Best practice spectrum auctions workshop, New Delhi 24 Sep 2014

Criteria for successful
spectrum auctions
Spectrum Auctions: Best Practices Workshop
New Delhi, 24th of September 2014
Stefan Zehle, CEO, Coleago Consulting Ltd
+44 7974 356 258 stefan.zehle@coleago.com
www.coleago.com

Agenda
Successful Spectrum Auction Design - © Copyright Coleago 2014 - www.coleago.com
1
1 Introducing Coleago Consulting
2 Industry trends influencing spectrum policy & management
3 Considerations for allocating spectrum through auctions
4  Available spectrum lots and competition safeguards
5  Reserve prices and licence payment terms
6  Licence terms
7  Licence extensions
8 Challenging the auction orthodoxy
9 Recommendations for best practice in spectrum auctions

About Coleago Consulting
A specialist telecoms management
consulting firm
2

Since 2001, Coleago has offered a wide range of advisory
services to the telecom industry
3
Strategy & Business Planning
 Strategy Development, Marketing Strategy
 MVNO and Multi-Brand Wholesale Strategy
 Business Planning and Business Modelling
Telecoms Regulation & Interconnect
 Accounting Separation, Regulatory Price
Control
 Interconnect Cost Modelling, RIO
 Regulatory Consultations
Transaction Services
 Commercial Due Diligence
 Tower Due Diligence
 Preparation of Information Memorandum
Business Transformation & Cost
Reduction
 Cost Reduction
 Mobile Network Sharing
 Restructuring and Turnaround
Spectrum Valuations and Auctions
 Spectrum Strategy
 Spectrum Valuation for Auctions
 Spectrum Auction Bid Strategy and Execution
 Beauty Contest Bid Books
Mobile Network Sharing
 Mobile Network Sharing
 Managed Services and Outsourcing
 Tower Due Diligence
 Network Audit

Coleago has carried out over 70 spectrum consultation,
valuation, auction and beauty contest licence projects
Completed in 2013/4
 Chile - 900/1800 and 700/AWS
 Canada - 700MHz
 Paraguay - multi-band
 Oman - 800MHz & 2.6GHz
 Belgium - 800MHz
 New Zealand - 700MHz
 Myanmar - greenfield
 Australia - 700MHz & 2.6GHz
 UK - 800MHz & 2.6GHz
 Sri-Lanka - 1800MHz
Completed in 2012
 Belgium - 2.6GHz
 Netherlands - multi-band
 New Zealand -1800MHz spectrum
trading
 Switzerland - multi-band
 Russia - 700MHz & 2.6GHz
 Pakistan - 2.1GHz valuation
 Bangladesh - 2.1GHz valuation
4

Industry trends influencing
spectrum policy & management
Mobile data drives the need for
spectrum
5

New technology enables operators to offer faster services and
enables to pass more traffic through a given amount of spectrum
GSM
 Not well suited for modern data needs
 Download speed of up to 384 kbps
with EDGE technology
3G HSPA
 Spectral efficiency: 0.7 bits / Hz / cell
 Download speed of 42Mbps
LTE and LTE Advanced
 Spectral efficiency: 1.4 bits / Hz / cell
(possibly twice that for LTE-A)
 Download speed of 150Mbps (300 for
LTE advanced)
6
3G HSPA

Demand for mobile broadband requires more mobile
network capacity
The mobile data tsunami
 Mobile data traffic is growing at a
very rapid rate in all regions of
the world.
 In many mobile networks data
now exceeds voice.
 Technology alone cannot deliver
the required capacity, additional
spectrum is required
7

Mobile broadband is a key ingredient for the development
of the digital economy …
 An increase of 10% in mobile adoption
boosts GDP growth by 0.8% (World
Bank, 2009)
 For a given level of total mobile
penetration, a 10% substitution from
2G to 3G increases GDP per capita
growth by 0.15 % points (Deloitte,
2012)
 Doubling broadband speeds for an
economy can add 0.3% to GDP growth
(Arthur D. Little, 2011)
8

… and there are tangible benefits to society which
illustrate the impact of mobile data
 A 12% increase in financial inclusion in
developing countries in India and
Bangladesh
 Healthcare: up to 70% improved
compliance for TB
 10-15% increase in farmer income
 mEducation solutions can significantly
improve the affordability of education
by up to 65%
9

Existing and new spectrum is required for mobile
broadband services
10
700/800 MHz
850/900 MHz
1800/1900 MHz
1700/2100 MHz
2600 MHz
?
New spectrum for LTE, in some markets previously
used for TV, referred to as “digital dividend” bands
Originally only used for GSM, progressive
redeployment to 3G HSPA and recently to LTE
Originally only used for GSM, progressive
redeployment to 3G HSPA and recently to LTE
Currently used for 3G, upgrading to dual carrier
HSPA+, LTE deployment in the Americas
New band for LTE
The mobile industry is seeking over 1GHz new
spectrum for mobile broadband

Implications for spectrum management and auctions
Supply of new spectrum
 Focus on making a maximum of
spectrum available for mobile
broadband as fast as possible
Allocate new spectrum
 Allocate new spectrum to mobile
operators to facilitate and encourage
rapid deployment of 3G HSPA and
LTE
New technology in existing spectrum
 Ensure that new technology, notably
LTE, can be deployed in existing
bands
11

Considerations for allocating
spectrum through auctions
Spectrum auctions provide transparency
but do not necessarily result in efficient
spectrum allocations unless they are
designed consistent with best practice
12

Auctions are the default mechanism for spectrum
allocations
Beauty contests were used at the start
of the mobile industry growth
 Difficult to administer
 Open to dispute due to subjectivity
Since 2000 auctions are the norm in
spectrum allocations
 Transparent process (no subjectivity)
 Policy objective: maximise economic
efficiency
 In theory, whoever values spectrum
the most will produce the greatest
social good
13

Implicitly, auctions focus on maximising revenue from
whatever is sold
14

Policy objectives for the allocation of mobile spectrum
are wider than maximising auction proceeds
 Promote the highest value use of
spectrum
 Ensure spectrum is deployed rapidly
and widely and the maximum spectral
efficiency is extracted
 Promote investment and innovation
 Promote rural broadband access and
increase digital participation rates
 Promote competition
 Promote customer convenience
 Provide a high net economic return to
the public
 Immediate revenue generation by
maximising auction proceeds
15

Spectrum auctions provide transparency but do not
necessarily deliver the policy objectives efficiently
Assessing best practice in spectrum
auctions should cover at a minimum
the following aspects:
1. Available spectrum lots and
competition safeguards
2. Reserve prices and licence payment
terms
3. License terms
4. Licence extensions or renewals
16

1. Available spectrum lots and
competition safeguards
Overcoming risks associated with
fragmentation and the implications of LTE
17

Available spectrum lots and competition safeguards
The quantity, nature and
packaging of the spectrum that
individual operators are able to
pursue will bear significantly on
auction outcomes.
Relevant aspects include:
 Supply of spectrum
 Spectrum set-asides, floors and/or
spectrum-acquisition limits
 Allocation of specific or generic lots
 Lot sizes
 The inclusion of expiring usage-rights
in combined auctions
 Competition safeguards
18

Efficient allocation requires the timely availability of large
amounts spectrum
 The Swedish telecoms regulator PTS
is at the forefront of best practice
spectrum allocation and management.
 PTS shall increase the availability of
useful spectrum through least
restrictive conditions, in the work for
international harmonisation,
assignments at a good rate to meet
demand, and promotion of secondary
trading. PTS Spectrum Strategy, Draft
summary of the consultation report, Feb 2014
 PTS’s spectrum strategy recognises
that spectrum which is held back and
hence not used, delivers no socio-economic
value.
The estimated total spectrum
requirements for both the RATGs
1 (pre-IMT, IMT-2000, and its
enhancements) and 2 (IMT-Advanced)
are 1,340 MHz and
1,960 MHz for lower user density
settings and higher user density
settings, respectively.
Future spectrum requirements estimate for
terrestrial IMT, Report ITU-R M.2290-0,
(12/2013)
19

When a new band is released, all of the spectrum in that
band should be made available at once
Auctioning small amounts of
spectrum is inefficient
 When few lots are on offer demand will
exceed supply by a greater factor.
High auction prices will reduce
investment.
 Small amounts of spectrum increase
deployment costs and prevent
operators from delivering true mobile
broadband services.
 LTE and LTE advanced require an
allocation of at least 2x10MHz or
2x20MHz of contiguous spectrum per
operator.
20
New spectrum for LTE: More and
wider is better
Band MHz
7 (2.6GHz) 2x70MHz + 40MHz TDD
28 (APT700) 2x45MHz

Spectrum set-asides, floors and/or spectrum-acquisition
caps can lead to inefficient outcomes
 Spectrum set-asides, floors and/or spectrum-acquisition caps are typically
designed to prevent excessive spectrum concentration.
 However, if at all, these measures need to be determined with great care to avoid
unduly distorting outcomes.
21
Potential impact of auction design Examples
Spectrum is acquired by inefficient users who
deploy little and fail to gain market share
Chile AWS auction (2009)
Canada AWS (2008)
Spectrum remains unsold and hence the
economic value is not extracted
Netherlands 2.6GHz (2010)
Belgium 2.6GHz (2011)
Spectrum is not deployed and held for resale
at a profit for private investors Canada AWS (2008)
Increased spectrum costs for incumbents
damage operator Netherlands 800MHz (2012)

Ensuring a minimum block size of 2x10MHz is key for
efficient LTE deployment
 Deploying LTE in 2x15MHz costs
around $3,900 per MHz; deploying in
only 2x5MHz costs $9,900 per MHz.
 The maximum downlink speed in
15MHz is 112 Mbps compared to only
35 Mbps in 5MHz.
 Potential solutions:
– Allocate wide enough bands to
individual operators
– Allow spectrum sharing so that
operators who hold, say 2x5MHz
each, may jointly deploy in 2x10MHz
– Allow spectrum trading
Capex per LTE e-NodeB
70,000
60,000
50,000
40,000
30,000
20,000
10,000
22
-
1 Carrier 3 Carriers
US$
eNodeB equipment 5MHz Carrier Cost

Auctioning specific blocks of spectrum in parallel may
lead to non-contiguous allocations
 Potential non-contiguous allocations are a key drawback of a regular SMRA
 Not technically efficient
 Vulnerable to anti-competitive bidding (e.g. attempt to isolate individual blocks)
23
APT Band Plan allocated in 2 x 5 MHz blocks
B1 B2 B2 B2 B2 B1 B1 B3 B3
703 MHz
& 758 MHz
748 MHz
& 803 MHz
Example: Bidders B2 and B3 have contiguous allocations, while B1’s allocation
is fragmented, thus increasing deployment costs and reducing efficiency

Market based solutions to ensure contiguous spectrum
holdings
A possible solution to prevent non-contiguous
spectrum holdings, is
to auction generic lots instead and
to assign contiguous ranges during
a separate process, after the
quantities secured by each bidder
are known.
 The use of generic blocks might skew
results if the position within a given
band has a significant impact on
values: uncertainty over the final
assignment may distort bid behaviour.
 An approach taken in some instances
(e.g. the UK multiband auction in 2013)
has been to include a single specific lot
in a band alongside generic lots, and
applying a contiguity constraint during
the assignment stage.
 This mitigates both the risk of technical
inefficiency and of distorting the
allocation process.
24

The inclusion of expiring usage-rights in combinatorial
(package bidding) auctions is poses great risks
In some cases, spectrum on offer in a
combinatorial auction included
expiring licences alongside new
mobile spectrum
 All existing spectrum in use was sold
together with the new Digital Dividend
frequencies, in ‘Big Bang’ combined
awards in Switzerland, Ireland and the
Netherlands during 2012.
 The Norwegian 800MHz auction that
concluded in December 2013 also
included expiring 900MHz and
1800MHz licences.
These auctions introduce a significant
risk to business continuity
 In Norway, incumbent Tele2 lost all of
its existing 2G holdings, leaving the
business crippled. Tele2 has since
announced its exit from the Norwegian
market, with a sale of its assets to rival
TeliaSonera.
 “Big Bang” auctions are vulnerable to
distortions caused by the adoption of
predatory bidding strategies by a
dominant operator, as evidenced
Switzerland.
25

Auction design no longer focuses on new market entry
but accepts a reasonable level of consolidation
 Wireless markets are mature. At the
maturity stage of the industry life cycle
we can expect consolidation but not
new market entry, at least at network
level.
 Ensuring efficient use of spectrum in
competitive markets with an optimum
number of operators becomes a policy
goal.
In Europe, regulators are now
accepting that the benefit of
consolidation from 5 or 4
operators to 3 exceeds the
potential negative competition
impacts, provided safeguards are
put in place.
26

2. Reserve prices and licence
payment terms
Setting appropriate reserve prices in the
context of policy objectives
27

High reserve prices are prone to distort auction outcomes
and harm the public interest in a number of ways.
 Spectrum may be left unsold and
hence unutilised. This represents a
productivity loss to society and
reduced auction receipts.
 National imbalances in spectrum
holdings may be exacerbated.
 An unnecessarily high cost-burden
may be imposed on the industry,
leading to adverse downstream
consequences in terms of roll-out,
competition and consumer choice.
28
Potential impact of auction design Examples
Spectrum remains unsold and hence economic
value is lost to the country
India 850MHz (2012, 2013)
Australia 700MHz (2013)
Higher costs are imposed on operators than
necessary and deployment slows down
Australia 700MHz (2013)
Belgium 800MHz (2013)

Driven by the desire to plug a hole in the budget,
Australia set extremely high reserve prices for 700MHz
0.09
0.32
0.30
0.30
0.47
0.49
0.60
29
700/800MHz Digital Dividend Spectrum Reserve Prices Compared
0.00
0.13
0.25
0.42
0.56
0.80
1.35
-
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
1.10
1.20
1.30
1.40
Australia - 5/2013
Italy - 9/2011
France - 12/2011
Portugal - 12/2011
Canada - Q4/2013
Spain - 7/2011
New Zealand - 10/2013
Finland - Q4/2013
Switzerland - 2/2012
UK - 2/2013
Sweden - 3/2011
Denmark - 6/2012
Netherlands - 12/2012
Germany - 5/2010
US$ / MHz / Pop

The Australian reserve prices where set higher than
prices paid at auction in other countries
700/800MHz auction prices paid vs. Australian reserve prices
0.49
0.73
30
1.28
0.91
0.58
0.81
0.56
0.88
0.37
0.65
1.35
Australia Reserve…
- 0.20 0.40 0.60 0.80 1.00 1.20 1.40
Average 700/800MHz
UK - 2/2013
Denmark - 6/2012
France - 12/2011
Portugal - 12/2011
Italy - 9/2011
Spain - 7/2011
Sweden - 3/2011
Germany - 5/2010
USA - 2/2008
US$ / MHz / Pop

The Australian APT 700MHz auction resulted in an loss to
society and is an example of policy failure
 Potential socio-economic gain for
Australia?
 Is the spectrum is actually used?
 Can operators deploy the 700MHz
band as cost effectively?
 Is there competition to drive down
prices?
 Between AU$ 7bn and AU$10bn
 2x15MHz of 2x45MHz unsold
hence not all of the potential socio-economic
gain is realised
 Only Telstra obtained 2x20MHz,
can deploy at lowest cost, Optus
obtained only 2x10MHz
 One operator, Vodafone, did not
obtain any spectrum and the
leading operator Telstra increased
its competitive advantage, thus
reducing competition
31

Lessons learned from the Australian 700MHz auction
High reserve prices are not a good
approach to spectrum auctions
 They have a market distorting effect
 Regulators might do not achieve their
policy objectives
 Even if a large amount of money is
raised up-front this is likely to reduce
overall economic value in the long term
32

The societal value of allocating spectrum
 The return to the community from
spectrum auctions goes well beyond
any direct payment made to
government for spectrum.
 Implicitly all governments recognise
the trade-off between spectrum fees
and wider goals.
 Otherwise they would simply auction
off monopolies which would
undoubtedly bring the highest direct
receipts.
33

Setting high prices for spectrum is problematic
Hazlett and Munoz, “What Really Matters in
Spectrum Allocation Design”, 2010
“[T]he ratio of social gains [is of] the order
of 240-to-1 in favour of services over
licence revenues…Delicate adjustments
that seek to juice auction receipts but
which also alter competitive forces in
wireless operating markets are inherently
risky. A policy that has an enormous
impact in increasing licence revenues
need impose only tiny proportional costs
in output markets to undermine its social
utility.
In short, to maximise consumer welfare,
spectrum allocation should avoid being
distracted by side issues like government
licence revenues.”
34

Benchmarking based on auction outcomes is not an
appropriate method to set reserve prices
 The value of spectrum to a mobile
operator is specific to its business and
based on a discounted cash flow
forecast for the business.
 Prices paid at an auction reflect
outcome based on the value operators
assign to spectrum in that country
under a particular set of
circumstances.
 More recently prices paid reflect high
renewal prices. For example, the cash
strapped government in Greece in
2011 set extremely high reserve prices
to renew spectrum. Operators had the
choice to pay up or shut down.
35
Benchmarking

Reserve prices should be set based on cost
 Two costs of allocating spectrum to
mobile can be identified.
 Reserve prices should be set to
compensate for the higher of either
one of those costs:
– The opportunity cost, i.e. the value
that would be generated from the
next best alternative use of the
spectrum.
– The cost moving incumbent users to
free up the spectrum for mobile use.
Example New Zealand, 700MHz
auction 2013 announcement
The reserve price for each of the
nine lots of 5 MHz paired has been
set at NZ$22 million [NZ$198
million in total].
The Government has spent $157
million clearing the 700 MHz band
to allow the spectrum to be used for
4G mobile networks.
Communications and Information
Technology Minister Amy Adams, 700MHz
auction announcement, 4 Sep 2013
36

Offering the option of upfront versus staggered payments
constitutes best practice
 “Allowing staged payment will enable mobile
network operators to invest immediately in building
their 4G networks to increase their service to New
Zealanders.” Communications and Information Technology
Minister Amy Adams, 700MHz auction announcement, Sep 2013
 In principle, operators can reflect the timing of
payments in their valuations. However,
requirements to pay the full amount as a lump sum
may have a disproportionate impact on more highly
geared operators, which could threaten allocation
efficiency.
 Offering the option of upfront versus staggered
payments, subject to a market-based interest rate,
reduces this risk and therefore constitutes best
practice.
37

3. Licence terms
Applying best practice to licence terms is
likely to have beneficial impact on
investment in mobile and the socio-economic
benefits which flow from this
38

Setting spectrum licence terms to maximise spectrum
value to society
 Spectrum has no intrinsic value. Value
is only created through the use of
spectrum.
 The more spectrum is used, the more
socio-economic value is created.
Therefore spectrum licence terms
should encourage the maximisation of
the use of spectrum. In practice this
means setting terms that encourage
investment in the radio access
network.
39

An illustration of the effect of network investment on
value extracted from spectrum
Operator A Operator B
There are two mobile operators in
a country.
40
 Holds 2x10MHz of
spectrum
 Invested $300 million
 2,000 sites
 1 million customers
 Monthly MoU: 200
 Monthly ARPU: $10
 Annual revenue:
$120.0 million
spectrum
 4,000 sites
 1.6 million customers
 Monthly ARPU: $11
 Annual revenue:
$211.2 million
Each holds the same amount of
spectrum.
Operator B invested twice as
much as operator A.
Since operator B has a much
better network, operator B
attracted 60% more customers
than operator A. As a result of
better network quality B’s
customers also make more calls
and generate a higher ARPU.

Greater network investment increases efficient use of
spectrum measured in minutes per MHz per year
41
spectrum
 1 million customers
 2,400 million minutes
per year
spectrum
 1.6 million customers
 4,224 million minutes
per year
Which operator has invested
more?
 120.0 million minutes
per MHz per year
 211.2 million minutes
per MHz per year
 2,000 sites
 4,000 sites
Both operators hold the same
amount of the scarce resource
that is spectrum
Which operator is delivering more
economic benefit to the country?
Which operator is passing more
traffic through a given amount
spectrum?

Operator B experiences a lower % return on investment
but delivers greater value to the country
42
Spectrum MHz 2x10 2x10
Investment $ mn 300 600
Number of sites 2,000 4,000
Customers million 1.0 1.6
Monthly ARPU $ 10 11
Annual revenue $ mn 120.0 211.2
EBITDA 40% 40%
EBITDA $ mn 48.0 84.5
Annual capex 5% initial 15.0 30.0
Free cash flow 33.0 54.5
Annual return on investment 11.0% 9.1%
Monthly minutes of use 200 220
Annual minutes mn 2,400 4,224
Minutes (mn) per year per MHz 120.0 211.2
Operator A extracts more
value for the private
investors
Operator B generates
more value for the country

Now the government decides to levy fee of 4% of revenue
and describes it as “spectrum usage charge”
No charge With 4% charge
43
Operator A Operator B Operator A Operator B
Annual revenue $ mn 120.0 211.2 120.0 211.2
Spectrum usage fee - - (4.8) (8.4)
EBITDA 40% 40% 36% 36%
EBITDA $ mn 48.0 84.5 43.2 76.0
Annual capex 5% initial 15.0 30.0 15.0 30.0
Free cash flow 33.0 54.5 28.2 46.0
Annual return on investment 11.0% 9.1% 9.4% 7.7%
Drop in free cash flow (4.8) (8.4)
Operator B’s profitability declines more that operator A and the return
on investment is even lower. A rational investor would invest less and
pursue operator A’s strategy. As a result the country would lose out.

Spectrum usage charges to recover administrative costs
are consistent with spectrum policy objectives
A fee per MHz of spectrum:
Encourages operators make as much
use of spectrum as possible, i.e.
encourages investment
Is easily calculated and transparent
Covers the cost spectrum
management
A fee based on revenue:
 Penalises operators who make
efficient use of spectrum
 Discourages investment in the network
 Reduces the socio-economic value of
spectrum
44
The initial licence fee already covers the opportunity cost of allocating the
spectrum for mobile use, hence society is already adequately compensated.
Calling a levy on revenue a “spectrum usage charge” is a
misnomer because it has nothing to do with the policy of objective
of maximising the use of spectrum. It is simply a form of taxation.

Relevant European Union legislation is generally
considered at the forefront of good practice
DECISION No 243/2012/EU OF THE
EUROPEAN PARLIAMENT AND OF
THE COUNCIL,
of 14 March 2012,
establishing a multiannual radio
spectrum policy programme
Article 2: paragraph 1 (a)
General regulatory principles
(a) applying the most appropriate and
least onerous authorisation system
possible in such a way as to maximise
flexibility and efficiency in spectrum use.
Such an authorisation system shall be
based on objective, transparent, non-discriminatory
and proportionate criteria.
45

Best practice relates spectrum usage charges to the cost
of spectrum management
DIRECTIVE 2002/20/EC OF THE
EUROPEAN PARLIAMENT AND OF
THE COUNCIL
of 7 March 2002
on the authorisation of electronic
communications networks and
services (Authorisation Directive)
Article 12, paragraph 1 (a)
Administrative charges
Any administrative charges imposed on
undertakings providing a service or a
network under the general authorisation
or to whom a right of use has been
granted shall:
 (a) in total, cover only the
administrative costs which will be
incurred in the management, control
and enforcement of the general
authorisation scheme and of rights of
use…
46

Best practice provides certainty over investment: Long
licence terms and expectation of renewal
High expectation of renewal
Certainty over business continuity
improves the business case and
hence 20 year terms are
recommended
 Investors are prepared to accept lower
returns and as consequence the ratio
of investment to revenue increases.
 Prices can be than they otherwise
would be.
High expectation of renewals at
reasonable terms constitutes now
best practice
 UK and New Zealand are leading in
this area.
 In the UK renewal will be the norm
unless there are spectrum
management reasons not to do so.
Administered Incentive Pricing is used
to determine the fee payable.
47
15 Years 20 Years

4. Licence extension
Why auctions may not be appropriate for
existing spectrum rights
48

Extending existing licences poses different problems
Licence extensions or renewals are
now common as many 15 or 20 year
licence come to the end of their term
 Not obtaining an licence extension or
renewal, may put a viable business at
risk.
 There is evidence that mobile
operators reduce investment as the
end of the term approaches.
 If licences are not extended,
considerable disruption would result
with a cost to consumers, employees
and the attraction for future investment
49

Auctioning spectrum due for renewal may create
asymmetries that invite predatory bidding
Licences expiring Licence continuing for another 3 years
50
Operator A Operator B Operator C
 Needs to renew
spectrum to continue
operating, hence the
value is very high, say
$500 million.
 Cash will be diverted
from investment to
paying for spectrum.
 Do not have to renew spectrum, but the additional
spectrum would have a value of $100 million to B & C.
 Operators B and C estimate the value of the expiring
spectrum licence to A at $500 mn, and hence bid
$300. They are unlikely to win, but have imposed a
high cost on A.
 B & C know that because financing is not unlimited
this will slow down A’s investment and thus reduce
competitive pressure for B and C to invest.
Collectively the industry will end up investing less and
the socio-economic benefit of spectrum is reduced.

Solutions for licences extensions or renewals
Harmonise licence expiry dates
 In many markets licences do not expire at the same
time. Some regulators harmonised licence expiry by
granting partial extensions, e.g. Ireland.
 All expiring spectrum could then be auction together,
thus eliminating the predatory bidding problem that may
otherwise arise from asymmetries.
Use an administered method with pricing based on
opportunity cost
 Recognise that auctioning spectrum that is in use by a
successful mobile operation is not necessarily the best
method.
 An administered process based on the opportunity cost
to other potential users (AIP) may be a better solution.
51

Challenging the auction
orthodoxy
Should auctions always be considered as
the right approach for spectrum
allocations?
52

53
Spectrum auctions worked fine in past,
so what’s different now?

We need to rethink the method of allocating spectrum in
the light of maturing mobile markets
Mobile markets have reached the
maturity phase of the industry life
cycle
 Many markets show flat, at least in real
terms) or declining revenues and
EBITDA
 This maturity industry life cycle stage
suggest that that policy goals should
be revised:
– Encouraging new network based
competition is not be appropriate
– Taking cash out of the industry is not
sustainable
54

Maturing markets are characterised by consolidation, not
new market entry
Mobile industry consolidation is in full
swing
 The pace and size of cross-border
M&A has been breath-taking, with five
mega-deals announced or completed
during the past three months.
 Markets with consolidation potential
include India, Indonesia, Canada, Italy,
Germany and Brazil - although
regulation is likely to be a constraint in
most of these.
 Not surprisingly, we are seeing
numerous infrastructure sharing deals.
Investors should expect further M&A,
but at a less frantic pace.
55

Given the existing spectrum, new entrants will have too
little spectrum to compete
In an LTE world, large contiguous
spectrum holdings confer particular
competitive advantage
 The exit of some operators in Europe
and the insolvency of Mobilicity in
Canada demonstrates that it is
impossible to succeed in the market
with small spectrum holdings.
 When industry logic has driven
consolidation, trying to reverse the
process by regulatory is unlikely to
produce societal benefits.
56

Using new spectrum auctions to increase network-based
competition is unlikely to succeed
Regulators may wish to consider:
 Allocating spectrum in a manner which
does not reduce competition while at
the same time maximising the benefit
of a wide band.
 Facilitating a transition from network
based competition to other forms of
competition.
 Focusing on other regulatory remedies
if competition is failing, such as a
regulated access price offer. The
conditions attached to Hutchison
Three’s acquisition of Orange Austria
can serve as an example.
57
Consolidation is
normal when the
industry life cycle
reaches the maturity
stage
Wide band allocations
are required for an
economically and
spectrally efficient
deployment of LTE

58
Alright, we are unlikely to get new network based
marketing entry, but why not still have auctions?

Options for spectrum auctions in mature markets
59
New market entry unlikely
Each incumbent gets a “fair
share”, but auction proceeds are
low because there is no real
bidding
Set high
reserve prices
Unfettered auction among
incumbents
Auction proceeds may be high,
but increased industry
consolidation and reduced
competition results
Spectrum caps to preserve
existing competition
Focus on
other policy
goals
What then is the
point of an
auction?

Setting high prices for spectrum is problematic
Hazlett and Munoz, “What Really Matters in
Spectrum Allocation Design”, 2010
“[T]he ratio of social gains [is of] the order
of 240-to-1 in favour of services over
licence revenues…Delicate adjustments
that seek to juice auction receipts but
which also alter competitive forces in
wireless operating markets are inherently
risky. A policy that has an enormous
impact in increasing licence revenues
need impose only tiny proportional costs
in output markets to undermine its social
utility.
In short, to maximise consumer welfare,
spectrum allocation should avoid being
distracted by side issues like government
licence revenues.”
60

61
But why can’t we simply set high prices for spectrum,
surely the industry can pay up?

To fulfil societal goals for mobile broadband connectivity,
mobile operators require large amounts of spectrum
Radiocommunication Study Groups
Document 5D/XX-E, Sep 2012
62
The GSMA has commissioned
Coleago to undertake some initial
spectrum requirement estimates for
IMT to the year 2020. A report on
this work from Coleago is attached
indicating the total spectrum
required for IMT of 1600 to 1800
MHz for the year 2020. The GSMA
believes this is a reasonable
number.

Extracting high spectrum fees from the mobile industry is
not sustainable
63
Demand for Mobile
Broadband and Spectrum
Requirement
Prices Paid for Spectrum
LTE Deployment and
Backhaul Capex
Tangible (Network) and
Intangible (Spectrum)
Capex
Revenue
EBITDA
Free Cash Flow
Impact on Operators
Balance Sheet
Return on Capital
Employed (ROCE)
EBITDA Margin %
+
=
+
=
The consequence: The only
way to maintain ROCE is to
reduce investment

When returns drop below the cost of capital, investment
ceases to flow into the industry, example Latin America
Cash flows from operations are
declining
 Operators in Latin America have seen
EBITDA margins decline in recent
years.
 In Q2 2013, the average service
revenue EBITDA % margin for Latin
America was 34.3% compared to
39.4% in North America and EBITDA
cash flows showed a significant year-on-
year decline.
Capital expenditure pressure is
increasing
 Capex in the Latin American mobile
industry is set to increase driven by
LTE deployment.
 However, this is only the investment in
tangible assets.
 Spectrum capex is a key variable in
determining total capex.
64

65
How can we do things differently?

2009 AWS auction in Chile focused on stimulating new
market entry, but resulted in policy failure
Spectrum caps guaranteed new
market entry …
 A spectrum cap of 60 MHz, effectively
excluding the three incumbent mobile
network operators - Movistar, Entel
and Claro.
 Cable television company VTR won
30MHz of the AWS spectrum paying
US$3.02 million, and Nextel won
60MHz, paying US$14.7 million.
 Revenue raised amounted to a tiny
$0.011 / MHz / pop.
… but failed to deliver timely
deployment and competition …
 The new entrants were unable to
launch their 3G mobile service until
May 2012, one and a half years after
the October 2010 deadline.
 VTR and Nextel together only have
1.3% market share, nearly three years
after the AWS spectrum licence award.
… and private investors may pocket
the new entrant discount.
 In a secondary market VTR and
Nextel are likely to sell the spectrum
for more than they paid.
66

The 2014 700MHz licence award in Chile broke new
ground and is likely to deliver the policy objectives
67
The 700MHz spectrum award process
focussed on connectivity and
competition policy objectives …
 connect 1,281 rural towns and 500
schools
 obligation to build fibre
 mandated MVNO access and roaming
… rather than extracting money from
the mobile industry.
 Auction proceeds amounted to a
relatively tiny 0.017 $/MHz/pop.
 The reserve price in Australia was set
at 1.35 $/MHz/pop - 78 times higher.

Recommendations for best
practice in spectrum auctions
Careful choices in spectrum auctions are
required in order to maximise the socio-economic
benefit of spectrum
68

A formal consultation process and rational argumentation
should precede any spectrum allocation event
Existing total
Supply of Spectrum
Spectrum auction
design
69
Market
competitiveness
Policy objectives
spectrum
holdings
RAN and
spectrum
sharing
Licence
duration and
extension
Demand for Spectrum
Number of operators in
a market

Best practice in spectrum auctions – #1 & #2
1. Apply spectrum floors in the event
spectrum-in-use is being re-auctioned,
to minimise the threat to
business continuity.
2. Set reserve prices to reflect the cost
of alternative use or the cost clearing
spectrum so as to avoid imposing
high costs on the industry that would
be passed on to consumers in the
form of higher retail prices and result
in reduced investment.
70

Best practice in spectrum auctions – #3, #4 and #5
3. Provide deferred licence fee payment
terms to equalise the opportunity for
operators to secure the resources
they need to pay for spectrum, in
light of possible differences in their
ability to raise substantial funds in
advance, and also to speed up
deployment.
4. Adopt auction rules that minimise the
scope for predatory bidding and
further efficiency.
5. Impose non-band specific roll-out
obligations provisions to minimise the
threat of spectrum hoarding.
71

Best practice in spectrum auctions - #6 and #
6. Avoid the use of auctions when an
obvious distribution of available
resources among operators can be
readily identified, to minimise the risk
of outcomes that are both perverse
and unexpected.
7. Adopt licence terms that encourage
efficient use of spectrum: Usage
charges to reflect spectrum
management costs only, long licence
terms and high expectation of
renewal by means of AIP.
72

Questions?
Stefan Zehle, MBA
CEO, Coleago Consulting Ltd
Tel: +44 7974 356 258 stefan.zehle@coleago.com
www.coleago.com

Additional information
1. Choice of auction format
2. How do mobile operators value
spectrum?
74

Choice of auction format
Regulators should examine their national
circumstances and reflect feedback from
operators in their ultimate choice of
auction mechanism
75

Three main formats are used for spectrum auctions
76
SMRA
Simultaneous
Multi-Round
Ascending
auction
 Bid on specific blocks of interest (between minimum and
maximum set by auctioneer for each block)
 ‘Standing high bids’ for each lot in each round
 Auction ends when there is no excess demand
 ‘First price’: pay what you bid
CCA
Combinatorial
Clock Auction
 Bid on packages of generic lots rather than on individual lots
 Pay ‘second price’: minimum needed to win and to avoid
‘unhappy losers’
 Separate assignment round for positioning in the band
 Also pay ‘second price’ for assignment
First-price
sealed bid
 Bid on individual lots
 First price: pay what you bid
 Auction ends when bids are opened

SMRA auctions with bidding on specific blocks carry a
fragmentation risk
Auctioning specific blocks of spectrum in parallel may lead to non-contiguous
allocations
 Key drawback of regular SMRA
 Threatens ‘technical efficiency’
 Vulnerable to anti-competitive bidding (e.g. attempt to isolate individual blocks)
2570MHz
& 2690MHz
77
B1 B2 B2 B2 B2 B1 B1 B3 B4 B4 B4 B4 B3 B3
2500MHz
& 2620MHz
Example: Bidders B2 and B4 have contiguous allocations, while B1 and
B3’s allocations are fragmented, creating significant problems for them

SMRA auctions lead to exposure risk
Risk of being stuck with an unwanted subset of the target package
 Potential value destruction: paying more than the final package is worth
 Key drawback of SMRA
 Package bidding (CCA) avoids this: either win entire package pursued or nothing
at all
Package
price
78
1 lot 2 lots 3 lots 4 lots
€
Values
Example: SMRA in a single band
 A package of 2 or 3 lots is still
profitable at current prices
 But one may ultimately be outbid
 And be left with a single,
unprofitable lot on which one is
Standing Highest Bidder

‘Winner’s curse’ arises in a sealed bid auction when a
bidder pays more than would have been necessary to win
 Typical of first-price, sealed-bid
auctions
 In the first Brazilian spectrum auction
in Bell South paid more than twice as
much as the next highest bid for the
Sao Paulo Metro licence
 Can also occur under SMRA: pursuing
a large package and failing can drive
up the price for the smaller package
ultimately secured
Over-payment
79
Winning
bid price
Next
Highest
bid
€

To avoid the winner’s curse, bidders may ‘shade’ their
bids
Demand moderation strategies in SMRA
auctions are analogous to bid shading
 SMRA invites a tacit ‘negotiation’
between rivals.
 The faster participants settle on the
final allocation, the less everyone
pays.
 But there is a risk to allocation
efficiency. By reducing demand too
much, a bidder could miss out on a
larger package that it should otherwise
have won.
Shaded
amount =
Surplus if win
80
Bidder’s
Valuation
Expected
rival
valuation
Shaded
bid value
€

First price sealed bid auctions should be avoided
“Under a first price rule, there is an incentive for bidders to reduce
the value of their bids to less than their full valuation in order to pay
as close as possible to the minimum necessary to beat other
bidders (bid shading). By doing so, they risk not winning at all
when it would be efficient for them to do so” Ofcom, UK
81

The Second Price Rule eliminates the winner’s curse …
Example1: Second Price Auction for a single lot
 Whoever values the resource most highly wins (economic efficiency)
 “2nd price” rule: pay no more than the minimum required to win
 Incentivises truthful bidding: no penalty for bidding full ‘walk-away value’
 No unhappy loser: Bidders A and B would not have been prepared to pay more
than the price paid by Bidder C
82
Bid for 1
lot 2nd price
Bidder A €40 -
Bidder B €50 -
Bidder C €100 €50
Winner
Bidder C wins (highest bid
amount) but only pays the
“opportunity cost” (amount the
auctioneer could have sold
the lot for if Bidder C were
absent)

…but this comes at a cost: a 2nd price auction for multiple
lots can lead to significant pricing differentials
Example 2: Second Price Auction for 2 identical lots
 Allocating 1 lot to bidder A and B maximised bid value and is therefore winning
 The “2nd price” (or Vickrey price) is the opportunity cost imposed by each bidder
– If Bidder A were absent, the auctioneer could have sold its winning lot to Bidder B
for €40 (the extra that B would pay for an additional lot = €100-€60)
– If Bidder B were absent, the auctioneer could have sold its winning lot to Bidder A
for €15 (the extra that A would pay for an additional lot = €75-€60)
 No-one has any grounds complain: neither Bidder A nor B were prepared to pay more
to win an extra lot, and Bidder C was not prepared to pay this price for any lots
83
Bid for 1 lot
Bid for 2
lots 2nd price
Bidder A €60 €75 €40
Bidder B €60 €100 €15
Bidder C €10 €20 -
Bidder A and B pay each
other’s marginal bid
values for an extra lot

Real Example: impact of second price rule in the Denmark
CCA based 2.6GHz auction in 2010
 Hutchison paid €0.9 million for 2x10
MHz of FDD plus 25 MHz of TDD
 The other bidders who acquired 2x20
MHz FDD paid ~20x more per MHz
 This dramatic outcome was a product
of a second price combinatorial
auction with tight spectrum caps:
– TDC, Telenor and Telia’s prices
reflected Hutchison’s bid value for
an additional lot of 2x10MHz FDD
– Hutchison’s 2x10MHz FDD could
not have been sold to anyone else,
hence the 2nd price was the
reserve price
H3G TDC Telia Telenor
84
Prices paid per MHz *

The CCA format has some technical problems that are
difficult to understand and communicate
 A CCA format may lead to inefficient allocations due to budget constraints in
conjunction with lack of visibility over price exposure.
 Given the second price rule in a CCA auction, participants have limited visibility
over their actual price exposure. This introduces a potentially intractable strategic
dilemma for bidders who are constrained by a fixed budget rather than by their
spectrum valuations.
– To minimise the risk of knock out, such bidders might focus their war chest on a
smaller package. Yet in so doing, they would be forced to squeeze their
marginal bids for extra resources, in order to respect the overall budget limit.
– This could prevent them from securing the larger package that, given the
relative valuations and constraints of each participant, it would be efficient for
them to win.
– Adopting an opposite strategy increases the chance of winning a more
ambitious prize, but also the risk of walking away with nothing.
85

Inefficient allocations due to budget constraints in
conjunction with lack of visibility over price exposure
Given the second price rule in a
CCA auction, participants have
limited visibility over their actual
price exposure.
This introduces a potentially
intractable strategic dilemma for
bidders who are constrained by
a fixed budget rather than by
their spectrum valuations.
 To minimise the risk of knock out, such
bidders might focus their war chest on a
smaller package. Yet in so doing, they would
be forced to squeeze their marginal bids for
extra resources, in order to respect the
overall budget limit.
 This could prevent them from securing the
larger package that, given the relative
valuations and constraints of each participant,
it would be efficient for them to win.
 Adopting an opposite strategy increases the
chance of winning a more ambitious prize,
but also the risk of walking away with nothing.
86

While no format appears to be perfect, policy makers
need to consider the problems associated with each
87
Vulnerability of auction format CCA SMRA First-price
sealed bid
Inefficient allocations due to budget constraints in
conjunction with lack of visibility over price
exposure
Material price disparities, bidders end up paying
different amounts per MHz
Risk of predatory bidding strategies designed to
raise the costs incurred rivals
Inefficiencies caused by difficulties in aggregating
optimal spectrum packages
Economic inefficiencies caused by demand
moderation strategies
Costly to execute, difficult to understand lack of
transparency

Best practice in choice of auction format
 First price sealed-bid auctions are the
most distorting type of auction and should
be avoided.
 Second price sealed-bid auctions are
better and retain simplicity.
 Opinions will differ on whether CCA or
SMRA is more prone to distortions overall.
– The relative levels risks will also
depend on circumstances as well as the
detailed rules.
 Regulators should examine their national
circumstances and reflect feedback from
operators in their ultimate choice of
auction mechanism.
88

Auction rules may matter more than auction formats
Rules are set to prevent gaming and
vexatious bidding while ensuring that
all spectrum is sold efficiently
 Spectrum packaging
 Spectrum caps
 Spectrum set-aside
 Activity rules
 Provision of information
 Bid increments
 Spectrum trading
 Reserve prices
89

How do mobile operators value
spectrum?
A brief overview of the methodology
operators use to value spectrum in order
to set auction bid limits
90

Spectrum valuation involves comparing an operator’s
enterprise value with and without the spectrum
91
Approach to Valuing Mobile Spectrum
Net Present Value with
NPV without
Value
Business value
with the new
spectrum
Business value
without the new
spectrum
Value of the new
spectrum

Relative difference in spectrum between operators impact
on the ability to serve customers and hence revenue
“with new spectrum”
competitive advantage
“without new spectrum”
competitive disadvantage More Spectrum
92
Reference case
Time
Customers / Revenue

Valuations are based on Discounted Cash Flow (DCF)
analysis
Need to include
spectrum-licence
93
Free Cash Flow Forecast
NPV of
10/15 year
forecast
Terminal
Value
Enterprise
Value
Valuation
$ renewal costs!
Past
$

Sources of spectrum value may include the following
94
Capacity
Access speed claim
Blocking value
Geographic & in-building coverage
Impact on other business
Avoid additional site build
Technology migration

Key operational drivers of spectrum value
95
Commercial drivers Network drivers
Capacity &
Blocking value
Coverage footprint
& Network
densification
needs
Spectrum package
Performance
(2x20MHz versus
Traffic
Indoor coverage
quality
(lower bands)
Costs of sale & overheads
2x10MHz)
Commercial impact
Network impact
Market share Network
Revenues capex & opex
Costs associated with freeing up spectrum / migrating customers need to be taken into
account in scenarios where existing spectrum is lost
Successful Spectrum Auction Design - ©
Copyright Coleago 2014 -
Spectrum roll-out
capex & opex
Network impact

Congestion puts customers and revenues at risk
96
Mbps per
site
Maximum capacity per
baseline site (including
max densification uplift)
Baseline sites
(ranked by peak demand)
Unmet demand
leads to loss of
customers and
revenues if
there’s a better
alternative
elsewhere
Unconstrained peak
demand per site
Constrained peak
demand per site
% of sites
overloaded without
spectrum package
If / where it is not
possible to build
oneself out of
congestion…
Mobile networks suffer from ‘Pareto’ effect: a minority of sites tends to carry a
disproportionate amount of traffic, with the bulk of sites under-utilised

Extra capacity can drive market share
97
% of sites
overloaded without
spectrum package
Mbps per
site
If / where it is not
possible to build
oneself out of
congestion…
Maximum capacity per baseline
site with spectrum package
Baseline sites
(ranked by peak demand)
Unconstrained peak
demand per site
Constrained peak
demand per site
Without
spectrum
package
Extra demand
addressable with
spectrum package
Increase in
spectrum
Need to take an industry-wide view: consider congestion levels versus any spare
capacity across all operators

Indoor coverage quality also drives revenue
 The superior propagation characteristics of low band spectrum yield a higher
probability of (deep) in-building service at a given performance
 This may be translated in terms of a % of demand that is only addressable with low
band spectrum (or with potentially expensive in-building solutions)
98
Probability of
service
With spectrum
below 1GHz
With spectrum
above 1GHz
Deep indoor % Demand
Offer LTE
performance
deep indoor
with 700MHz

www.coleago.com
Stefan Zehle, MBA
CEO, Coleago Consulting Ltd
Tel: +44 7974 356 258 stefan.zehle@coleago.com

Best practice spectrum auctions workshop, New Delhi 24 Sep 2014

More Related Content

What's hot (20)

Similar to Best practice spectrum auctions workshop, New Delhi 24 Sep 2014 (20)

More from Coleago Consulting (8)

Recently uploaded (10)

Best practice spectrum auctions workshop, New Delhi 24 Sep 2014