time t < 1 and bids v at t = 1.
− If bidder j deviates, i.e., places a bid at t < 1, then bidder i bids v at t′ ∈ (t, 1).
• The object for sale can be “Fake” with probability pF or “Genuine” with proba-
bility pG = 1 − pF .
− If the object is genuine and bidder I places a bid at some t < 1, then bidder U
can infer that the object is genuine and outbid at some time t′ ∈ (t, 1).
− If the object is genuine and bidder I does not bid before t = 1, then bidder U
wouldn’t bid whenever pG(1 − p)(H − m) − pF m < 0, in which case bidder I
indeed places a bid at t = 1.
5 Amazon versus eBay
• However. . .
− Amazon: If there’s a bid in the last 10 minutes, the deadline is extended for an
additional 10 minutes.
• Roth and Ockenfels (2002) and Ockenfels and Roth (2006) compare those two.
• Hypothesis about late bidding
Non-strategic hypothesis
– Delaying
– Bidders unaware of proxy bidding No difference
– Valuation increases (endowment effect) between Amazon
– Bidders don’t like bids hanging & eBay
5.1 Data analysis
− Computer: private values, retail prices are easily available, bids convey little
information.
− Feedback# reflects the number of transactions a bidder participate in, and thus
the bidders’ experiences.
− Sniping is less likely to occur when valuations are private (as opposed to inter-
dependent or common values).
• The use of auctions on eBay is declining, while the use of posted price increases.
References
Ockenfels, Axel and Alvin E. Roth (2006), “Last-Minute Bidding in Second Price Internet Auctions:
Theory and Evidence Concerning Different Rules for Ending an Action,” Games and Economic
Behavior, 55, 297-320.
Roth, Alvin E. and Axel Ockenfels (2002), “Last-Minute Bidding and the rule for Ending Second-
Price Auctions: Evidence from eBay and Amazon Auctions on the Internet,” American Economic
Review, 92, 1093-1103.