UMTS: DTI Consultation Document
Multimedia Communications on the Move
Comments and Proposals of Scientific Generics Ltd
16 October 1997
Scientific Generics Limited
Cambridge CB2 5NH
A member of The Generics Group plc
Scientific Generics is a major European technology and business consulting
organisation very much concerned with the exploitation of technology and the
process of innovation. It offers the following comments in the interests of
stimulating the level of competition and innovation in services which potentially
cover a wide range of fixed and mobile applications.
1. SUMMARY AND RECOMMENDATIONS .............................................................. 1
2. INTRODUCTION ......................................................................................................... 3
3. EVOLUTION OF UMTS .............................................................................................. 4
3.1 Which paradigm applies? ................................................................................... 4
3.2 First and Second Generation Service Providers failed. They have a role
this time. ................................................................................................................ 6
3.3 Use a Railtrack model: “Radiotrack”? ............................................................... 7
4. SPECTRUM PRICING .................................................................................................. 9
4.1 Competition, Pricing, and Economic Surpluses .............................................. 9
4.2 Maximising Consumer Surplus........................................................................ 11
4.3 Uncertain, Derived Demand ............................................................................. 12
4.4 What have spectrum auctions to do with efficient use of spectrum? ......... 12
5. CAN THE RADIO TRACK MODEL OFFER ADVANTAGES OVER
MANAGED OLIGOPOLY? .................................................................................... 15
5.1 More Economics ................................................................................................. 15
5.2 Proposed Arrangements for competition and spectrum pricing and
efficiency ............................................................................................................. 16
5.3 Proposed Spectrum Pricing .............................................................................. 17
6. WHAT HAPPENS IF UMTS DEMAND IS PREDICTABLE?............................... 20
1. SUMMARY AND RECOMMENDATIONS
Scientific Generics’ principal comment on the DTI paper is to recommend that
there should be a single, Railtrack–style, network operator and a very open
market in Service Provision. UMTS Service Providers should be offered more of
the value chain than under first and second generation services and should also
get access to second generation networks at an equivalent level in the value
pyramid. Economic rent for spectrum should be charged so as the clear the
market for spectrum rather than to maximise government revenue and should
also be priced and charged as close to real time as possible. A succession of short
licence periods for Service Providers is recommended as the means for doing
More Specific Recommendations are:
The single network operator should be licensed under a succession of five
year BOT contracts, the main selection criterion being the contractual price
for a unit of spectrum load.
The network operator will have no power to issue network addresses and
will consequently have no HLR (Home Location Register).
The network operator and his controlling shareholders may not have
interests in Service providers which in aggregate control more than, say, 15%
of network operator revenues
The network operator will be required to make a minimum investment in
infrastructure (but not one implying more than majority coverage of the UK)
and to comply with ETSI standards. He will be under no other obligations as
regards services supported or coverage.
All available spectrum should be made available to the network operator
from the start of operations.
Service providers (SPs) should bid for licenses as described in Section 4.3
below in the case that the government’s objective is to clear the market for
spectrum. Failing this, the alternative arrangements aimed at maximising
government revenue should be adopted.
SPs should be pre-qualified on a minimum financial commitment and
backing and proof that they have the capability to issue network addresses
and interconnect with the network operator.
All interconnections with the network operator shall be on an equal basis,
meaning equal terms and prices for equivalent network load, with no
preference for high volume SPs.
Existing GSM/DCS operators should be required to allow all licensed UMTS
SPs to interconnect to their networks on an equal, most favoured partner
basis (e.g. on terms at least as good as the GSM operator’s own UMTS
service). The connection would be similar to that of an existing GSM roaming
partner. Should there be problems with achieving this, those existing
operators refusing to co-operate should be denied the right to any interest in
UMTS service provision.
Any additional interconnection between the UMTS network and existing
GSM/DCS networks requested by the UMTS network operator (e.g. to allow
call handover) should be supported by the existing networks.
The Radiocommunications Agency would have the final say in determining
the formulae for establishing equivalent network load (which is used for
pricing and measuring total network demand). While the process for
establishing these should be open, care should be taken to prevent this
becoming a matter for litigation.
In the case of spectrum being allocated per the DTI Scenario 2, first, the
deregulated use of 1900-20 MHz is supported but because of possible
radiation problems in the down-link band, shared use might be best confined
to the up-link; second, the reversal of duplex separation for the benefit of
asymmetric traffic is strongly supported.
Besides some issues of spectrum planning, The DTI paper seems concerned to
address two objectives:
Providing ‘licensing certainty’ for Third Generation Mobile (subsequently
referred to here as UMTS) in a way that will provide a competitive and
stimulating environment for the development of multimedia services. This
will be done within the framework of the relevant ETSI standards and of
allocation of frequency bands specified in the consultative paper.
The introduction of spectrum auctions for the initial allocations of UMTS
spectrum as the first instance of a better method of spectrum pricing than the
existing system of administrative pricing.
The two objectives are, of course, linked by the desire to make the most
productive use of a relatively scarce resource - or, at least, an irreversible decision
on spectrum usage.
In the commentary offered below, it is been thought helpful first to consider the
drivers of and barriers to use of UMTS. This informs our view of the
appropriateness of the DTI’s proposals for spectrum licensing and leads us to
suggest an alternative and much more open approach to spectrum usage.
This is then complemented with a more general discussion of spectrum pricing
and suggestions for pricing which could be applied to UMTS and conceivably
could be used as a paradigm for all spectrum pricing and allocation.
EVOLUTION OF UMTS
3.1 Which paradigm applies?
“GSM takes over from TACS”, or
“ISDN TAs fight modems”
Much of the DTI paper gives the impression that UMTS will take over from
GSM/ DCS in the same manner that GSM is taking over from the analogue
networks. This is in spite of the fact that it is recognised that substantially new
applications will be enabled by UMTS and that it provides the first infrastructure
intended to cover both fixed and mobile usage.
GSM - as a replacement to analogue - offers those who want voice services:
Better bandwidth utilisation, which ultimately affects price and overcomes
Better paging/forced alerts
It is not clear that UMTS improves on any of these very much:
If the bandwidth utilisation improvement is only of the order offered by
CDMA IS95 vs. GSM, it is not very marked.
GSM and DCS 1800 licensees in the UK as a group have sufficient bandwidth
for mobile voice traffic for some time to come: enough to pursue “wirefree”
strategies of substituting for both voice and data fixed communications. The
paper also considers re-farming GSM bandwidth for UMTS - another
indicator that it is not pressure on spectrum that warrants a move to UMTS.
Second generation operators will have partially depreciated network assets
when UMTS is launched and could well face a lower charge for bandwidth
than UMTS providers will (if the economic rent of the spectrum/operating
license for UMTS is maximised via auction).
In short, a UMTS operator in competition with existing networks for voice traffic
has no significant benefit to offer, yet he has handicaps in the form of a greater
financing burden and more costly and bulkier handsets.
The service - where available - would offer great potential. But why would users
migrate? It is like migrating to ISDN. Those who only use voice and G3 fax have
no reason to do so.
The service will come into its own - and drive migration - through high
bandwidth applications. (An EU funded study of demand points to large file,
audio clip, and video clip downloading as the major revenue potential.) Such
applications may not be much in demand when users are driving along
motorways but may be more in demand as a fixed line substitute and in specific
locations, such as hotels and airports.
This view of service evolution - but more generally the lack of obvious continuity
from GSM to UMTS service applications - argues against presuming that UMTS
is above all the next generation mobile service for which very similar conditions
to those for existing operators should apply: For example:
requirements regarding geographic coverage, and
the expectation that competition among a few operators is appropriate,
preserving continuity with the present GSM/DCS arrangements as traffic is
naturally siphoned over from these services to UMTS.
With GSM operators looking to provide packet data services and at least 64 kbps
data circuits, there is certainly an aspect of UMTS that looks like that of ISDN
competing with lower cost analogue lines and continuing improvements in
At the time of writing it is not yet clear how different from 900 and 1800 MHz
GSM the ETSI standards for UMTS will be. Anyone offering mobile UMTS
services in the first few years of service will have to offer dual capability
GSM/UMTS phones and the cost, size, and weight penalty of supporting UMTS
(over and above vanilla GSM) will depend on how big a change ETSI standards
herald. (A Service Provider intending to offer only limited mobility data services
via UMTS could, however, offer a separate UMTS data terminal with a GSM
handset; existing GSM services being used for voice coverage and perhaps low
data rate wider mobility.)
We take first the case that ETSI introduces a new (to Europe) air interface
(probably CDMA) in order both to support a truly world standard and to
introduce an air interface better suited than GSM to high bandwidth data
applications. In Section 5 below we consider the implications of UMTS service
being a largely predictable extension of GSM services, a scenario which may also
be associated with UMTS being implemented as GSM in new frequency bands.
In the first case there would be a considerable cost and perhaps portability
penalty to be overcome to attract users to the new service. Does this mean that
the service is not needed and what steps can one take to give it the best chance of
proving successful with end-users?
The short answer to whether the service is needed (in 2002) is that we do not
know. There is a fair prospect of Web surfing and file retrieval being the main
demand drivers; but it is probably more helpful to assume that demand in 5-10
years time is, in the detail required seriously to design an appropriate
competitive framework for it, unknowable. We believe that the best way of using
the UMTS opportunity is to encourage the birth and demise of many new
services through maintaining a highly competitive service provision
environment and low costs of market entry - fostering the kind of creativity we
have seen with the Internet.
In sum, the role of service providers needs to be revisited.
3.2 First and Second Generation Service Providers failed. They
have a role this time.
The UK government tried to ensure widespread competition with first and
second generation mobile networks (prior to the PCNs) by establishing a two tier
structure of network operators and service providers. In broad terms this has
failed, as the network operators have in effect controlled functionality and end
user pricing and - ultimately - the service providers themselves.
One reason SPs failed first time round was that all the useful functionality was in
the network operator domain and each SP was in a relatively weak position to
prise functionality from the operator. Even now SPs generally do not even
control the voice mail platform of their clients. In retrospect, the take-over of the
SPs by the operators seems inevitable.
If a SP were to design a new service for a target market, he would like as much
control over the underlying technical capacity of the system as possible. If he has
to deal with service packages in the hands of an oligopolistic network operator,
then his value added is severely constrained. If a similar division of
responsibilities between SPs and network operators applies to UMTS, then it is
hard to countenance an outcome other than the existing GSM/DCS operators
acquiring the UMTS licenses and using them for an oligopolistically managed
competition both between themselves and with the fixed networks in which the
same parties are likely to have a vested interest too.
But the failure of SPs to-date is not a reason for accepting an oligopoly of a few
network operators for UMTS. It is possible to create an environment in which
there is widespread competition between SPs or at least in which costs of entry
are kept relatively low.
3.3 Use a Railtrack model: “Radiotrack”?
The answer is to have a single utility providing the very basic network, under a
controlled pricing regime and open to all, and to create a very free market in
service provision and spectrum usage.
With a single utility there is no duplication of investments, which hurt the
economics of network deployment and hence delays the roll out of coverage.
By the same token there is no need to set-up up interconnection agreements
between several UMTS networks in order to avoid a Telepoint-like lack of
coverage. All licensed SPs would have rights to interconnect on equal terms.
The single network aids conformity in interface standards and terminals and
also minimises the environmental impact of the network deployment. There
are also predicted to be spectrum efficiency benefits from using all the
available spectrum as one unit.
The Network Operator is selected by auction to offer the lowest tariff to SPs
under a Build, Operate, and Transfer (BOT) contract, whereby network assets
are compulsorily transferred at book value to the succeeding contractor for
the next period. The contract period might be five years.
SPs own the telephone numbers (subject to customer number portability). In
GSM terms, they have the HLR and their customers ‘roam’ onto both the
UMTS and GSM networks. Radiotrack therefore only supports ‘visitors’.
Licensed SPs will also obtain the same right to interconnect with existing
GSM /DCS operators.
The tariff bid by Radiotrack bidders could be per Mbyte for each class of
service with the relative price of each service class equal to an agreed ratio of
network load per Mbyte for the service. Any new service class would thus be
automatically priced dependent on its relative network load.
Resubmitting the Radiotrack contract to auction every five years allows for
correction for outcomes departing from expectations and prevents the operator
developing an historic incumbent culture.
Strong competition among SPs would be encouraged in the following ways:
Bids would be invited for an unlimited number of SP licences which entitle
users to interconnection with both Radiotrack and the GSM/DCS networks
under regulated roaming agreements for each service class. The regulation is
intended to ensure the right to interconnect on the same terms as anyone else.
Note that the GSM interconnect rights are needed inter alia to prevent undue
advantage lying with existing network operators.
The fact that SPs do not carry the investment on the network infrastructure
means that very short license periods can be used. Certainly not more than 5
years, perhaps as little as 1 year. One might think in terms of 30 months.
(This also allows spectrum to be repriced as uncertainty peels away and
demand pressure builds up.)
Competition would be supported by limiting the proportion of Radiotrack’s
total revenue from SPs that is controlled by the same interests to, say, 20%.
For both Radiotrack and SPs, a licence issued in one period confers no rights
or preference as regards licences in later periods.
The value of having many more SPs is:
There would then be a real prospect of competition at the service level - as
distinct from the essentially oligopolistic environment that would follow
from a simple extension of existing arrangements.
The high level of competition will be reflected not only in better prices for
users but also will be crucial to the stimulation of new services on which the
successful exploitation of UMTS depends.
Separating the more basic network functionality from the customer specific
ones (which are in the competitive SP domain) one can have the advantages
of a single, common network infrastructure without imperilling the
competitiveness of service provision.
The DTI paper proposes to auction spectrum for the first time in connection with
the initial UMTS operator licenses. This is against an historic background of
administrative pricing which was originally intended only to cover Radio
Communications Agency administrative costs, but which will be increased in the
case of existing GSM and DCS spectrum in areas that are defined as ‘congested’
in order to encourage network investment in bandwidth conserving measures.
Auctioning bandwidth (for the highest price) implies trying to maximise the
economic rent to the government of the spectrum or, in practical terms, of the
conditional licenses that are the government’s to confer. This is obviously a
departure from previous practice and the broad economic consequences of such
a policy are worth review.
4.1 Competition, Pricing, and Economic Surpluses
Figure 1 depicts the traditional demand curve (dD) against (for simplicity’s sake)
a horizontal cost curve (which implies that marginal costs equal average cost).
Under perfect competition the price in the market, which no individual supplier
can influence, is 'A' and the area dAc represents the consumer surplus - the
excess of consumer value over price paid.
A monopolist in the same market who could control the price paid by each
consumer for each item used would supply the same quantity as under perfect
competition, but in this case dAc would all accrue to the supplier as producer’s
surplus and each consumer would pay a price equal to the full value to him or
her of each item consumed.
One could say that in principle the government should be indifferent as between
these two extremes: the resource in question has been used to the same extent in
both cases - up to the point where the marginal value to society equalled the
marginal cost. However, in the case that consumers are overwhelmingly
nationals and producers could represent foreign interests, then there could be a
presumption in favour either of maximising the consumer surplus or of taking
the surplus into government.
Figure 2 shows the case of a monopolist who can control the price but cannot
price items differently by quantities taken or by client; there is but the one price.
In this case, winning marginal customers with a lowered price reduces revenues
from all existing customers, so that the net marginal revenue is lower than the
price and can easily be negative. The marginal revenue curve is “da”.
The producer will then equate marginal revenue with marginal costs, at point
“a”, which implies a price “P”. The consumer surplus is then dPp; the producer
surplus is pPac, and the area PaA is value that is lost to society altogether.
If some competition is offered to the monopolist, then each supplier is tempted to
lower prices more than in a monopoly situation, because he is then able to attract
existing users over from other suppliers as well as encourage more usage. Prices
will therefore fall to the extent that the producers try to win market share from
each other via pricing. Figure 3 represents such a situation with the price at “R”
and rRbc, rRd, and RAb representing producer surplus, consumer surplus, and
unexploited value respectively.
When the government auctions spectrum (de facto rights to operate) to a few
operators, one could judge that it has accepted the fact of an oligopolistic market
consistent with price “R” in Figure 3 and is basically aiming to capture the
producer's surplus, rRbc, for itself, allowing operators to make normal levels of
profit but not to gain a windfall from the rights they have obtained.
There are three points to note about this:
It clearly does not maximise consumer surplus and leaves some potentially
valuable resource unexploited
It has problems, well rehearsed with American PCS licence auctions, that the
demand evidenced in the auctions (by potential service suppliers) is a
demand derived from expectations of future end-user demand. These
expectations can be very wrong, and this affects the merits of charging for
spectrum in this way.
It has nothing, per se, to do with ensuring efficiency in spectrum usage,
which has been the stated purpose of increased administrative pricing of
We take these in turn.
4.2 Maximising Consumer Surplus
To drive the price down to “A” in Figure 3, one could aim to ensure a highly
competitive market with many suppliers. It is judged that this is not possible
the limited amount of spectrum in relation to the minimum foreseen for an
effective network (e.g. 2x20 MHz);
the disincentive to invest in networks if the potential demand is fragmented
among too many suppliers;
the need to limit the environmental impact of network deployment.
These considerations revolve around economies of scale which affect the
network infrastructure strongly but have much less effect on the value-adding
functions of service providers. A Radiotrack model of single network operator
who is not given rights to control prices, paired with an overlay of strong
competition between many service providers, would come closer to maximising
consumer surplus than an oligopoly of vertically integrated network
4.3 Uncertain, Derived Demand
Initial spectrum allocations, including, one assumes, those anticipated with
UMTS auctions, are long term commitments by government, intended to be
matched by long term commitments from network operators who carry the risk
of the potential value of a demand which is to a greater or lesser extent unproven
and is far into the future.
The potential for error is great. If companies bid too much for licences, then they
are hamstrung by the financial burden of the spectrum price. Either they do not
perform or the government ends up absorbing the risk. If companies bid too
little, the government misses out on the real economic rent of the initial
allocation. It may be able to get nearer its true value in subsequent spectrum
allocations, but it will be hampered by the need to prevent undue concentration
of market power in the industry and hence to keep new spectrum allocations
more or less in balance with historic ones.
There is no need to take bets on future licence values in this way. If one could
devise a “pay as you use “ type of pricing, then the information which informs
the price will be largely related to established demand rather than to forecast
4.4 What have spectrum auctions to do with efficient use of
First let us consider what is happening with administrative pricing of spectrum
intended to ensure efficient use of spectrum. The general principle involved is to
charge, for new allocations of spectrum, more or less the same as it would cost to
achieve the same increase in capacity by investment in more infrastructure (e.g.
micro cells) instead. This should ensure that where extra capacity is readily
obtained by some extra network components using existing spectrum, then that
is what operators do. And when extra capacity would be very expensive in terms
of infrastructure, then operators will opt to purchase extra spectrum.
It may sound as if this is encouraging efficient use of spectrum but, of course, as
long as spectrum is not used at all – as it is if it is waiting in abeyance pending
eventual purchase – then it is not being used efficiently in an economic sense. If
all of a frequency band is being set aside for use with, say, GSM networks, then
that band will be put to best use if all of it is available to GSM operators. Since it
is a limited resource, the operators will naturally invest in network elements that
will enhance capacity as and when needed. It is, in fact, an inefficient use of
resources to force operators into such investment ahead of a real shortage of
Why, then, is this new principle for administrative pricing of spectrum being
introduced? It is actually to do with controlling the nature of competition. If all
the spectrum that would ever be available for, say, GSM and DCS is apportioned
between the operators at the start of their operations, then that apportionment
plays a large part in determining the market shares that each will achieve. If one
operator is very successful and runs demand up against his spectrum limits, then
extra capacity at the margin for him will be more costly than for someone who
still has lots of spectrum to spare. The initially less successful operator can then
start undercutting the successful one and redress his poorer market share.
To check this kind of behaviour, the government holds back some of the
spectrum and is prepared to allocate new tranches to operators if they can show
that – in terms of some industry norm for capacity enhancement – they have
done all that is reasonable to win capacity from their existing spectrum
allocation. This stops one operator gobbling up spectrum allocations for purely
strategic purposes; and it allows more successful operators to win a greater share
of the spectrum than their initial portion – although still presumably subject to
some over-riding constraint intended to prevent dominance of the market by any
Taken to its logical conclusion, this approach would imply taking spectrum away
from operators who were not using it to the level of efficiency of those operators
facing the greatest demand pressure. This would ensure that marginal costs of
capacity were the same for all players. However, it is one of the features of the
existing regimes for operators that spectrum allocations are more or less
irrevocable (long term investment decisions being predicated on them).
In summary, then, we can say that the new arrangements on administrative
pricing of spectrum (and, indeed, the technical efficiency benchmarks that were
relied upon earlier) are intended not to ensure efficient use of spectrum from a
national view point but rather as a means of keeping competition between
How does the introduction of spectrum auctions compare in its operation?
Clearly, if all available spectrum were to be auctioned in one go to a few
this imposes constraints on changes in market shares, which may curb
market dominance by one party but encourages cartel-like behaviour; and
spectrum efficiency is left entirely to the operators; spectrum unused or
poorly used by one operator is not available to another.
If initially some spectrum is auctioned and some retained for further allocation,
then one should distinguish between the case where there is enough spare
spectrum to expect several new entrants to the market and the case where the
most probable outcome is that the first operators bid for extra capacity as
demand picks up. In the latter case (which seems the better fit with UMTS
allocations) we are back to the same kinds of considerations we have rehearsed
with administrative prices, for example:
What should trigger auctions? Measures of congestion?
Should there be a reserve price? The new-style administrative price?
What stops the most profitable operator buying capacity for purely strategic
If anything, administrative pricing would seem the better way of ensuring
equitable competition through the allocation of spectrum tranches additional to
the initial ones. The initial tranches themselves are different. The main thing that
an operator buys with the initial allocation is one of a very limited number of
operator licenses for a long period of time, together with enough initial spectrum
for a national operation for some years and some implicit protection from
domination of the market by any other operator. In effect, it is the producer’s
surplus under oligopolistic conditions that is on offer, and, as discussed above,
there are good reasons for the government to try to capture as much of this as
possible. In principle auctions are a good market clearing device but there are, as
also noted above, problems in trying to take all this value in one up-front
payment before demand has established itself and without a clear view of
competition from alternative technologies.
CAN THE RADIO TRACK MODEL OFFER
ADVANTAGES OVER MANAGED OLIGOPOLY?
We have noted above how a single network operator under suitable controls for
both pricing and value added can provide the platform for very strong
competition among service providers, with the concomitant benefit of
stimulating innovative services and maximising consumer surplus. Below we:
develop a view of suitable controls on both the operator and the service
providers to achieve this;
attempt to deal with the problem of how to extract economic rent more in
real time and not all up-front; and
Show how spectrum is used more efficiently than under the oligopolistic
Before getting into details, there are again some general economic points to
5.1 More Economics
First, if we achieve the objective of a single network operator under a cost plus
normal profit regime and strong competition among service providers, then - as
long as the supply cost curve is not rising - there is no producer’s surplus; i.e.
there is no economic rent which can accrue to the spectrum owner. A surplus
only materialises when demand for bandwidth exceeds supply, as is illustrated
by Figure 4.
If the initial demand curve is dD and is insufficient to use up all the bandwidth
available, then strong competition pushes prices down and usage up until prices
equal costs of production (incl. “normal” profits). If spectrum is limited (i.e. costs
become infinite at a certain point) and demand grows over time to d’D’, then the
market clearing price will rise above costs and an economic rent or tax on
spectrum of (OP-Oc) can be raised.
5.2 Proposed Arrangements for competition and spectrum
pricing and efficiency
We have described above the terms on which the network operator would
receive a license, including two in particular:
The fact that he would be bidding to offer a fixed (or even falling) price and,
other things being acceptable, would be selected on the basis of having the
lowest such price. The price regime would therefore be a fundamental part of
his contract. In spite of being the only supplier, he does not have the
monopolist’s ability to control prices. He should also be committed to
investing some minimum amount in the network. He therefore has an
incentive to support, and even stimulate, as much traffic as can be serviced at
a marginal cost less than the agreed price.
He has a contract for only five years. The initial investment required of him
must not be so high that he is unable to make a profit in this time. He will, of
course, receive back the book value of his investment. One might also
stipulate that his successor would take on the liabilities associated with
employees. In effect, the successor takes on the network as a going concern,
and its value to the outgoing contractor is not reduced by redundancy
liabilities etc., that would arise if the business had to be closed down at the
end of the contract. The five yearly contract rebidding does not require that
the contractor change every five years. In fact changes will probably be the
exception and the contractor for the first period may even gamble on making
most of his profits beyond his first contract period. Nevertheless, the process
plays an important part on keeping him on his toes and giving those who
think they can do a better job a chance to get the business – with the users,
not the owners of the inefficient business, getting the benefit of the change.
We believe that these arrangements offer sufficient controls over the single
network operator to ensure that he is efficient and responsive to user and SP
needs; and that he is not in a position either initially or later to extract monopoly
profits from the uniqueness of his status. He is also not permitted to have a large
enough stake in the SP market to influence that market.
Under these arrangements any economic rent obtained for the government is
paid by SPs and not by the network operator. We have also described above how
competition among SPs is achieved, including the stipulation that no group can
control more than say, 20%, of the network operators revenues. SPs would also,
we noted, have even shorter contract periods than the network operator – we
have suggested 30 months. The process of granting many SP licenses and putting
them all up for reselection at frequent intervals should ensure about as
competitive an environment as you can hope to get in a domain where there are
clear economies of scale.
It also has the useful feature of allowing the government to adjust the economic
rent for spectrum at regular intervals throughout the lifetime of this
technology:spectrum combination. The shorter the contract period for SPs the
more accurately can government track the real economic value of spectrum.
Clearly such an approach works much less well if one tries to extract the value
from network operators, who need longer contract periods in order to justify
their investment. This approach therefore has two huge advantages:
The tracking of the spectrum economic value gives much better economic
signals than up front auctions – better even than the new form of
administrative pricing. Spectrum is probably free at the start of operations,
when it is plentiful in relation to demand. When it comes under demand
pressure, then all of it (not just new tranches) is available at a market clearing
There is very much less uncertainty in the pricing of spectrum than in up-
front auctions. Spectrum is priced only for the short term, which to a
significant degree is predictable – certainly much more so than values which
have to be discounted over 15-20 years into the future. This greater certainty
reduces the various costs of risk, which can range from too much timidity to
the crippling effects of excessive financing burdens on service deployment. It
is also more equitable.
5.3 Proposed Spectrum Pricing
How then is the government to price the spectrum? It does so via the SP licenses.
On the assumption that government seeks to maximise consumer surplus, it will
invite bids for SP licenses with a minimum of prequalification (e.g. evidence of
probity, a minimum of financial commitment and support, and access to a HLR).
Each SP bids to pay a level of spectrum tax (which can be zero) and commits to
pay (on a DCF basis) a minimum amount over the license period. Government
will, in fact, only charge a tax in excess of its administrative costs in the event
that demand at a zero tax exceeds the spectrum available.
Because demand for spectrum does not reveal itself until the bids are in - and
because each SP’s bid will be dependent upon market assumptions, which are
liable to change as other bids are revealed - it would not be appropriate for
government to cash in the first round bids. Two or three rounds of bidding are
probably needed. After the first round, the Government will publish (as a
minimum) the amount of spectrum bid for in aggregate and the price (tax) which
would have cut off all the demand in excess of supply. The subsequent rounds
should result in a convergence of bid prices close to a market clearing rate. In
order to keep the individual bids serious, SPs will be held to their bids in the
final round, but strong convergence in prices is likely to result in most of the
economic surplus over and above a single market clearing tax rate accruing to
In the case that the bid price is zero, SPs must state the maximum amount of
bandwidth they will need at any time during the licence period, on the
understanding that an auction for spectrum will be immediately invoked as soon
as any SP reaches his maximum. This condition is imposed to prevent SPs
bidding a zero tax and claiming very low bandwidth needs ex-ante and
exceeding them ex-post. The condition might be adjusted when there is a tax on
spectrum to invoke subsequent bids when the actual demand exceeds the
aggregate maximum demand bid.
If government aims to maximise its economic rent (at the expense of consumer
surplus), then it does not clear the market by matching demand with all the
spectrum available. Instead it looks to maximise its tax revenue from guaranteed
payments at bid tax rates. Any zero bid therefore immediately disqualifies itself.
After the first round of bidding, government would assess what is the lowest tax
rate it would have accepted, consistent with maximising its revenues, on the
premise that it would have to apply the same rate to all. (This premise in effect
anticipates the convergence of bids induced by the multi-round process.) The
government’s calculation is therefore that every reduction in this rate reduces the
revenue from higher bidding SPs, while increasing the aggregate demand by
capturing more lower priced bids. It would publish this rate along with at least a
measure of the total bandwidth demanded, and perhaps the implied bandwidth
commitments of each SP as well.
As in the first case, the individual bids in the final round would be binding.
The latter arrangements may provide less scope for abuse of the process by
bidders than that for clearing the market for all spectrum. It shares much the
same benefits of taking frequent measures over time of the value of spectrum.
But it is clear that it will discourage productive uses of spectrum and is
particularly distorting in the early years, when the case for taxing spectrum at all
is very weak and the new service, with strong competition from GSM/DCS,
could well do without this burden.
In its general effect, maximising government’s economic rent in this way is
similar to auctioning spectrum to a few network operators: some productive
spectrum goes unused and some consumers’ surplus is lost. However, one
would expect a wide range of highly competitive SPs to create a higher level of
demand in total than three operators, each with its own (probably rather similar)
idea of where it should be “focused”. One can therefore expect this more
competitive model for service evolution to take the price point (and the use of
spectrum) further down the demand curve. When the alternative (and
recommended) objective of clearing the market is adopted, then there is still
better use of spectrum.
We have only to note the point made in the industry that spectrum efficiency
would be improved if all operators had access to all the spectrum to cap our case
that the combination of the proposed spectrum allocation procedures and the
creation of a Radiotrack–style single network operator will result in more
productive use of spectrum than creating an oligopoly of vertically integrated
network operators via up front auctions for a few operating licenses.
WHAT HAPPENS IF UMTS DEMAND IS
To a degree, the arguments for the proposals we have made have depended
upon the demand for UMTS services being both to a large extent unpredictable
and very diverse. We promised to consider the implications of demand turning
out to be instead a relatively narrow extension of what GSM offers today – which
might also be associated with a minimal change of GSM technology to allow it to
operate in the new frequency band. One could characterise this demand as that
for higher speed data access to support Web browsing and file downloads to PCs
One can imagine the evolution of such service taking one of two forms. In the
first, demand for higher speed data access develops relatively slowly and users
are not so impressed by the price:performance characteristics of GSM data
services as to use them heavily. In this case, there would be perceived to be a
weak demand for data services of even higher speeds; investment in UMTS will
be conservative; and the migration to UMTS will be driven by existing
GSM/DCS operators who simply want more bandwidth.
The second case is simply the reverse. Users clearly want a lot more data
bandwidth for mobile/wireless access and are quick to leave GSM behind as
soon as something better is available. The more limited mobility at first offered
by UMTS is not in practical terms very limiting. In this case it is also conceivable
that there will be considerable demand for UMTS as a pure data service with
continued reliance on GSM for voice. The interconnection of the GSM and UMTS
worlds at a high level of functionality ( e.g. to allow SMS messages to be sent to
either network) is important.
For each of these scenarios we can ask whether consumers would be better
served by our proposal or by allowing only a few network operators, likely to be
strongly vertically integrated over time and also with strong representation from
existing GSM/DCS operators.
For the first, slow growth scenario, it is surely self-evident that our proposals
would provide much more stimulus to demand and are to be preferred for that
reason. This stimulus comes from:
the greater strength of service providers vis-à-vis network operators, giving
them both more incentive and more ability to fashion service for different
the greater level of competition implied by the large number of suppliers;
the freedom from the initial financial burden of up-front license payments
that are not only the wrong economic signal at that point in the technology
life cycle but also create inequity between the new services and established
In the second case, one could expect a more lively scene for UMTS, at least if one
or more new entrants gains a UMTS license and starts using its attractions to win
high usage customers away from GSM/DCS. This would also happen under our
proposals – only more so. The level of competition and service innovation will be
higher, yet without a diluting effect upon investment in network infrastructure.
By contrast, network coverage under our proposals should be in advance of that
under a multi-operator regime and there are also higher levels of spectrum
efficiency achievable under a single network, resulting in better service quality
for users. Finally, the spectrum/licence pricing regime results in better economic
signals for usage throughout the technology life cycle than an up-front auction.
16 October 1997
Contact at Scientific Generics:
Tel: +44 1223 875200
Fax: +44 1223 875201