Professional Documents
Culture Documents
Cramton-Spectrum-Auction-Design-Slides 1
Cramton-Spectrum-Auction-Design-Slides 1
p Auction Design
g
Peter Cramton*
P f
Professor off Economics,
E i University
U i it off Maryland
M l d
23 April 2009
• Government
G t objective:
bj ti Efficiency
Effi i
– Make best use of scarce spectrum
• Recognizing competition issues in downstream market
Main points
• Enhance substitution
– Product design
– Auction design
• Encourage price discovery
– Dynamic price process to focus valuation
efforts
• Induce truthful bidding
g
– Pricing rule
– Activityy rule
Simultaneous
ascending auction
Simultaneous ascending auction
• Simultaneous
– All lots at the same time
• Ascending
– Can raise bid on any lot
• Stopping rule
– All lots open until no bids on any lot
• Activity rule
– Must be active to maintain eligibility
g y
Simultaneous ascending auction
• Strengths
g
– Simple price discovery process
– Allows arbitrage across substitutes
– Piece together
g desirable p
packages
g
– Reduces winner’s curse
• Weaknesses
– Demand reduction
– Tacit collusion
– Parking
– Exposure
– Hold up
– Limited substitution
– Comple bidding strategies
Complex
Limited substitution: US AWS
90 MH
MHz, 161 rounds,d $14 billi
billion
US AWS band plan: something for everyone
1710 1720 1730 1740 1755
Uplink A B C D E F
Bandwidth 20 MHz 20 MHz 10 MHz 10 MHz 10 MHz 20 MHz
Partition Small Medium Medium Large Large Large
Regions 734 176 176 12 12 12
Downlink A B C D E F
B
734 CMAs
C
176 EAs
Day 3 8
D
E
6 REAGs
F
Day 4 12
C
Day 5 16
D
Stage 2 31
D
B 40% discount
C
Final 161
D
0M 100M 200M 300M 400M 500M 600M 700M 800M 900M 1000M 1100M 1200M 1300M 1400M 1500M 1600M 1700M 1800M 1900M 2000M 2100M
Sum of pwb amount per 10 MHz for each block broken down by round. Color shows details about pw_bidder. Size shows details about license_s ize_mhz. The view is filtered on pw_bidder and round. The pw_bidder filter excludes . The
round filter keeps 8, 12, 16, 31 and 161.
license_size_mhz
10
20
Limited substitution: 700 MHz
62 MH
MHz, 261 rounds,
d $19
$19.66 billi
billion
Block A B C
Bandwidth 12 MHz 12 MHz 22 MHz
Type paired paired paired
Partition 176 734 12
Price $1.16 $2.68 $0.76
Verizon AT&T
CEPT band plan from Electronic Communications Committee Decision (05)05
Type Paired (FDD uplink) Unpaired (TDD) Paired (FDD downlink)
Lot 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6
Frequency 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8
0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5
Let auction determine band plan
Increase in unpaired spectrum maintaining 120 MHz duplex spacing
Increase in unpaired spectrum maintaining 120 MHz duplex spacing
Type Paired (FDD uplink) Unpaired (TDD) Paired (FDD downlink) Unpaired
Lot 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 2 3 4 5 6 7 8 9 10 15 16 17 18
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6
Frequency 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8
0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5
Lot 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6
Frequency 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8
0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5
Key design choices
• Generic 5 MHz lots
– Lots are perfect substitutes
• Package bids
– No exposure problem
• Clock stage
g
– How many paired? How many unpaired? Supply = 38
– Continue until no excess demand
• Supplementary
pp y bids
– Improve clock bids; add other packages
• Principal stage
– Find value maximizing
gggeneric assignment
g and base p
prices
• Assignment stage
– Contiguous spectrum
– Top-up
p p bid to determine specific
p assignment
g
P i i rule
Pricing l
Pricing rule
• In clock stage? In assignment stage?
• Pay-as-bid pricing
– Incentives
I ti for
f demand
d d reduction,
d ti bid shading
h di
• Bidder-optimal core pricing
– Maximize incentives for truthful bidding
Bidder-optimal
Bidder optimal core pricing
• Minimize payments subject to core
constraints
• Core = assignment and payments
such that
– Efficient:
Effi i t Value
V l maximizing
i i i assignment
i t
– Unblocked: No subset of bidders prefers to
offer seller a better deal
Optimization
• Core point that minimizes payments readily
calculated
– Solve Winner Determination Problem
– Find Vickrey prices
– Constraint generation method
(D and
(Day dRRaghavan
h 2007)
• Find most violated core constraint and add it
• Continue until no violation
• Tie-breaking rule for prices is important
– Minimize distance from Vickrey prices
5 bidder example with bids on {A,B}
• b1{A} = 28 Winners
• b2{{B}} = 20
• b3{AB} = 32 Vickrey prices:
• b4{A} = 14 p1= 14
• b5{B} = 12 p2= 12
The Core
Bidder 2 b4{A} = 14 b1{A} = 28
Payment b3{AB} = 32
Efficient outcome
20
b2{B} = 20
The Core
12
b5{B} = 12
14 28 32 Bidder 1
Payment
Vickrey prices: How much can each winner’s
bid be reduced holding others fixed?
Bidder 2 b4{A} = 14 b1{A} = 28
Payment b3{AB} = 32
b2{B} = 20
20
The Core
12
Vickrey
prices b5{B} = 12
Problem: Bidder 3
can offer seller
more (32 > 26)!
14 28 32 Bidder 1
Payment
Bidder-optimal core prices: Jointly reduce
winning bids as much as possible
Bidder 2 b4{A} = 14 b1{A} = 28
Payment b3{AB} = 32
b2{B} = 20
20
The Core
12
Vickrey
prices b5{B} = 12
Problem: bidder
bidder-
optimal core prices
are not unique!
14 28 32 Bidder 1
Payment
Core point closest to Vickrey prices
Bidder 2 b4{A} = 14 b1{A} = 28
Payment b3{AB} = 32
b2{B} = 20
20
Unique
core prices
15
12
Vickrey
prices b5{B} = 12
Minimize
incentive to
distort bid!
14 17 28 32 Bidder 1
Payment
Why core pricing?
• Truthful bidding nearly optimal
– Simplifies bidding
– Improves efficiency
• Same as Vickrey if Vickrey in core
(
(substitutes)
)
• Avoids Vickrey problems with complements
– Prices that are too low
• Revenue is monotonic in bids and bidders
• Minimizes incentive to distort bids
A i i rule
Activity l
Activity rule: Eligibility points
• Clock stage: Cannot increase package size
• Supplementary bids: Whenever reduce package
size,, value on all larger
g p packages
g capped
pp by y
prices at time of reduction
– Example
• Bidder drops from package of size 10 to 6 at prices p
• For all packages q of size 7 to 10, bid q p
• Implication
– Profit maximization is poor strategy
– Bid to maximize package size subject to profit 0
Full-scale test of design
(M l d and
(Maryland d GMU PhD students)
d )
• Experienced subjects
– PhD course in game theory and auctions
– Prior participation in package clock auction
• Motivated subjects
– Average subject payment = $420
$
• Realistic scenarios
Result
Supply
4a 4b 6a 6b
max profit max size max profit max size max profit max size max profit max size
100%
23 28 number of bids
79
40
max profit average efficiency = 95% (46 bids on average) 69
95%
38
45
90% 45
85%
69
38
77
69
55
80%
max size average efficien
ncy = 79% (62 bids on avverage)
39 89
ency
75%
Efficie
59
70%
65%
60%
55%
50%
5 15 5 15 5 15 5 15 5 15 5 15 5 15 5 15
b ( q ) b ( qs ) ( q qs ) ps
Properties with substitutes
• Bidding on most profitable package is best
• Clock yields competitive equilibrium with
efficient assignment and supporting prices
• No supplementary bids needed
• Final assignment = clock assignment
• Prices reduced to opportunity
pp y costs
Properties in general
• Bidding on most profitable package is nearly
best
• If no unsold lots at end of clock,, then
– Final assignment = clock assignment
– No supplementary bids needed
• If unsold lots at end of clock, then
– Supplementary bids needed
– Clock winner can guarantee it wins final clock
package
(raise by final clock price of unsold lots)
Experimental results
• 100% efficiency in nearly all cases
• Safe strategy adopted by bidders
– Clock
Cl k stage
t
• RP: Bid on most profitable package
• EP: Bid on largest profitable package
– Supplementary round
• Bid full value on all relevant packages
– Assignment stage
• Bid incremental value for specific assignments
Model, simulation, and experiment
U
Unpaired
i dD Demand
d (PPUU - inelastic)
i l ti )
600
500
Experiment
Simulation
Theory
400
Eligibility Point (AV)
Eligibility Point (AV)
Price
300
200
0
0 5 10 15 20 25 30 35
Quantity Demanded
C
Conclusion
l i
Conclusion
• Package clock auction
– Eliminates exposure
– Eliminates gaming
– Enhances substitution
– Allows auction to determine band plan
– Readily customized to a variety of settings
– Many
M other
th applications
li ti ((airport
i t slot
l t auctions)
ti )
Conclusion
• Harness p
power of markets
• Improve pricing
– Efficient decisions, short term and long term
– Innovation from price incentives
• Enhance competition
– Price
P i ttransparency
– Enhanced substitution and liquidity
– Reduced transactions costs
• Mitigate market failures
– Market power, coordination, externalities, …