Professional Documents
Culture Documents
Auctions
Auctions
2
To clarify the difference between first and
second-price auctions, suppose that we
have three bidders, 1, 2, and 3, bidder
1 submits £100, bidder 2 £50, and bidder
3 £25.
• In a first-price auction bidder 1 gets
the object and pays £100.
• In a second-price auction bidder 1 also
gets the object but pays £50.
Let’s now consider a simple model for
sealed-bid auctions. We have
• a seller who owns an object that for
simplicity is worth nothing to her and
hopes to raises money by selling it at
auction;
• two potential buyers, bidder 1 and bid-
der 2. The monetary valuations of the
objects for these bidders are v1 and v2,
respectively.
If bidder i wins the object and pays x, his
payoff is
vi − x.
The bidder who does not win the object
pays zero. The payoff of the seller is x.
3
The valuations of the bidders are private
information:
• each bidder knows only his valuation
and the seller knows neither valuation.
For instance, the object is an ancient arte-
fact for which each buyer has a personal
valuation that is only known to him.
4
Let’s suppose that the valuations of the
bidders are chosen independently and uni-
formly between the lowest possible value
v and the highest possible value v. Then,
the probability that valuation vi of bidder
i is less than y in [v, v] is
y−v
v−v
5
Let’s first consider a second-price auc-
tion. We will now show a classic result:
in a Bayesian Nash equilibrium each bid-
der bids his own valuation, that is,
bi(vi) = vi.
So let’s suppose that Nature has chosen
the valuation v1 for bidder 1. We will first
see that any bid x of bidder 1 such that
x > v1
cannot achieve a higher payoff than bid-
ding v1, whatever the bid b2 of bidder 2
will turn out to be.
1. If b2 > x > v1, bidding x or v1 makes
no difference for bidder 1 since bidder
2 wins the object.
2. If x > b2 > v1, by bidding x bidder 1
wins the object and pays b2, the sec-
ond highest bid, thus getting payoff
v1 − b2 < 0. By bidding v1, the payoff
is equal to zero since bidder 2 wins the
object: bidder 1 is better off bidding
v1.
3. If x > v1 > b2, regardless of whether
bidder 1 bids x or v1, he wins the ob-
ject and pays b2 thus getting a pos-
itive payoff v1 − b2: bidding x or v1
makes no difference. 6
We have so far considered only ‘strict’ in-
equalities. What if the bids are tied, and
thus the second highest bid is the same
as the highest bid? Whatever the ‘tie-
breaking’ rule for allocating the object, by
bidding x > v1 bidder 1 cannot be better
off:
• If x = b2 > v1 and bidder 1 ends up
winning the object by bidding x, he
pays pays b2 and gets a payoff v1 −b2 <
0. By bidding v1 the payoff is equal
to zero since bidder 2 wins the object.
• If x > b2 = v1, by bidding x bidder
1 wins the object and gets a payoff
v1 − b2 = 0. By bidding v1 the payoff
is also equal to zero whether or not
the object is won by bidder 1 since
v1 = b2.
The case of a bid x < v1 is analogous and
is left as an exercise.
14
Remarkably, the revenues of the first-price
and second-price auctions are identical.
This is a celebrated result known as the
Revenue Equivalence Theorem:
When the valuations of the bidders
are identically and independently dis-
tributed, first-price and second-price
auctions (and more) yield the same
revenue.
How about the ascending auction? It is
not difficult to see that an ascending auc-
tion is essentially equivalent to a second-
price auction.
17
This observation has some important con-
sequences on bidding behaviour. If a bid-
der wins the auction, he must lower his
ex-ante expected valuation of the object:
• a bidder wins when the other bidders
value the object less according to their
own information;
• thus winning the auction reveals ‘neg-
ative’ information to the winner who
should lower his valuation.
This phenomenon is called the winner’s
curse. For instance, winning the drilling
rights for an oil field reveals that the other
bidders’ geological surveys probably found
less oil than winner’s survey.
18
Auction design has been a widely studied
topic, theoretically and experimentally. Spec-
trum auctions, where governments sell the
rights to transmit telecommunication sig-
nals have received a great deal of atten-
tion from auction designers, with mixed
success. Two objectives are prominent in
auction design:
• Efficiency: assigning the object to the
bidder who values it most.
• Revenue generation: maximising the
seller’s revenue.
In addition to the bidding behaviour in dif-
ferent auctions, two considerations, which
we have not addressed directly, play a role
in auction design:
• Encouraging the entry of bidders.
• Preventing collusion among bidders.
Both considerations are central for effi-
ciency and revenue generation. We will
see the effects of an increase in the num-
ber of bidders in an exercise. Collusion
can be a problem when many items are
auctioned simultaneously with a small num-
ber of bidders: bidders can ‘divide up the
items’ cooperatively and bid less aggres-
sively on each item. Ascending auctions
where the bidding process is observed are
especially vulnerable since early ‘visible’
bids can be used to communicate the bid-
ders’ intentions.
19
Many auctions have a reserve price, that
is a minimum bid and payment necessary
to acquire the object. Of course, when
the object has value for the seller it is
reasonable to have a reserve price that is
at least as high as her valuation.
• Can a reserve price that is above the
seller’s valuation increase the revenue
of the seller?
Let’s return to our example where the
seller’s valuation for the object is equal
to zero and two bidders have indepen-
dent valuations uniformly distributed on
[0, 900].
• Would this seller generate a higher ex-
pected revenue by setting a positive
reserve price?
Recall that the seller’s expected revenue
for all the auctions that we considered is
equal to 300.
21