Lecture 1: Spectrum Auctions (23 February)
Lecture 2: Electricity Auctions (1 March)
A revised handout is available on the course web page.
General reference: Paul Milgrom (1989), Auctions and Bidding: A Primer, The Journal of Economic Perspectives 3, pp. 3-22.
References
• Paul Milgrom (2000), Putting Auction Theory to Work: The Simultaneous Ascending Auction, Journal of Political Economy 108, pp. 245-72.
• Ken Binmore and Paul Klemperer (2002), The Biggest Auction Ever: The Sale of the British 3G Telecom Licenses, Economic Journal 112, pp. C74-C96.
• Tilman Börgers and Christian Dustmann, Strange Bids: Bidding Behavior in the United Kingdom’s Third Generation Spectrum Auction, available at: http://www.ucl.ac.uk/~uctpa01/strange2.pdf .
Outline of the Lecture
1.Introduction
2.An Overview of 3G Licensing in Europe
3.Why Auctions?
4.The Simultaneous Ascending Auction
5.The UK Experience
6.Conclusion
1. Introduction
• Mobile telephone networks use radio spectrum.• Spectrum users need permissions (licenses). • Licenses are issued by a government agency (in
the UK: formerly Radiocommunications Agency, now Ofcom).
• Around the year 2000, licenses for the next (3rd) generation of mobile telephone networks needed to be issued.
• International agreement had specified spectrum bands and technology (UMTS).
The Objectives
1. Create sustainable competition in the third generation mobile telephone market.
2. Allocate the licenses to the “right” license holders.
3. Raise government revenue.
The Constraints
1. The available spectrum: 2x60 MHz of paired spectrum and 25 MHz of unpaired spectrum.
2. Technological constraints (minimum spectrum: 2x10 MHz paired spectrum).
Method I Method II Method III
Mechanism by which the Number and Size of Licenses is determined
Bureaucratic Bureaucratic Auction
Mechanism by which the License Holders are chosen
Bureaucratic Auction Auction
The Methods
• Economic theory makes predictions regarding these methods.
• Before the auction: Recommendations based on theory were made.
• A variety of methods were tried out.• After the auction: Check whether predictions came
true.• Our focus will be on the UK.• But first we give an overview of the European
picture.
2. An Overview of 3G Licensing in Europe
Method I (Beauty Contest)
Finland, Spain, Sweden, Portugal, France, Ireland, Luxembourg
Method II (Auctions With Exogenous Licenses)
UK, Netherlands, Italy, Belgium, Denmark
Method III (Auctions With Endogenous Licenses)
Germany, Austria, Greece
Country No of Licenses
Rev (€ ) per head
Method I Finland
February 1999
4 0
Spain
December 1999
4 15
Sweden
September 2000
4 0
Portugal
September 2000
4 40
France
January 2001
2 168
Ireland
March 2002
3 27
Luxembourg
March 2002
3 2
The Results
Country No of Licenses
Rev (€ ) per head
Method II UK
March 2000
5 642
Netherlands
June 2000
5 372
Italy
September 2000
5 213
Belgium
February 2001
3 44
Denmark
September 2001
4 96
Country No of Licenses
Rev (€ ) per head
Method III Germany
July 2000
6 619
Austria
November 2000
6 103
Greece
July 2001
3 46
3. Why Auctions?
Objective:
• “Put licenses into the hands of those who value them the most.”
Allocate licenses so that the sum of valuations is maximized.
• Promote competition.
How to maximise the sum of license valuations?
Identical licenses:
Give to the bidders with the highest valuations.
Heterogeneous licenses:
More complicated.
Example:
One small license (S), one large license (L).
Solution: A gets S and B gets L.Marginal valuations matter!
Bidder Value for S Value for L
A 9 11
B 4 7
Example (continued):
One small license (S), one large license (S).
Solution: A gets S and B gets L.
Bidder Value for S Value for L
A 9 11
B 4 7
C 5 6
The key problem:• The government doesn’t know the valuations.
The solution:• Auctions.• Competitive bidding reveals values.• Competitive bidding maximizes the sum of
valuations.
4. The Simultaneous Ascending Auction
A very rough classification of auctions:
Single Unit Auctions
Multi-Unit Auctions
Sealed Bid Auctions
Open Bid Auctions
x
Open, Multi-Unit Auctions
• The most prominent open, multi-unit auction is the “simultaneous ascending auction”.
• The “simultaneous ascending auction” is the most popular auction format for spectrum auctions.
• It is often attributed to Milgrom, Wilson, McAfee, …
• Here we study it in the simplest context: Multi-unit supply, but single-unit demand.
Simultaneous, Ascending Auction Rules:
• All licenses are auctioned simultaneously.• Bidding starts at minimum bids.• In each round, the currently leading bidders are
committed to their bid, and can’t change it.• All other bidders have to overbid one currently
leading bid by a minimum increment, or drop out.• Bidding continues until the number of active
bidders = number of licenses.• Bidding on all licenses closes at the same time.
Straightforward Bidding:• Surplus = Value of a license – Minimum bid.• Bid for the license for which surplus is
maximised.• Bid the minimum bid.
Theorem:• Straightforward bidding is rational (a Nash
equilibrium).• Under straightforward bidding, licenses are
allocated so that the sum of valuations is maximised. (Milgrom, 2000)
Example (continued)
• Bid price of L up to 1.• C switches to S• Bid price of L up to 2.• A switches to S, and then switches back and forth.• When price of S reaches 5, C drops out.• A wins S and B wins L.
Bidder Value for S Value for L
A 9 11
B 4 8
C 5 6
Why is the Theorem true?
Straightforward bidding is rational (a Nash equilibrium)…
• At the time at which a bid is placed, you get the highest surplus.
• If prices of other licenses go up, your choice looks even better.
• If your bid is overbid, then you can still switch on other licenses.
Is it always best to bid straightforwardly?
• Not if the others play a strategy of the type “Bid the price of license X up, if bidder Y ever bids for license X”.
• Yes if the others also bid straightforwardly.
This is why the Theorem only says that it is a Nash equilibrium to bid straightforwardly.
Under straightforward bidding, licenses are allocated so that the sum of valuations is maximised…
• Suppose some other license allocation had a higher sum of valuations.
• Then also the difference between sum of valuations and current prices were higher.
• Then some bidder would have to have a higher surplus in the new allocation.
• Why did this bidder not bid for this license?
5. The UK Experience
• I focus on the case of the UK.• There were five licenses: 2 large ones (A and B),
and 3 small ones (C, D and E).• The 4 incumbents were not allowed to bid for A.• There were 13 bidders.• 6 March 2000 - 27 April 2000.• 150 rounds.
The Winners:
Incumbents in red
Licence Company Winning Bid
A TIW £ 4,384.7 m
B Vodafone £ 5,964.0 m
C BT3G £ 4,030.1 m
D One2One £ 4,003.6 m
E Orange £ 4,095,0 m
The Losers:
Company Withdrew in Round
NTL Mobile 150
Telefonica 133
Worldcom 121
One.Tel 100
Epsilon 98
Spectrumco 97
3GUK 95
Crescent 94
Our main question:
• Did bidders reveal consistent marginal valuations?• Under straightforward bidding: bid for large
license if
vL-pL > vS-pS
pL-pS < vL-vS
• Valuation differences should be revealed by bids.
Possible Interpretation of BT’s Behaviour:
• Naïve rule of thumb (minimize £/MHz)• Easy to justify towards shareholders.• Drive up the price that Vodafone has to pay.
Possible Interpretation of TIW’s Behaviour:
• Strategy change after the first phase.• Tried to conceal their true identity in phase 1.
Possible Interpretation of NTL Mobile’s Behaviour:
• Learning in the first phase.• Budget constraints in the last phase.
6. Conclusion
• More research is needed to understand bidders’ motivation structure in license auctions.
• The outcome of the UK auction might have been efficient …
• … but not for the reasons we expected.• Auction designs need to be tested further.
Open Questions
• Do we need to license use of the radio spectrum (“Open Spectrum Movement”)?
• Should licenses specify the use that is made of the spectrum?
• Should we allow trade in licenses?• Should license trade be regulated?• How does trade in licenses affect bidding in
license auctions?• Do customers get a good deal when firms have to
pay high license fees?