Auction Theory

1
Selling a single good
Seller has a single good that she wants to sell how
can she identify the right buyer and the right price?
Why an auction why not just set a price?
Seller may not know what price to set. Might set it too high
and not sell, or too low and have excess demand.
Potential buyers know what theyd pay, but arent telling.
Auction serves as a mechanism for price discovery, and
also as a way to induce competition among buyers.
Today: different ways to run an auction, the strategic
response of buyers to rules, and the Revenue
Equivalence Theorem. 2
Auction methods

Sealed bidding
Seller asks for single round of bids, high bidder wins and
pays their bid.

Common in selling real estate, timber, commodities, construction

or service contracts (where low bid wins!)
Variant: high bidder wins but pays second highest bid not
common, a little like an eBay proxy auction with sniping.

3
Auction methods

Ascending auction
Price starts low, bidders (or auctioneer) raise price until
only one bidder is left.

Common for art, antiques, cars, charity auctions, real estate,

company take-overs, job market (if youre lucky!).
Variant: price starts high and slowly drops until someone buys
used in Dutch flower auctions, clearance sales.

4
Canonical model
Potential buyers
Two bidders with values v1 and v2
Values v1,v2 drawn from uniform distribution on [0,100].
Each buyers value is his own private information.

Seller gets to set the auction rules.

Note: same ideas would apply with N bidders or with values

drawn from a distribution other than U[0,100].

5
Ascending auction
Price starts at zero, and rises slowly.
Buyers indicate their willingness to continue bidding
(e.g. keep their hand up) or can exit.
Auction ends when just one bidder remains.
Final bidder wins, and pays the price at which the
second remaining bidder dropped out.

How should you bid?

6
Ascending Auction
Optimal strategy (regardless of what opponents do):
keep bidding until the price just equals your value.

Bidder with highest value will win.

Winner will pay second highest value.

Example with three bidders

Suppose values are 25, 33, and 75.
First exits at 25, second at 33 and auction ends.

7
Ascending auction revenue
What revenue can we expect from the ascending
auction? (Assume two bidders with values U[0,100]).
Revenue will be equal to the second highest value.
Expected revenue is equal to the expected second highest value.

If we repeatedly draw two values from U[0,100],

On average, the highest draw will be 66 and 2/3.
On average, the second highest will be 33 and 1/3.

So the average (or expected) revenue is 33 and 1/3.

Question: how would this be different if there were three bidders,
or if values were distributed U[0,200], or U[100,200]?
8
Aside: Order Statistics
One draw

0 1/2 1

Two draws

0 1/3 2/3 1

Three draws

0 1/4 2/4 3/4 1

9
Competition
How is the value created in the auction divided
between the seller and the winning bidder?
Two bidders with values U[0,100].
Expected value of winner: 66.6
Expected price: 33.3
If seller has zero value, equal split of price, consumer surplus.
With three bidders
Expected value of winner: 75 and Expected price: 50
With more demand, expected value creation is higher and
seller is able to capture a higher price (and p/value?).
With 10 bidders, exp. value of winner is 10/11, exp. price is 9/11 of 100.
10
Second price auction
Bidders submit sealed bids.
Seller opens the bids.
Bidder who submitted the highest bid wins.
Winner pays the second highest bid.

How should you bid?

11
Second price auction
Claim. An optimal strategy in the second price auction
(regardless of what opponents do) is to bid your value.
Proof
Compare bidding v to bidding some b>v.
Only way the higher bid matters is if opponent bid is above v but less
than b. Then you lose if you bid v and win if you bid b, but you pay
above your value. So better to have bid v.

Compare bidding v to bidding some b<v.

Only way the lower bid matters is if opponent bid is above b but less
than v. Then you lose if you bid b and win if you bid v. So bidding v is
better you win and pay less than your value.

12
Second price auction

Bid true value v

0 If opponent bid is here,

v If opponent bid is here, 100
win and make money lose the auction

Bid b>v

0 If opponent bid is here, v If bid is b If opponent 100

win and make money here, win bid is here,
but lose lose auction
money!
13
Second price auction

Bid value v

0 If opponent bid is here,

v If opponent bid is here, 100
win and make money lose the auction

Bid b<v

0 If opponent bid b If bid is v If opponent bid is here, 100

is here, win and here, lose lose auction
make money auction.
14
Second price auction
In equilibrium, everyone bids their value.
Bidder with highest value wins.
Pays an amount equal to second highest value.

Exactly the same as the ascending auction!

Same winner, same price/revenue.

Example: three bidders with values 25, 33, 75.

Winner has value 75, pays 33, as in ascending auction.

15
Vickrey auction
Second price auction is an example of the more general
Vickrey auction that can be used to sell multiple units
or goods in a way that makes truthful bidding optimal.

Everyone submits their value(s)

Seller allocates to maximize value
Winners pay an amount equal to the value their goods
would have in their next best alternative use.

Think about how the 2nd price auction does this!

16
Sealed bidding
Bidders submit sealed bids.
Seller opens the bids and
Bidder who submitted highest bid wins.
Winner pays his own bid.

Now what is the optimal strategy?

17
Sealed bidding, cont.
Best to submit a bid less than your true value.
How much less?
Submitting a higher bid
increases the chance you will win
increases the amount youll pay if you do win
Optimal bid depends on what you think the others
will bid (unlike in the second-price auction!).
We need to consider an equilibrium analysis.

18
Nash equilibrium
Defn: A set of bidding strategies is a (Bayesian) Nash
equilibrium if each bidders strategy maximizes his
expected payoff given the strategies of the others.

In an auction game, a strategy says how much you should

bid for every possible value.
Bidders do not know their opponents values, i.e. there is
incomplete information.
So each bidders equilibrium strategy must maximize her
expected payoff accounting for the uncertainty about
opponent values.
19
Solving the sealed bid eqm
Suppose 2 uses the strategy: bid b2=*v2.
Bidder 1 understands 2s strategy (the eqm assumption), but
doesnt know 2s bid because he doesnt know v2.
If Bidder 1 bids b:

0 If opponent bid is here,

b If opponent bid is here, 100
win and pay b lose the auction

Event Probability Net utility

b2 < b Pr(v2<b) = Pr(v2<b/) v-b
= b/(100)
b2 > b 1-b/(100) 0
20
Solving the sealed bid eqm
Event Probability Net utility
b2 < b Pr(v2<b) = Pr(v2<b/) v1-b
= b/(100)
b2 > b 1-b/(100) 0

Expected utility = b/(100)*(v1-b) + 0

So bidder 1s bidding problem is
maxb (b/100) *(v1 - b)
where (b/100) is prob. of winning and v1 b is profit if he
does win. The goal is to maximize expected profit. 21
Solving for equilibrium
Problem: maxb (b/100) * (v1 - b)

First order condition

(1/)(v1-b) (b/) = 0

Cancelling out (1/) and re-arranging

v1-b = b
b =(1/2)v1

So if bidder 2 uses the strategy b2 =*v2, bidder 1s best

response is to use the strategy b1 =(1/2)*v1
22
Sealed bid equilibrium
There is an equilibrium of the sealed bid auction in which
b1=(1/2)v1, b2=(1/2)v2 . Both bidders bid half their value.

Bidder with the highest value wins in eqm.

What is the revenue, on average?

Values are distributed U[0,100], so bids are distributed U[0,50].
High value, on average is 66. High bid, on average, is 33.

Same as the ascending and second price!

Same for each realization of bidder values, or just on average?
Example: suppose bidder values are 75 and 50.
23
Wait how did that work?
We compared two sealed bid auctions
High bidder wins and pays own bid
High bidder wins and pays next-highest bid.
Result: on average the seller gets the same revenue.
But why? Dont you get more by charging the winner his
own bid, rather than a lower amount?

When you change the game, you also change behavior!

The equilibrium strategy goes from b=v to b=v/2.
What is surprising is that the change in behavior just
exactly offsets the change in the rules.
24
The Equivalence Theorem
Weve seen that several standard auctions share the following
properties if bidders play according to Nash equilibrium:
The high value bidder wins.
The expected auction price equals the expected value of the
second highest bidder.

The result that efficient auctions (where the high value

bidder wins) are revenue equivalent (yield the same price on
average) is quite general (see RET handout).
It actually has been the source of two Nobel Prizes: William Vickrey in
1996 for the first version of the result, and Roger Myerson in 2007 for
using a generalization of it to describe optimal auctions.

25
Other forms of auction
Descending price (Dutch) auction
Price starts high, drops continuously until someone buys.
Strategically equivalent to a first price auction.

All pay auction

Bidders submit bids, high bidder wins, and everyone has to pay their bid.
Not commonly seen, but sometimes used as a stylized model of political
lobbying donations, or contests where money is spent to win.

Penny auctions
Ascending auction, with the twist that you pay for every bid along the
way. Popular on-line, e.g. Swoopo.

26
Elements of Auction Design
Sellers have a variety of ways to modify the basic auction
forms and try to increase their revenue, or achieve other
objectives.

Attracting bidders to participate in the auction.

Setting a reserve price (or minimum bid).
Subsidizing weak bidders to create competition.

In many settings, such as procurement, there is also the

challenge of making sure bidders compete (i.e. dont collude).
The Klemperer paper on the reading list emphasizes this.
27
Are bidders really strategic?
Game theory models weve looked at assume that
bidders understand the environment and behave
strategically. A good assumption?

No simple answer to this question

Lots of evidence against the assumption when looking at bidders
in novel environments with little feedback.
But potentially a reasonable assumption for experienced and
sophisticated bidders in familiar environments.

Consider a few examples

28
29
Swoopo Auction
Auction rules
Price starts at $0, and clock starts to run down.
Placing a bid costs $0.50.
A bid re-starts the clock, price increases by $0.10.
If clock expires, auction ends. Standing high
bidder is the winner and pays the current price.

Swoopo sells common items (electronics,

sometimes even cash).
30
Excess bidding on Swoopo?
Ned Augenblicks Stanford dissertation
Data from 650 auctions of $50 bills: average revenue: $104!
More general data from over 100,000 Swoopo auctions in 2008:
revenue was 52% above Amazon price; Swoopo making millions.
Maybe bidders in auctions just get carried away?

However, in March 2011, Swoopo filed for bankruptcy.

Maybe the bidders caught on? Google trends

31
32
Excess bidding on eBay?
Lee and Malmendier (2011)
Auctions for Cashflow 101 in 2004.
Available from two eBay sellers at posted price: $129.95.
Price went higher in 42% of the auctions!

But . Einav, Kuchler, Levin & Sundaresan (2012)

Hundreds of thousands of cases where an eBay seller sold a
product by auction while also selling it at a posted price.
Auction prices exceed posted prices fairly often in 2003, but by
2009 in relatively few auctions, and usually by just a few dollars.

33
 eBay Auction Prices (2003)
0.3

0.25

0.2

0.15

0.1

0.05

Auction Price / Items Average Posted Sale Price

34
Auction prices often exceeded posted prices in 2003.
 eBay Auction Prices (2009)
0.3

0.25

0.2

0.15

0.1

0.05

Auction price / Items Average Posted Sale Price

35
Auction prices were usually well below posted prices in 2009.
Summary
Auctions are a useful way to sell when
The appropriate price is unknown
The item is scarce relative to buyer demand
There are many ways to run an auction
Sealed bidding, ascending auctions and more unusual variants.
Strategic bidders adapt behavior to compensate for rule changes.
Surprisingly, these changes can result in many standard auctions
being revenue equivalent an important idea in auction theory.
Game theory analysis of auctions assumes that bidders
are sophisticated and strategic. The evidence is mixed.

36