(Premajhen) Zaslužek iz zraka
Univerza v Ljubljani
Kardeljeva ploščad 17
Tel: 01 589 2400
Fax: 01 589 2698
Making (not enough) money from thin air
Faculty of Economics
Kardeljeva ploščad 17
Tel: +386 1 589 2400
Fax: +386 1 589 2698
I am grateful to Damir Cibic for sharing his insight on the topic. I would also like to thank Brane Miklavčič and
Urša Manček from Mobitel d.d. and Darko Pretnar and Ludvik Uhan from Iskra Tel d.d. for their help and
perspectives on the matter.
V članku analiziramo javni razpis za podelitev druge licence za mobilno telefonijo GSM v
Sloveniji. Napovedi teorije dražb in izkušnje podobnih razpisov iz drugih držav nas prepričujejo,
da bi odprta, naraščajoča, angleška dražba prinesla večji izplen kot uporabljeno lepotno
tekmovanje. Istočasno bi tako lahko tudi izboljšali učinkovitost končne alokacije licence.
Namesto, da je skrbela za kriterije, katerim so ponudniki morali zadoščati, bi morala vlada
ponuditi več licenc in tako privabiti več ponudnikov, večjo konkurenco.
Ključne besede: Deregulacija, Učinkovitost, Telekomunikacije.
The tender for operating the second license for GSM mobile telephony in Slovenia is analyzed.
Using what auction theory teaches and what similar tenders held elsewhere in the World show it
is argued that using an open ascending English auction in place of a beauty contest may have
generated much higher revenue. At the same time the efficiency of a final allocation of the
license may have been promoted. Instead of worrying abut criteria bidders competing for one
license only must have fulfilled the government should have provided a higher number of
licenses and attracted a higher number of bidders.
Key words: Deregulation, Efficiency, Telecommunications.
The introduction of mobile telephony to Slovenia dates in 1991 when joint Slovenian-Croatian
analog Nordic Mobile Telephone (NMT) network was brought to life. The partnership broke a
year later and the Mobitel NMT system took over the Slovenian part. At that time Slovenia was
already lagging behind in the development of the digital mobile telephony system GSM.1 This
system offers much broader service and is spread all over the Europe. With this in mind the set
up of appropriate network was initiated and in 1995 Mobitel d.d. was awarded a license to
operate the GSM network using a 50 MHz frequency band, a half of available spectrum. Mobitel
is owned by Telekom Slovenia a mostly state owned provider of fixed telephony services and has
been a monopolist on the market until March 1999 when the second licensed operator Simobil
started to sell GSM services. Before we discuss the market mechanism used to assign the second
license let us briefly describe the state of the market prior to Simobil’s entrance.
I.1 State of the market
Growth of the number of Mobitel GSM customers is amazing, from zero on July 31st, 1996, to
200,000 in March, 1999. Mobitel GSM has therefore reached approximately one half of average
penetration rates of mobile telephony characteristic for West European countries in mid 1998.
These growth rates surprised even market researchers at Mobitel since they are continuously
revising their expectations about future growth upwards (Mobitel, 1997, and Mobitel, 1998).
Group Special Mobile at first, Global System for Mobile now. This system is a European
counterpart of US Personal Communication System (PCS).
Their forecasts from 1999 for the end of year 2000 were 295,000 GSM and 25,000 NMT
customers. What really happened in the aftermath of the Simobil's entrance were significant price
drops that spurred the demand and in the mid 2000 the total number of users of mobile telephony
services is over 900,000 with Mobitel serving around 90% of these.
Mobitel’s revenues went from $41 million in 1996 to $64 million in 1997 and profits from $6
million in 1996 over $11 million in 1997 to $22 million in 1998.2 These numbers show the
mobile telephony market is certainly a nice market to be in and Mobitel was very happy enjoying
the monopoly profits. This was about to end in 1999 since the government decided to auction off
the other half of available frequency range within 890-915 MHz and 935-960 MHz bands, to the
second GSM provider.
II. The selling mechanism
It is by now a well known fact that the optimally chosen auction is the best mechanism to
promote the efficient allocations when selling goods in various contexts one of which seems to
be the governments selling off the spectrum rights. There are success stories like the notorious
US FCC spectrum auctions (Cramton, 1997) and stories showing that not thinking things through
may prove costly like those of similar auctions in New Zealand in 1988 and Australia in 1993
(Mueller, 1991, and McMillan, 1994).
In 1999 dollars.
II.1 Rules and the outcome
The government decided to set up the following mechanism (Ministry for Traffic and
Communications of Republic of Slovenia, 1997) to sell one license (in addition to existing
Mobitel's one) for operating a GSM network. Every bidder had to provide the warranty stating its
seriousness about the offer in terms of $110,000 toward the failure to sign a contract if awarded a
license and $1.1 million toward the failure to comply with the terms of the contract after a
specified amount of time from signing the contract.
The mechanism itself was a version of a beauty contest. Every bid consisted of seven factors each
of which was evaluated on a given point scale. Seven scores obtained this way were summed to
get a total score for a particular bid. The factors in question are shown in Table I.
Table I: Categories evaluated in a beauty contest.
(a) Speed of network setup and extent of signal coverage, (0-190 points)
(b) Price level of the services, (0-190 points)
(c) License fee not less than $9.5 million, (0-190 points)
(d) Available technological and human resources, (0-150 points)
(e) Amount of domestic capital and managing rights, (0-120 points)
(f) Quality and extent of service and consumer satisfaction, (0-110 points)
(g) Use of domestic human and other resources. (0-100 points)
A professional committee that issued a recommendation to the government, which gave the final
approval and announced the winner, performed the evaluation of bids and categories they were
consisted of. The winning bidder was to pay the license fee stated in its bid in full amount in 60
days after signing a contract. Given that the fee that bidders were prepared to pay for a license is
one of the graded categories this mechanism has a slight scent of a first price auction.
Nine potential operators showed interest in the auction by buying the tender documentation (for
$1100). Three firms underwent serious preparations to enter the auction. A day before the auction
one of them dropped out, so, two firms participated in the auction and the license for the second
GSM operator was awarded to Simobil which paid $13.9 million for it or approximately $7/pop. 3
It is worth noting that the other remaining bidder, Digitel, offered $20 million for a license but
was turned down due to inferior results in other categories.
III. Assessment of the rules: theory and practice
Both theoretical work and wide spread practice in privatizing and deregulating various fields of
state controlled economies suggest that decision to use an auction instead of a beauty contest to
sell the second half of the GSM frequencies would be the way to go. We refer to a wide body of
theoretical literature on auctions, some studies about similar auctions held elsewhere, especially
in USA, and occasionally on plain common sense to argue that a simple ascending English
auction would have outperformed the beauty contest held. We structure our thoughts around five
III.1 The amount of documentation and short time to prepare it
The amount of documentation to be submitted by the bidders was extraordinary and the cost of
preparing it was to be born solely by the firms. The time available to submit it was also quite
$/pop = license fee / population size.
short, namely three months. All aspects of the bidding firm’s organization, capital structure,
references and detailed chart of future actions were to be covered. This was done to scare off
small upstarts not being sure about whether to commit to the job or not. It could have, on the
other hand, also turned away some foreign operators estimating the small market of population of
two million not being worth a hassle. We will comment on this in more detail later when we
touch domestic content issues. This aspect of the selling procedure has therefore probably
reduced the number of competitors.
The same goes for the use of warranties against bidders not committing to and defaulting on the
contract. However, there is a strong evidence from similar tenders that such warranties are
necessary (McMillan, 1994), so this was definitely a right thing to do.
III.3 Composite bids
It is clear that the government by using different factors in evaluating the bids (Section II.1 (a)-
(g)) tried to achieve the allocation that would be a combination between a social optimum and a
revenue maximizing one. Our concern is, though, that concentrating on social efficiency was
probably unnecessary and even harmful from the revenue maximization point of view since
simple auction formats themselves take care of efficiency while keeping the optimal strategies
bidders should follow relatively simple. This is not something the mechanism employed did.
III.3.1 Bidders' optimal strategies are difficult to find
Consider a simple model resembling the beauty contest held. Seller is selling a single indivisible
license. There are two bidders, each receiving two independent signals vi and ui, i = 1, 2, which
are independent from the opponent’s signals also. All the signals are private and drawn from
commonly known distribution. Signal vi is a value the obtained license yields to bidder i while
signal ui represents a level of some other category, called quality, bidder is able to maintain upon
receiving a license, that is also being evaluated by the seller in comparing the bids. Bidder i’s
equilibrium strategies are two bidding functions bi (vi , ui ) and di (vi , ui ) , giving the fee bidder i is
prepared to pay and the level of quality bidder i will maintain, respectively. Seller sums each
bidder’s bids, bi and di, and assigns the license to the bidder with the highest score. We assume
keeping up with the quality level di is associated with constant marginal cost c. We have
described a mechanism featuring characteristics of both a beauty contest and a first price sealed
bid private value auction.
What are the equilibrium strategies bi (,) and di (,) ? The symmetric solutions are obtained from
the following expected profit maximization problem for bidder 1:
max (v1 b1 c d1 ) Pr( b1 d1 b2 (v2 , u 2 ) d 2 (v 2 , u 2 )) , (1)
b1 , d1
ˆ ˆ ˆ ˆ
with symmetry conditions b1 b1 (v1 , u1 ) b2 (v1 , u1 ) and d1 d1 (v1 , u1 ) d 2 (v1 , u1 ) .
First, note that the proposed model falls under the partial equilibrium paradigm. It is a usual
practice in auction literature to treat the bidder’s valuation of an object as given, not being an
explicit result of some future optimizing behavior. It is harder to convince ourselves this ought to
be the case with qualities ui in our model as well. The two most illustrative characteristics from
the beauty contest setup that can take the place of ui in our model are price level and the quality
level of services (Table I, (b) and (f)). A complete characterization and solution of the model
would involve the optimizing behavior on the part of consumers and incumbent firm in
determining the optimal price and quality levels offered. We will argue that solving (1) is very
hard the way it is, hence, doing it in general equilibrium setting would just add another level of
The solution to (1) is beyond the scope of present paper but it should be clear that it would
present a lot of tractability problems. Another point is that our assumption of private values and
independent signals is yet again too simplifying in that the real environment is one of common
values and affiliated signals. In such an environment even the strategies in a simple first price
auction are quite challenging to find (Milgrom and Weber, 1982, pp.1107), so, taking that into
account in our illustration problem would make it even harder to solve. Therefore, it appears that
bidders did not face an easy task in forming their bidding strategies. A reserve price in actual
beauty contest just adds another level of difficulty..
The inability to come up with a sensible strategy that would be close to some kind of equilibrium
may cause even highly rational players to make considerable mistakes with resulting off
equilibrium paths easily leading to inefficient outcomes.
III.3.2 Composite bids are not necessary
Next, we argue that categories (a), (b) and (d)-(g), included in composite bids, although of
perceived importance may be as a matter of fact of questionable relevance, so that the problems
described in previous section might have been avoided in the first place.
III.3.2.1 Entrant’s pricing
Categories under (a), (b), (d) and (f) deal with things the entrant should posses and offer
effectively if he wants to succeed in the market. At the time of the tender bidders had to forecast
what the prices and the level of services on the market would be when entering approximately a
year later. Including these forecasts in composite bids just introduces noise and may lead to an
inefficient allocation of the license. When entering the market in March 1999, Simobil’s prices
were lower and the amount of services offered higher than stated in their bid. This was due to a
big preemptive marketing campaign held by Mobitel during the months before that. They
lowered their prices by some 40% and extended the array of services offered, engaging in
second-degree price discrimination. That certainly did not come as a surprise, but for Simobil to
envision expected Mobitel’s behavior and the expected consumer demand in spring 1999, when
they were preparing their bid in winter 1997-98 was improbable. Therefore, it seems fair to say
that categories (b) and (f) were redundant in the beauty contest.
III.3.2.2 Entrant’s competence
Speed and the extent of signal coverage along with technological and human resources available
(categories (a) and (d)) are more important since they indicate the bidder’s ability to really
perform on the level needed to make a progress from social point of view. It needs to be said
though that the easiest level of offer in any category to check is the license fee offer. Namely, the
amount offered will be paid in 60 days otherwise $1.1 million warranty will be called upon and
license awarded to someone else. To evaluate the statements about future speed of the network
setup and the extent of signal coverage is a harder task. Every bidder has a clear incentive to lie
about it at the time of the tender for at least two connected reasons. First reason is of course to
score as many points possible to win the license. What makes this kind of thinking sustainable?
Second reason does. Once the winner has a license and the competition for it is eliminated, he
can start weighing the costs of meeting the levels specified in the bid against the probability that
the government would actually default him if those specifications are not met entirely. It is
reasonable to believe that there would be some tolerance on government’s part coming from the
fact that if the entrant is defaulted another tender should be probably needed and that some social
costs from still having a monopoly would be incurred in the mean time. At the time of entry
Simobil has for instance gone through some child diseases not anticipated in their bid, but the
government did not consider acting on the contract provisions of cashing in the warranty and
defaulting them. We see a realization of a classic incredible threat that could have been
recognized by a rational bidder. Hence, categories (a) and (d) may have also added just some
more noise to the composite bid.
A better way to deal with the issue of bidders’ competence to perform the job in hand would
probably be to decide on which the firms allowed to compete for the license are prior to the
beauty contest. This might have been done based on the information submitted and evaluated
against some minimum threshold to be achieved by eventual bidders. The level of the threshold
is a variable government could have set in accordance with standards demanded. This would still
not eliminate the firms’ incentives to overstate their capabilities but would take care of clouding
the bids that lead to the final allocation with point scores based on misrepresented data. One can
argue that the committee deciding about who the participants in the auction will be would be in a
hard spot trying to remain impartial in a face of extensive lobbying that would probably take
place. One the other hand, the same can be said for the committee evaluating the composite bids
in the contest held, since the choice of points awarded to a particular bidder was strongly affected
by members’ subjective opinions.
III.3.2.3 Domestic content clauses
Categories (e) and (g) deal with the domestic content issues which are extensively debated in the
international trade literature and usually looked upon as welfare reducing. There are two more
factors apart from usual considerations diminishing foreign firms’ will to enter such markets.
First and connected to the extensive amount of documentation to be submitted, the enterprise
entering the auction had to be registered in Slovenia and owned and managed in its majority by
Slovenian hands. Given that there were just three months to find a Slovenian firm with some
experience and will to enter the market and that such firms were scarce along with the fact that
most of the profits are to be kept in Slovenia some bigger international players that showed
interest in auction decided not to participate.4 Second, given that the majority of telephony know-
how in Slovenia is concentrated in one firm, Iskra Tel d.d., producer of telephony switch boards
among other things, the bidder in a good relationship with them would probably enjoy a big
comparative advantage. Simobil indeed managed to associate with Iskra Tel and have won the
license. Although somewhat justifiable with domestic jobs and profits these two categories have
decreased the extent of the competition in the auction by discouraging foreign firms. Another
matter is again credibility of government positions on the issue. What if sometimes in the future
the ownership mix of the winning firm changes away from what the statement in the bid was?
What if foreign capital’s share eventually exceeds 25% allowed? Will government terminate the
contract? The aftermath: in July 2000 Simobil’s shareholders received an offer from Mobilkom,
mobile telephony provider from Austria, who expressed a wish to buy more than a half of
Simobil. There was no response from the government.
If domestic contents clauses were really necessary the better way to deal with them would
probably be to set some minimum levels and check the potential bidders against them prior to the
Simobil’s Swedish partner Telia is the tenth biggest provider of mobile telephony services in
terms of the number of customers.
auction. The government should have also set higher warranty premiums to support its positions
with respect to the firms’ competence and domestic content issues.
Having said all this, we conclude that simple bids in form of license fees offered might have
done a better job for the following reasons:
bidders would have had easier job identifying optimal strategies avoiding potential
there would have been no noise from unneeded categories in the composite bids,
again, avoiding potential inefficiencies,
potential domestic-content welfare losses would have been avoided and the extent of
competition would have been higher.
If we convinced ourselves that simple bids would be better we need to think about a mechanism
that would be most suitable for the task at hand.
III.4 Open vs. sealed bidding: English auction and more competition?
Milgrom and Weber (1982, pp.1105 and 1109) show that in the common value and affiliated
signals environment ascending English auction yields the highest expected revenues among
simple auction formats when biding is on the price for the auctioned object only. This is due to
the linkage principle, i.e. the bidders remaining in the auction revise their assessments of the
object’s value based on the prices at which exiting opponents dropped out, lessen the potential
winners curse and bid more aggressively. This way bids come closer and closer to what the
object is really worth to bidders. This effect is on average even amplified by seller making any
information he might have public. It is this line of reasoning that usually speaks in favor of open
auction versus sealed bid auction. Open auction is, on the other hand, more susceptible to
collusion but when biding for only one license around which a winner builds his business
collusion probably does not pose a problem since a license is viable for every bidder’s survival.
The other caveat of an open auction is that risk averse bidders tend to bid less than they would in
a sealed bid auction (Riley and Samuelson, 1981). Firms bidding large amounts of money are
surely risk averse but estimating whether this effect overpowers positive effects from information
content is an empirical issue.
FCC for instance decided to go with what is known and adopted an open auction which despite
the complex environment resembled English auction and is called simultaneous multiple-round
The case for English auction is strengthened by Bullow and Klemperer (1996) who show that
employing English auction with n + 1 competitors yields higher revenues than optimal selling
mechanism with n competitors.
English auction with a larger number of competing bidders and simple bids on the license fee
would have therefore probably generated higher revenue not hurting the overall efficiency of the
III.5 Reserve price
Reserve price is important in auctions with small number of bidders since it steps in for the
missing competition and prevents the object from being sold for next to nothing as in the case of
New Zealand (Mueller, 1991). In our case with two bidders the reserve price probably played a
significant psychological and informational role in forming the bids and was a good move. Its
level and the expected revenue to be captured are discussed next.
IV. Were licenses sold cheap?
Although we have been describing the tender for one GSM license the title of this section has
license in plural. This is because the eventual compromise with respect to the license Mobitel
was already operating was that they do not enter the competition and get to keep their license, but
have to match the winning bid in every category. This makes the revenue generated in the auction
even more important, since it was going to be doubled by Mobitel’s payment. It is impossible to
say what a license’s exact worth is, but given the preceding discussion and comparison to the
FCC auctions in USA and prices at which licenses sold in similar Central European countries we
can form an educated guess about the revenue it could have generated.
Slovenian environment is somewhat comparable to the environment in which C-block auction in
USA was held (Cramton, 1997, Section 6). In their case small businesses with annual revenues
less than $40 million competed for the final 30-MHz block in 493 basic trading areas (BTAs).
Average price at which these licenses sold was $39.88/pop, higher than in any other FCC auction
held. Blocks A and B of equal width for which big players were competing in MTA auction
yielded $15.54/pop on average.
Small bidders in C-block auction really went at each other. Government offered attractive
installment payments spread over ten years at the 10-year Treasury note rate which gave them a
lot of leverage. Cramton estimates this option accounts for $16 of the $24 spread between the
two average prices. The rest goes to the competition. There were 255 small bidders in this
auction compared to 30 big enterprises in the MTA A and B-block auctions. Many of the small
bidders were startups that would be out of job if not winning a license.
Big businesses entering A and B-block auction formed strategic alliances reducing the number of
competitors, bided cautiously due to larger stakes involving larger areas (MTAs) and
geographical agglomerations and wanted to keep prices low to avoid winners curse.
Any sort of collusion or complicated strategic behavior was probably impossible in C-block
auction due to the large number of bidders, so the prices went higher.
It is also worth noting that C-block auction lasted 184 rounds from December 18, 1995, till May
6, 1996. This time span gave firms and their creditors time to make difficult decisions and
reassess their positions deciding whether to continue or not.
This reasoning also applies in Slovenia’s case. Government should have gone for an open auction
with simple rules to provide the bidders with a lot of information on the license’s expected value
and with relatively simple equilibrium strategies. They should have also allowed a larger number
of firms to compete at the auction to accentuate the competition effect. Multiple round auction
held over couple of days would have also given bidders some more time to reconsider their
actions and reevaluate their expectations on the future market performance given competitors’
bids. English ascending auction would seem to fit the job perfectly.
What might have government expected to get for the second license, i.e. was the reserve price
that probably gives a good idea about what was expected and hoped for set adequately?
There are two simple ways to form a decent guess about what the winning bidder might be
willing to pay in English auction.
First, a common belief from experience in other European countries and some evidence from
Salant (1995) suggest that the buildup costs are on average expected to be on average at least
double the license fees. Higher costs of course speak for lower license fees. The costs depend on
the configuration of the terrain and since there are a lot of possibilities to cover relatively big
areas from surrounding mountains in Slovenia, they should not be much higher than average.
There were estimates at the time of the auction that Mobitel’s future investment in its network
would be in the range of $150-200 million. Simobil claims to be able to achieve the same signal
coverage with less transmission stations for approximately half that costs. Let $250 million be a
rough estimate of the joint level of investment in setting up appropriate networks by Mobitel and
Simobil planned at the time of the auction. This would suggest that the two licenses would be
worth $125 million or $62.5 million apiece or around $30/pop each.
Second, a simple comparison to the FCC C-block auction can be made. We take the fact that no
installment payments plans were made available in Slovenia into account. The average price in
this case estimated by Cramton (1997, pp.474) would be around $24/pop. This way we arrive at
the estimate $24/pop · 2,000,000 pop = $48 million per license. If government decided to provide
installment payment option similar to the one offered in USA and if this had similar effect on
bidding as in FCC’s C-block auction this estimate would be more around $80 million.
To put these numbers in proper perspective, suppose we settle for the lower estimate of around
$50 million per license. The two licenses would have yielded $100 million, which would be
almost 1.8% of Slovenia’s national budget in 1999. One can argue that these estimates should be
corrected for lower purchasing power of Slovenian consumers. One can also point out that
Mobitel has encountered rapid expansion of the number of its customers that does not seem to be
slowing down as they have already achieved 13% penetration rate (Spring, 1999). Apart from the
time lag of the beginning of GSM telephony in Slovenia, this rate and its growth are on the pace
of the rates in West European countries. Simobil too was surprised about the number of
customers they have after being on the market for two months. They expected 20,000 customers
only by the end of the year 1999. With prices not lower than in Western Europe the case for
smaller value of license as compared to Western countries is weakened somewhat.
There is another quick comparison we can make with countries that are closer to Slovenia than
USA. The second license in Austria sold for $320 million or $40/pop, licenses in Poland for
$400 million or $10/pop and in Slovakia for $10 million or less than $2/pop.
Nevertheless, to be on the safe side, discounting our lower estimate of the license’s value by
factor two yields a number around $25 million or $12.5/pop for the price the license might have
sold for. This conservative estimate is 2.65 times the reserve price set by the government and
more than 1.8 times the price auction realized.
V. The aftermath and conclusion
In the fall 2000 two additional tenders considering mobile telephony will be held in Slovenia.
First, three licenses to operate DCS 1800 system will be sold and second, three licenses for third
generation UMTS systems will be auctioned off. The first tender will be very similar to the one
presented. The only difference is that there are no domestic content clauses to be evaluated in the
beauty contest to be held. If we are lucky we will get three new mobile telephony providers in
Slovenia, while by the worst case scenario both incumbents will get a license, so there will be
space for only one new operator. As for the UMTS system, the government finally gave in and
decided to perform a combination of the two mechanisms discussed. From the pool of all the
applicants five competitors will be chosen first on a basis of their professional capabilities and
than the auction will allocate three UMTS licenses among them. After all these licenses are
allocated we could have between three and eight mobile telephony providers in Slovenia. The
best number for consumers would be of course eight, but in reality a market as small as ours is
can probably not support such a number of firms. The optimal number in our opinion is around
four or five providers.
There are still a lot of state-owned enterprises in Slovenia that will eventually be privatized and
there are still sectors that will have to be deregulated (fixed telephony with Telekom Slovenije
and other infrastructure fields). Government is certainly on the right path to use public tenders for
the job, we only show that their setup must be carefully thought through. Based on what is
known from the theory and practice we have argued that using a simple English ascending
auction with threshold levels the bidders must comply to on some issues, instead of rather
complicated beauty contest would have yielded at least twice the revenue generated, not hurting
the efficiency of allocation.
Numbers 190, 150, 120, 110 and 100 as the upper bounds on possible points scored in different
categories (Table I, (a)-(g)) suggest that serious effort was put in designing the rules of the beauty
contest, therefore, it just remains to channel it in more relevant directions.
Bulow, Jeremy, and, Paul Klemperer (1996), ‘Auction Versus Negotiations’, American
Economic Revue, 88, 180-194.
Cramton, Peter (1997), ‘The FCC Spectrum Auctions: An Early Assessment’, Journal of
Economics & Management Strategy, 6, 431-495.
McMillan, John (1994), ‘Selling Spectrum Rights’, Journal of Economic Perspectives, 8, 145-
Milgrom, Paul R., and Robert J. Weber (1982), ‘A theory of Auctions and Competitive Bidding’,
Econometrica, 50, 1089-1121.
Ministry for Traffic and Communications of Republic of Slovenia (1997), Javni razpis za
podelitev koncesije za uporabo radiofrekvenčnega spektra za storitve mobilne telefonije GSM,
Mobitel (1997), Annual Report 1996, Mobitel d.d., Ljubljana, Slovenia.
______ (1998), Annual Report 1997, Mobitel d.d., Ljubljana, Slovenia.
Mueller, Milton (1991), ‘Reform of Spectrum Management: Lessons from New Zealand’, Policy
Insight, No.135, Reason Foundation, Los Angeles.