Spectrum auctions in Sweden

41  Download (0)

Full text

(1)

Bachelor Thesis in Economics

Spectrum auctions in Sweden

A theoretical study of spectrum auctions in Sweden

Gustaf E. Smedman & Timo A. Kervinen 31 May 2020

School of Business, Society and Engineering Mälardalens Högskola

Bachelor thesis in Economics: 15 credits Course code: NAA305 Supervisor: Christos Papahristodoulou

(2)

Acknowledgements

We want to extend our deepest gratitude to our supervisor Christos Papahristodoulou, who has supported us with constructive feedback throughout the whole writing process. We would also like to extend our gratitude to our respective fiancées, Amanda, and Martina, for their unconditional love and moral support.

- Gustaf E. Smedman & Timo A. Kervinen

(3)

Abstract

This paper seeks to find whether the spectrum auctions in Sweden have been conducted efficiently and if there is a de facto model that suits all auctions. The efficiency is conditions that emphasise truthful bidding, price discovery and limits collusive behaviour. The paper compares three different auction models used in Sweden, a beauty contest used in the allocation of 3G spectrums, and the auction model selected for the upcoming 5G spectrum auction. The auction models are as follows: first and second-price sealed-bid auction, SMRA and CCA. We found that beauty contests should not be used in any spectrum allocation as it did not meet the criteria of efficiency outlined in this paper. The first-price sealed-bid auction is not a suitable format for spectrum auctions. According to the theory, it generates equivalent revenues on average as the second-price format, which shows a higher degree of efficient allocation. We found that depending on the blocks sold, both SMRA and CCA can result in somewhat efficient results, but they are not suitable for a single object auction. We found that no de facto auction format is suitable for every spectrum auction to be conducted in the future, but instead that the auction format is dependent on the characteristics of the individual auctions.

Keywords: Vickrey, CCA, SMRA, Radio spectrums, Single object Auction, Multiple object Auction, Price discovery, Truthful bidding, Collusive behaviour

(4)

Table of contents

Table of Abbreviations ... 4

1 Introduction ... 5

1.1 Problem ... 6

1.2 Aim ... 6

1.3 Limitations ... 7

1.4 Methodology... 8

2 Literature review ... 9

2.1 Sealed bid auctions ... 9

2.1.1 First-price sealed bid auctions ... 9

2.1.2 Second price sealed bid auction ...10

2.2 Multiple unit sealed bid auctions ...12

2.2.1 Revenue Equivalence ...13

2.2.2 Simultaneous Multiple Round Auction (SMRA) ...14

2.2.3 Combinatorial clock auction (CCA) ...16

2.3 Beauty Contest...19

2.4 General remarks ...21

2.4.1 Winner’s curse ...22

2.5 Swedish Spectrum Market ...23

3 Results ...24

3.1 Single object auctions 2008-2017 ...24

3.2 Multiple object auctions 2003–2018 ...27

3.3 5G auction in 2020 ...31

4 Conclusion ...32

Appendix A ...34

Appendix B ...36

References ...37

Various internet sources and legal documents ...38

Figures

FIGURE 1.GRAPHICAL REPRESENTATION OF 3G AND 4G SURFACE COVERAGE IN SWEDEN DURING 2017, FOR A DATA SPEED OF 10MBIT/S, COMPARING THE EFFECT OF NET1’S CONTRIBUTION TO THE TOTAL SURFACE COVERAGE IN EACH COUNTY. ... 26

Tables

TABLE 1.VCG-PAYMENT EXAMPLE ... 18

TABLE 2.SINGLE OBJECT AUCTIONS HELD BETWEEN 2008-2017 ... 24

TABLE 3.2,6GHZ SMRAAUCTION 2008 ... 29

TABLE 4. 800MHZ SMRAAUCTION 2011 ... 29

TABLE 5. 700MHZ SMRAAUCTION 2018 ... 29

(5)

Table of Abbreviations

Abbreviation Explanation

CCA Combinatorial Clock Auction

FCC Federal Communications Commission FDD Frequency Division Duplex

GHz Gigahertz

Mbit Megabit

MHz Megahertz

PTS The Swedish Post and Telecom Authority SMP Significant Market Power

SMRA Simultaneous Multiple Round Auction SÄPO Swedish Security Service

TDD Time Division Duplex VCG Vickrey-Clarke-Groves VoIP Voice over Internet Protocol WDP Winner Determination Problem

(6)

1 Introduction

Auctions have been used over several millennia to distribute resources for the ones who value them the most. Some of the first documented cases of auctions have been witnessed in Herodotus’s book The Histories, where he described a Babylonian wife market in 500 B.C.E, in where women were sold to the highest bidders (Böheim & Zulehner, 1996).

Today auctions are used in the allocation of a wide range of objects, and the values of these objects can vary tremendously. The auctions are conducted both by governments and auction houses around the world; some notable auction houses are Sotheby’s, Christie’s, and eBay.

These institutions sell everything ranging from art and jewellery to houses and natural resources. Several other auctions may catch the attention of an economist, such as, a wide range of commodity auctions, treasury bills, assets under privatisation, and public utilities such as radio spectrums (Koutroumpis & Cave 2018).

This paper focuses on the Swedish spectrum market, in which multiple different auction mechanisms have been used to allocate spectrum rights, and it evaluates the efficiency of each auction model used. The Swedish Post and Telecom Authority (PTS) regulates the market and conducts the auctions.

Approaching the spectrum auctions from an auction theory perspective is complicated but fascinating. The spectrums are heterogeneous products but similar in certain aspects which leads to high competition for the spectrum licenses among companies. The companies are not locked into using specific technologies but instead develop ones that suit their needs the best.

The complex structure of substitutes and complements places the spectrum auctions among the most challenging auction settings seen in practice (Cramton, 2013). The complicated auction setting combined with the allocation of a scarce natural resource, which is vital to the technological development of any nation, emphasises the fact that a mechanism which maximises the societal wellbeing is needed.

(7)

1.1 Problem

There have been instances in Sweden where the use of inefficient allocation mechanisms has led to dissatisfactory outcomes. The three large telecom companies Telia, Tele2 and Vodaphone1, gained significant market power (SMP) and charged taxpayers above the market price for their services. In other instances, the companies have failed to meet the demands set by the Swedish government in regards of the network coverage and instead of investing more in construction to meet these demands, they have left the market which leads to some bandwidths not being used in Sweden (Welin, 2006).

Without careful regulation and only focusing on the revenue maximisation from the auctions, the government faces a risk of oligopolies, or a monopoly, forming in the market. Such outcomes reduce the efficiency of the spectrum allocation. The lack of efficient allocation can escalate the negative aspects that have been witnessed in the market today.

The three most prevalent problems we identify are: dissatisfactory expansion of the radio network, limited technological development and lack of competition between the telecommunication companies. These outcomes would lead to Sweden falling behind on the evolution of radio technologies and reduce the societal wellbeing that could be achieved.

A more competitive market will likely generate higher government revenues in the long term while remaining efficient. This can be accomplished by an auction mechanism that allocates the spectrum licenses to competitive firms which use the licenses to their foremost potential (Cramton, 2013).

1.2 Aim

This paper analyses different allocation methods the PTS has applied in the radio spectrum market and investigates whether there is an auction format that is more efficient than the previously used methods. The PTS has prior applied many different approaches, and we try to find if there is one auction format that outperforms the alternatives in terms of efficiency for all allocations in the future.

1See Karlberg, L. A. (2004, December 12)

(8)

1.3 Limitations

1. The study is a theoretical review from the perspective of the government. The viewpoint is limited to the extent that the mechanisms incentivise truthful bidding, price discovery and limits collusion. This angle excludes all the costs incurred by the companies as well as the revenue streams.

2. Our analysis is based on previously published research about auctions and to a large extent, papers published about spectrum auctions. Due to the complexity of the underlying mathematics in advanced auction mechanisms, some mathematical statements are taken as given, i.e. for the mathematical proof, we will refer to the authors' original paper.

3. There is limited access to data concerning the previously conducted auctions in Sweden, such that only the winning bids are published for most of the auctions. The lack of information about all the bids placed in the auctions limits our analysis of the bidders' behaviour.

4. There is no accessible data concerning the exact utilisation of individual radio spectrums; this leads to high limitations while empirically evaluating the efficiency of the outcome of different auctions.

5. The paper is limited to the auction formats in their general form. Thus, there is a possibility that a better alternative exists as there are multiple different variations of the auctions which can alternate the outcome significantly.

6. We have a strong assumption while theoretically evaluating the auctions, that if the bidder bids his truthful valuation and wins the spectrum license, he will utilise the spectrum to its foremost potential in line with the vision of the Swedish government.

7. The decision process behind the allocation mechanism chosen by the government is influenced by local market conditions and the current political agenda. This is beyond the scope of discussion in this paper.

8. The auction mechanisms concerning this paper can have one or multiple winners and different payment rules. Therefore, the comparison between the simpler mechanisms and complex mechanisms is limited.

(9)

1.4 Methodology

An extensive literature review of auction theory and recent studies about spectrum auctions, as well as a thorough reading of legal documents from the PTS regarding the spectrum auctions in Sweden was conducted. This paper gives an introductory level literature review about radio spectrums (see Appendix A).

We introduce the theoretical models of the auction formats and the positive and negative aspects of them and see how they performed when the PTS used them. We also introduce some of the fundamental theorems needed to understand how auction mechanisms work, such as the revenue equivalence theorem, winners curse, winner’s determination problem and VCG- mechanisms.

To conduct the study, we have formulated three questions and defined efficiency according to three different criteria. We believe that, if an auction format can fulfil these criteria, it is deemed efficient and should be used by the government for future spectrum auctions.

i. Were the allocations methods applied previously efficient?

ii. Is there an auction format that outperforms the previously used ones in terms of efficiency?

iii. Is there a de-facto mechanism that can be used for all spectrum auctions?

Our criteria of efficiency are defined to be the following, and we deem a mechanism to be efficient when it fulfils all three.

Truthful bidding ensures that the object is won by the bidder who values it the most. This criterion is in line with the principles of auction theory, which states that the auction is efficient when such an event occurs (Krishna & Krishna, 2009).

Price discovery provides the bidders with more accurate valuations, ensuring that the truthful bid is in line with the competitive market price of the object.

Limiting collusive behaviour, as, without the limitation, the auction can result in a dissatisfactory outcome. The bidders can block other competitors from entering the market, reduce or increase the prices paid for the spectrum licenses and form oligopolies in the market.

(10)

2 Literature review

The presentation order of the formats is such that single-unit auction models are first, which is followed by multiple-unit formats. The formats are introduced in ascending order of complexity, such that a more complex form follows the simplest form. Along with introducing the different methods, we highlight the problems that each method has for the auctioneer selling the spectrum blocks. We will then introduce more common problems that every auction format shares, namely the winner’s curse and collusive behaviour.

2.1 Sealed bid auctions

Historically the PTS has used sealed bid auction formats for allocating both a single spectrum for a specific frequency bandwidth, i.e. a single object auction, and spectrum licenses for multiple frequency bands simultaneously, i.e. a multiple object auction. Although the single object auction formats are not as common as the multiple auction formats, the auctions are still important to analyse. These auctions can occur when the PTS has analysed the usage of a particular spectrum and seen a possibility for the opening of another block, or when the PTS wants to limit the use of a specific spectrum by not auctioning all the available blocks to increase societal benefits.

The cause behind not auctioning all blocks at once be, for example, that opening up another block and allocating it may favour the current operators in the market instead of increasing competition. Or that the marginal benefit of an increase in the spectrum is zero or hard evaluate.

We will start by introducing the single object auctions for simplicity and then present the multiple object auctions. We will discuss the mechanisms incentive to truthful bidding as well as their problems.

2.1.1 First-price sealed bid auctions

In a first-price sealed-bid auction each bidder submits their bid anonymously, the participants do not know the other bids. The highest bid wins and its paid by the participant who submitted it. A result of the anonymous circumstances the bidder's strategy is simply a function of his or her maximum valuation of the object; therefore, no dominant strategies exist in these auctions (Bichler, 2018). To show the trade-off the bidder's faces, we can represent the payoff mathematically.

(11)

In a first price sealed bid auction, the participants bids are represented by 𝑏𝑏𝑖𝑖. Given the fact that no other information than the bidders own valuation is known, the payoff function for bidder 𝑖𝑖 is expressed as:

𝛱𝛱𝑖𝑖 = �𝑣𝑣𝑖𝑖 − 𝑏𝑏𝑖𝑖 𝑖𝑖𝑖𝑖 𝑏𝑏𝑖𝑖 > 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗

0 𝑖𝑖𝑖𝑖 𝑏𝑏𝑖𝑖 < 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗

Where 𝑣𝑣𝑖𝑖 represents bidder i’s valuation of the object and 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗 represents other bidders max bid not equal to i.

Studying the payoff function, we notice that there is no incentive for the bidder to place a bid such that 𝑏𝑏𝑖𝑖 = 𝑣𝑣𝑖𝑖 or 𝑏𝑏𝑖𝑖 > 𝑣𝑣𝑖𝑖, as this would result in a payoff of zero, even if the bid is the winning bid. Considering a scenario where the bidding behaviour of the other participants is fixed, such that, at any bid, there is no guarantee of winning nor losing. In this scenario only a simple trade-off needs to be considered by the bidder, that is, as the bidder increases the probability of winning by increasing his or her bid, the gains from winning decreases (Krishna

& Krishna, 2009).

The first price auction format has a simple mechanism: the auction is one of the more cost- efficient mechanisms used when allocating spectrums. It has been the motivating factor behind its previous use in Swedish spectrum auctions. Since the format is more straightforward, it is easier to grasp and requires very little administrative work to conduct both from the sellers and the bidder’s perspective.

In first price sealed bid auction, the bidders have no knowledge of the other bidders’ actions and have no knowledge of how many bidders participate in the auction. This creates a problem where the bidders have difficulty with setting a correct value for the bid. The difficulty of valuation is an effect of the mechanism having no possibility for price discovery among bidders.

Shachat & Wei (2012) described problems in the first price sealed bid format; they stated that the Nash equilibrium predictions regarding prices and individual bids are not accurate in the model.

2.1.2 Second price sealed bid auction

Another type of sealed-bid auction used by the PTS is the second price sealed bid auction, also known as the Vickery auction, was first introduced by Vickrey (1961).

(12)

The second price sealed bid auction has a similar bidding procedure as the first price sealed bid auction. The difference between the auctions is that the winner of a second price sealed bid auction does not pay the winning bid. Instead, the winner pays the second-highest bid alternatively in some cases the second-highest bid plus a small increment but still less than the highest bid.

In a second price sealed bid auction the submitted bids are represented by 𝑏𝑏𝑖𝑖, given the fact that no other information then the bidders own valuation is known, the payoff function for bidder 𝑖𝑖 is represented as:

𝛱𝛱𝑖𝑖 = �𝑣𝑣𝑖𝑖 − 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗 𝑖𝑖𝑖𝑖 𝑏𝑏𝑖𝑖 > 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗 0 𝑖𝑖𝑖𝑖 𝑏𝑏𝑖𝑖 < 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗

Where 𝑣𝑣𝑖𝑖 represents bidder i’s valuation of the object and 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗 represents other bidders max bid not equal to i.

The auction is designed to incentivise bids closer to the participants' true valuation of the object;

we notice this as we study the payoff function. This is consistent with the fact that a weakly dominant strategy is to do bid one's true valuation (Krishna & Krishna, 2009). This can be proven mathematically as follows:

Let us consider a bidder 1 and suppose that the highest competing bid in the auction is 𝑝𝑝1 = 𝑚𝑚𝑚𝑚𝑚𝑚𝑗𝑗≠𝑖𝑖𝑏𝑏𝑗𝑗 which is unknown to the bidder. If bidder 1 bids, 𝑣𝑣1 > 𝑝𝑝1 then bidder 1 wins the auction. If 𝑣𝑣1 < 𝑝𝑝1, he loses the auction. Assume bidder 1 is indifferent between winning and losing if 𝑣𝑣1 = 𝑝𝑝1.

Let us say bidder 1 bids a non-truthful bid 𝑤𝑤1, instead of 𝑣𝑣1, such that 𝑣𝑣1 > 𝑤𝑤1 ≥ 𝑝𝑝1, then bidder 1 wins the auction and he gains 𝑣𝑣1− 𝑝𝑝1. On the other hand, if 𝑣𝑣1 > 𝑝𝑝1 > 𝑤𝑤1 bidder 1 loses the auction. In this scenario, if bidder 1 would have bid his true valuation 𝑣𝑣1, he could have won and gained a positive profit. Thus, bidding less than one’s true valuation will never increase the profit but may under some circumstances decrease it. (Krishna & Krishna, 2009).

The second price method relies heavily on the idea that every bidder is rational and will bid their true valuation of the block at the auction. However, this idea fails as bidding one’s true valuation is only a weakly dominant strategy. The gap between 𝑣𝑣1 and 𝑤𝑤1might be significant, i.e. as long as the bid 𝑤𝑤1 is marginally higher than 𝑝𝑝1 the bidder still wins. Thus, the winning bid might not be truthful.

(13)

If a bidder believes that all other bidders are bidding truthfully and want to elevate his probability of winning, bidding more than their true valuation would increase their chances that they will win the auction. Because of the method used, they would only have to pay the second- highest bid. However, if all the bidders were to act this way, the block would end up being overvalued, and the winning bidder will obtain no surplus or a negative surplus.

If two bidders were to bid the same amount that would win the auction, the winning bidder has to be chosen by a lottery, and they would have to pay their maximum bid. Such outcome causes two problems, as paying the maximum bid, their truthful valuation, will not generate any surplus for the winner. The lottery aspect would run a risk of the losing bidder to challenge the auction method in court, which would create additional fees for all parties involved.

2.2 Multiple unit sealed bid auctions

The first price auction has no multiple object counterpart; however, the second price auction does. The multiple unit auction counterpart to the second price sealed bid auction is a Vickery-Clark-Groves (VCG) auction. The auction provides a truthful mechanism designed to achieve the socially optimal solution when allocating an object, as it incentives participants to bid closer to their maximum valuation of the object (Bichler, 2018).

The difference between second-price and VCG is the payment rule, “each bidder is asked to pay an amount equal to the externality he exerts on other competing bidders”2. In this auction, it is still a weakly dominant strategy to bid one’s true valuation. We will introduce a simple mathematical example of the VCG-mechanism in the upcoming section concerning combinatorial clock auctions.

A key issue that arises in the VCG mechanism is its core idea of requiring bidders to submit their truthful valuations of the blocks being auctioned. Some participants may be hesitant in revealing their true valuation as it might contain revealing information about the participants themselves.

The mechanism has a risk of generating zero revenues for the seller due to the payment mechanism used, where the price of the blocks is determined from the opportunity cost to other bidders. Ausubel & Milgrom (2006) formulate an example of this:

2See Albano, G. L. (n.d.). An Introduction to Multiple Object Auction (p.25)

(14)

“For example, consider a hypothetical auction of two spectrum licenses to three bidders.

Suppose that bidder 1 wants only the package of two licenses and is willing to pay $2 billion, while bidders 2 and 3 are both willing to pay $2 billion for a single license. The VCG mechanism assigns the licenses efficiently to bidders 2 and 3. The price paid by bidder 2 is the difference in the value of one license or two licenses to the remaining bidders. Since that difference is zero, the price is zero! The same conclusion applies symmetrically to bidder 3, so the total auction revenues are zero. “(p. 11)

The above example is an extreme case and assumes there are no reserve prices set. This scenario is unlikely to happen in real life. Regardless of the auction outcome giving zero revenues to the auctioneer, the allocation is still efficient as the licenses were allocated to the participants who valued them the most.

As the payment mechanism relies on the number of bidders participating and their subsequent bids, the results can lead to a non-monotonic outcome of the auctioneer’s revenues. In the previous example, three bidders were participating. However, if the third bidder did not attend, the seller would receive 2 billion in revenues from one of the licenses sold to bidder 2. It creates a possibility of collusion, where the bidders can invite other bidders to participate in the auction in order to lower the prices they would have to pay (Ausubel & Milgrom, 2006).

The problems in the mechanisms will never occur if the bidders have substitutes preferences.

However, if there is one bidder whose preferences violates the substitutes condition, there is a possibility of all the latter mentioned problems occurring. Due to these weaknesses, it is more often used as a payment-rule following another type of auction, rather than being used as an auction format itself.

2.2.1 Revenue Equivalence

Studying the two sealed bid auction formats in single object auctions in the sections above, we notice that they differ in the aspect that the second price auction has a weakly dominant strategy, while the first-price auction has no dominant strategy. However, the equilibrium bidding strategies are the same when deriving3 them.

A crucial theorem used to derive the equilibrium strategies in the single object auctions is the revenue equivalence theorem, first introduced by Vickery, (1961).

3For a complete derivation of the equilibrium strategies see Krishna & Krishna, (2009).

(15)

More formally, the theorem states “Suppose that values are independently and identically distributed and all bidders are risk neutral. Then any symmetric and increasing equilibrium of any standard auction, such that the expected payment of a bidder with value zero is zero, yields the same expected revenue to the seller.” (Krishna & Krishna, 2009, p.28).

According to the revenue equivalence theorem, the sealed bid auctions generate the same level of expected revenue to the seller in the single object auction case; in fact, this is true for all simpler auction formats. (Krishna & Krishna, 2009) However, it is important to note that equal revenues do not imply equivalent efficiency of allocation.

2.2.2 Simultaneous Multiple Round Auction (SMRA)

Multiple-lot auctions are the most frequently accruing situation in spectrum auctions.

There is a need for alternative methods to the simpler auction mechanisms such as the first price and second-price auction. Therefore, the PTS adopted a globally used method for spectrum auctions, known as the simultaneous multiple round auctions.

Simultaneous Multi-Round Auctions, as the name implies, conduct multiple different auctions at the same time. It means that all the licenses that are available at the same time to the bidders, who can freely choose which license(s) to bid on. In a similar fashion to an English auction, the bids are ascending, and each bid needs to be higher than the previous bid.

The bidding continues in multiple scheduled rounds, in which each bidder can bid on as many lots as they want. The round ends when there are no more new bids, and the auctioneer announces the highest bid on each lot, and declares the bidders holding the highest bids as the

“standing high bidder”. After the announcement, a new round of bidding starts. Only the lots that received bids on the previous round are included4.

Due to the complex nature of the bidding process, as there are multiple auctions at the same time, bidding is done electronically. The software used in the bidding will update itself continuously during the bidding process to ensure that the bidders have real-time knowledge of the current highest bid. Along with the highest bid, the bidders also know who the highest bidder is (Keuter & Nett, 1996). In the case of spectrum auctions, no physical presence is needed to participate.

4See DotEcon. (2009). Liberalisation of spectrum in the 900MHz and 1800MHz bands.

(16)

The bidding ends when there are no longer any higher bids submitted. The auctioneer then discloses the Winner of the auction which coincides with the highest bid submitted on that license. Depending on the auction, either all the bids, including the non-winning ones or just the winning bid is published.

SMRA uses an activity rule, which requires the bidders to participate in the auction from the beginning and it this way eliminates to the possibility of playing snake-in-the-grass during the auction in which the participant would monitor the bidding activity and only submit their bid towards the end of the bidding activity. Another rule imposed to the bidders is monotonicity, which makes the bidders not eligible to bid on a higher number of lots later in the rounds.

The auctioneer might set a price reserve and define the bid increments in the auction. The bid increments force the bidders to bid sequentially and forbid jump bidding, in which a bidder might bid a value that is much higher than what the current bid is. These rules are there to minimise the possibility of a monopoly forming after the auction has ended. The rules also help to minimize the risk of a participant not winning enough blocks to operate efficiently (Bichler, 2018).

As the licenses for each block tend to be homogeneous, the end prices paid by the bidders tend to be close to each other. Due to the nature of the participants having information about rival bids, the uncertainty of their valuation of the spectrum is limited. The price discovery takes away the pressure of the participants to do extensive research on historical bidding behaviour of the others before the auction. Due to these benefits, the winner’s curse is unlikely to happen.

Another benefit of the mechanism is the ease of the auction format compared to other more complex models, and the bidders are free to bid on the specific licenses that they need. (Keuter

& Nett, 1996)

They have motivated its use of the model by the facts that the model has been used successfully in other spectrum auctions and that it is a format suitable for when auctioning multiple licenses at once. In the case of the 700MHz auction5, PTS set limitations for the information that the bidders can see. The information provided on which licenses the bidder has a standing high bid and the amounts, bidders own activity in the previous round and eligibility for future rounds, and how many extensions the bidder has to use in the auction.

5See Dnr: 17-9908 (2018, July 4)

(17)

Due to the nature of the blocks being substitutes and complements, the bidders face an exposure problem, which might result in a non-core outcome as some bidders might obtain blocks at low prices. The bidders can also see other bids and their competitor’s behaviour which causes a risk of tacit collusion in the form of retaliatory bidding to raise the prices and signalling others with the bid amounts. (Bichler, Shabalin & Wolf, 2013).

As each block is auctioned independently, there is a negative side-effect of the bidders who seek to obtain more than one block, might not be able to secure contiguous blocks and have a risk of fragmentation. It also runs a risk of a bidder who seeks to obtain more than one block only getting one or no blocks (Damic, Peteh & Petric, 2012).

2.2.3 Combinatorial clock auction (CCA)

Ever since the combinatorial clock auction was first proposed by Ausubel, Cramtom and Milgrom (2006), it has garnered much attention from academic circles. The model was quickly adapted in to practice by the UK. In 20086, it was used for two spectrum auctions replacing the standardised SMRA. Since then, several governments have used the auction format. The model has not yet seen use in Sweden, but it is the chosen model for 5G auctions in 2020.

The CCA combines two different auction formats with a VCG-mechanism to combat issues that are in other auction formats. The authors claim that the format is especially useful in situations where the products sold are complements and/or substitutes, such as when auctioning spectrum licenses. (Ausubel, Cramton & Milgrom, 2016)

The first stage is a multiple clock round auction. The auctioneer announces a descending price per block. The participants then respond with how many blocks they wish to obtain at the current price level. This procedure creates a linear demand curve. The responding bids have to be submitted within a specified time frame decided by the auctioneer. The bidding amount is calculated by multiplying the announced clock price with the number of blocks demanded. If a bidder does not respond with any demand within the time frame, it indicates that the bidder has zero demand at the current price. The bids which match the asking price are called standing bids, and the bidder is considered standing if he has at least one standing bid. If the aggregate demand of one clock round exceeds the supply, a higher price is announced for the next round (Bichler, 2018).

6See World first for DotEcon’s combinatorial clock auction. (n.d.)

(18)

If some bidders in this scenario were to lower their demand, other bidders could place what is known as an exit bid. An exit bid announces the bidder’s willingness to win more blocks than the amount won in the previous round. The number of blocks a bidder can submit an exit bid on is limited, to the total number of blocks he bid on during the last round. Once the demand and supply are equal, the clock phase will stop, and the bundles are allocated according to the previous bids submitted (Bichler, 2018).

The clock auction has specific activity rules, such that as the price rises, the blocks demanded by bidders cannot increase. While this works when the auctioned items are homogeneous but, in a case, where there are multiple heterogeneous items, the rule might be too restrictive.

Ausubel, Cramton & Milgrom (2017) introduced a modified activity rule which removes some of the restrictions that might hinder the auction process. The modified rule allows bidders to increase their demand on individual lots, as long as, the increased demand in those lots matches that of the demand of another heterogeneous lot.

The clock stage provides a simple and effective way for price discovery, which is helpful when going forward into the second stage of the auction, known as the supplementary round, which starts after the first stage ends. The second stage gives a sense of security to the bidders as the bidders can correct the valuation errors; they might have had during the clock phase. It is important to note that the bidders cannot decrease their bids in the second stage, but can only increase them. (Ausubel, Cramton & Milgrom, 2016)

During the supplementary round, bidders can bid on as many additional bundles as they want as well as improve the bids placed during the clock phase using a sealed bid auction format. In the CCA auction, the bidders must bid on all the possible combinations of blocks and individual blocks for them to be eligible to win some. All bids placed during both phases of the auction are entered into what is known as a winner determination problem (WDP), to determine which bids are winning (Ausubel & Baranov, 2017). The auctioneer solves to the problem with the intent of finding allocations that maximise the revenues from the auction (Damic, Peteh &

Petric, 2012).

The winner determination problem is an optimisation problem that occurs when conducting multiple-object auctions with stages consisting of different auction formats, such as the CCA.

The WDP solves the universal value maximising assignment of the blocks. ”Winning

packages are determined by finding an allocation that maximises the total value (as reflected in bids) subject to feasibility constraints: each item can be sold only once, and only one bid

(19)

from each bidder can be part of the winning allocation”(Ausubel & Baranov, 2017, P.2). In the scenario of there being more than one value maximising allocation, the winner is chosen randomly. (Ausubel & Baranov, 2017).

The third stage of the CCA is known as the assignment stage; this is where the VCG- mechanism is applied. The WDP finds the winning valuations for each block, and the VCG- Mechanism is used to solve for what price the winners should pay for each respective bundle or single block. How a simple example of how the mechanism works is presented below in Table 1.

Table 1. VCG-Payment example7

Bidders {X} {Y} {X, Y}

Bidder A 30 21 53

Bidder B 24 24 49

Note. Assume a combinatorial auction, hence the bidders are obligated to submit bids for all bundles and individual items. Suppose two items are to be sold, object X and object Y. Object X and Y are complements and the bidders’ value the combination of them more than when obtained separately.

As seen in Table 1, the maximum revenue of the auction would yield 30+24= 54, in this case, object X is sold to bidder A, and object Y is sold to bidder B. However, the bidders do not pay their maximum bid as they receive a form of deduction known as a Vickrey-payment.

Bidder A receives a Vickrey payment of 54-49= 5 since without his participation the total revenue of the auction would have been 49, as bidder B would have won and paid for the bundle {X, Y}. The net payment of bidder A is then 30-5=25 for object X, which is the highest standing bid for object X reduced by the Vickrey-payment.

Bidder B receives a Vickrey-payment of 54-53=1 since without his participation the total revenue of the auction would have been 53, as bidder A would have won and paid for the bundle {X, Y}. Bidder B then has a net payment of 24-1=23. After the VCG-payments have been deducted, the auctioneer receives a total revenue of 25+23=48. Neither of the participants receives a bundle, but instead the object they had the highest standing bid on.

As shown by the example above, as VCG-payments are introduced, the bidders are incentivised to place bids equal to their true valuation. It is important to note that the auctioneer does not

7Modified example retrieved from (Bichler 2018, p 101-102)

(20)

implement VCG-payments to maximise revenues but to incentivise truthful bidding. The combination of a WDP-algorithm to maximise the auctioneer’s revenues and the VCG- payments to induce truthful bidding provides CCA-Mechanism with the possibility to maximise revenues while securing an efficient allocation of the objects (Bichler, 2018).

The CCA faces a problem due to the VCG-payments, as the winning bidders might not end up paying the same price for identical blocks, thus violating the law of one price. The model is also open for spiteful bidding, in which a company tries to purposefully increase the price that the winner ends up paying and thus limiting his or her surplus gained from the auction. While this is the case for many formats, the two-stage nature of CCA gives rival bidders two distinct chances for such activity while ensuring that the spiteful bids will not be the winning ones.

(Bichler, Shabalin & Wolf, 2013).

As bidders are obligated to place bids on each combination of blocks in the auction, they have to value blocks which are of no interest to them. Price discovery may, therefore, be limited to a certain extent as bids placed on individual blocks or bundles will not be proportional to others.

Empirical studies comparing the CCA and SMRA have been conducted by Ihle, Marsden &

Traber (2018) which show that the CCA generates higher revenues than other open auction formats such as the SMRA and this was occurring more frequently for lower frequencies from 600-900Mhz. Ihle, Marsden &Traber (2018) also emphasise the fact that governments should be careful when using the CCA for these frequencies as the empirical evidence suggested that the mechanism promotes deviations from straightforward bidding, this causes higher prices and in turn, promotes an inefficient allocation. On the other hand, Cramton (2018) argues that the CCA has advantages over the SMRA as “It eliminates the exposure problem; it eliminates most gaming behaviour; it enhances substitution; and it encourages competition”(Cramton, 2018, p.187).

The model is much more complicated than other previously used ones, bidders might run a risk of not being adequately educated on the nuances of the auction and the optimal strategies.

2.3 Beauty Contest

An alternative to different auction mechanisms, many governments have opted to use a beauty contest instead. Beauty contests look into the capabilities of the buyer, and the seller tries to find the best option to maximise the social welfare instead of the revenues generated from the auction. (Madden, Bohlin & Tran, 2013) In the context of this specific subtopic about

(21)

a beauty contest, the term “bid” is not necessarily a monetary payment towards the government but more of a “proof of capabilities that match the requirements of the government”.

The fundamental principles of a beauty contest are to evaluate the suitability of the bidder for the network operation in a way that they can meet different demands set by the seller (Madden et al., 2013). When beauty contest is used for telecommunication networks to evaluate different firms to find the most suitable one, the seller can set up several criteria, which may be weighted differently. These criteria might involve things such as financial capability, research and development investments, speed and capacity of the technology and the speed in which the network can be built. (Prat & Valleti, 2000)

This model enables the seller to review the offers more closely and give the right to build the network for operators that match the criteria the best. The model transfers the cost of research when it comes to investigating the potential value of the market and investments needed to make meet the requirements set by the government from the seller to the buyer (Welin, 2006).

Winning a beauty contest relies solely on the evaluation of the capabilities of the entrants by a committee and exposes itself to a multitude of problems. The members of such committee might have differing opinions on which aspects of the contest should have the highest priority, and they run a risk of having members who are subject to bribery and favouritism (Prat & Valletti, 2000). The committee might have difficulties in verifying the information submitted by the participants and must trust that the information provided is correct (Cartelier, 2003).

As governments which oversee the committees sometimes have part ownership of some of the companies participating in the contest, the decision to allocate the spectrum to such a company might be biased. The model also opens possibilities of lobbying by the participants to be treated more favourably.

The risk exists that the bidders submit offers that are not truthful such that they cannot deliver on what was agreed on and could thus face fines imposed by the government. These fines are counter-intuitive for the concept of beauty concept as it tries to maximise social welfare, and the brunt of these fines will end up as increased prices for the end-user (Welin, 2006).

As the process of evaluating different bids can be a lengthy process with multiple different variables, it runs a risk of being challenged in the court of law by the losing bidders (Welin, 2006). This process will add costs to the process and lower the public trust in the system.

(22)

While the government is to impose specific operating fees that the buyer must pay, the amount of fees is hard to determine. If the license sold is for a long duration, the value of such a license will raise. This raise in valuation means the operators have to pay much more for the right to operate and thus charge more for the end-user to access the network. This method will then lower the social welfare that the government is looking for initially with the concept of a beauty contest (Madden et al., 2013).

The winning participants do not have to pay the government to obtain the license, which will cause an apparent drop in government revenues when compared to other formats used. This model deprives resources that could be used for other government projects and in a way, is subsidising profits to the participating companies (Cartelier, 2003). As the companies do not have to make an initial investment, they have a higher incentive to abandon the operations concerning building the network.

2.4 General remarks

Although the auctioneer wants to maximise the efficiency of the auction, there might be varying degrees of efficiency depending on the format used. Still, the behaviour of bidders is essential. All auction formats introduced in this paper share a common problem of collusion.

The methods of collusion can vary, but they all tend to achieve one of two possible outcomes.

Either to reduce the amount paid for the blocks or to increase the amount paid by the rival bidders.

Explicit collusion by the companies is when they agree before the auction starts on who bids on what. This way, each company gets what they want at the lowest price possible. Such a model of collusion depends on every participant agreeing on colluding. An outside bidder has the advantage of raising the prices of the colluding firms, possibly winning some blocks that were supposed to go to a member of the collusion. Separate agreements can be made between buyers, in which the bidder who wins the auction will share the surplus equally among each other (Agranov & Yariv, 2017).

Tacit collusion is another typical form of collusion that can happen during auctions, in which the bidders do not need to have an outside agreement before the auction. During the auction, a bidder might jump bid in which they submit a bid which is much higher than the current highest bid. The bidder does this to signal their valuation of the object. Retaliatory bidding occurs when a bidder submits a bid on a block that they are not interested in acquiring but instead want to minimise the surplus of the winning bidder. (Bajari & Yeo, 2008)

(23)

Bajari & Yeo (2008) exemplify a way of tacit collusion. They observed that bidders would use the last three digits in their bids to signal to other bidders that they were going to defend a specific license. Tacit collusion can be observed in cases where the bidders can see the actions of other bidders, such as in every open auction format. Explicit collusion can appear in every case, regardless of the format used. The auctioneer can take measures to limit tacit collusion, such as setting a limit on price increments but limiting explicit collusion is exceptionally challenging.

2.4.1 Winner’s curse

One closely related phenomenon to the previously introduced revenue equivalence theorem is the winner's curse. Winner's curse is a concept that describes a situation where the winning bidder of an auction has overestimated the value of the item and thus experiences a loss. In this case, the winner might have been better off by not winning the auction. The curse cannot happen if the bidders act rationally, and the curse is an anomaly in the market (Thaler, 1988).

Rational behaviour can be challenging to establish during the auction as it relies on accurate valuation of the object and that the bidders have a full understanding of the auction format in use. In the case of a first-price auction, the number of bidders can have a direct impact on the chances of the winner's curse occurring (Thaler, 1988). As the number of bidders increases, the probability of winning the auction goes lower, which the bidders can combat by bidding aggressively during the auction. If the bidder bids aggressively up to their valuation and stops, the curse does not occur. However, they also run a chance of not winning the auction, especially if all the participating bidders have somewhat similar valuations.

The optimal strategy to avoid the winner's curse is thus to have a thorough understanding of the auction format and the valuation of the item. While it is always imperative to understand the auction format, holding firmly to one's valuation might run a secondary risk of never winning an auction. The participants will need to evaluate, which is more important, to avoid the winner's curse or win an auction that enables the company to operate in their field of business.

It is important to note that the winner's curse can occur in every auction format, and it is not mutually exclusive to a specific mechanism.

(24)

2.5 Swedish Spectrum Market

In Sweden, a spectrum license provides the holder with the right and obligation to manage the frequency bandwidth for a specified period, generally between 10 to 15 years depending on the spectrum in question. The operator has to pay an annual fee, which is dependent on the spectrum license they hold.

The PTS limits entry to the auctions from bidders who seek to resell the licenses or keep them non-operational, as PTS seeks to fulfil their vision “To maximise the long-term societal benefit of radio spectrum in Sweden”8. The PTS pursues the vision by attempting to allocate the spectrum licenses to companies that utilise them to their utmost potential.

The PTS has a set of entry requirements for auctions, that can be seen as barriers of entry. The three most prominent barriers are:

- A formal bank guarantee as proof of the financial capacity. The amount varies depending on the auction format. In some instances, it is configured as a multiple of a fixed amount and the number of blocks the bidder wishes to bid on, in others, it is the reservation price.

For sealed bid auction formats consisting of one round, it is equivalent to the bid placed.

This guarantee works as a security measure, such that if the participant refuses to pay the winning bid, the PTS will keep the guaranteed amount.

- A security check by the Swedish Security Police (SÄPO). It ensures the legitimacy of the companies participating, such that no subsidiaries of the participant companies are involved.

- Demands for coverage and maintenance of the network. Depending on the spectrum license in question, the PTS has varying coverage requirements that must be met, regardless if it is profitable or not for the companies. It is important to note that coverage is much more than just providing a signal in a specified geographic area; the strength of the signal must meet individual requirements

According to the EU-Commission, the PTS, like all other European regulatory authorities, have a responsibility to regulate companies with significant market power (SMP)9. SMP- regulations aim to create a long-term sustainable competition on the market for electronic

8See Dnr: 13-7510 (2014, P.6)

9 See Konkurrensreglering (SMP): PTS. (2016, October 14)

(25)

communication. The PTS does not regulate the end price to consumers unless the companies have significant market power and are using it to overcharge consumers.

All companies who hold a spectrum licence in Sweden have to offer other operators access to their mobile network at a competitive price10. The competitive price should be such that it covers the costs and gives a reasonable return on capital to the company which rents the license. This way, smaller companies can provide their mobile services in the market, regardless if they own a spectrum license.

The Swedish spectrum market is a lucrative business for the companies who can meet the demands of the PTS, as there is limited price regulation for the end product sold by the companies to the consumer market. The PTS conducts annual market reports, the latest publicly accessible report11 regarding the market in 2018, states that revenues for the mobile communication market amounted to a total of 30.7 billion SEK. The high revenue streams generate an incentive for companies to acquire spectrum licenses regardless of the barriers.

3 Results

3.1 Single object auctions 2008-2017

PTS has historically allocated single block licenses using first and second-price auction models in allocating blocks from 1900MHz, 1800MHz and 450MHz frequencies, as shown in Table 2.

Table 2. Single object auctions held between 2008-201712

Year Bidder Frequency Auction Format Bid in Millions

2008 iBand AS 1900MHz First-price sealed bid 0,061 SEK

2016 Hi3G Access AB 1800MHz First-price sealed bid 100,05 SEK

2018 Netett 450MHz Second-price sealed bid 91,3 SEK

Note. A representation of the results of the single spectrum auction conducted by the PTS between 2008-2017. The table includes all participants who won, frequency, auction format and the winning bids.

10See Kalkylarbete mobilnät: PTS. (2016, October 13)

11Dnr: 18-39364, (2019, May 24),

12See Dnr: 08-418 (2018), Dnr: 16-9255 (2018) and Dnr:15-11708 (2016)

(26)

The revenues generated from the 1900MHz auction were significantly lower than what was generated from other auctions with similar formats. The reason behind this is that the spectrum was, in 2008, used for analogue television. The PTS revoked this license following the closure of the analogue network in Sweden. The administrative fees from the auction amounted to 130 thousand SEK, which were higher than the auction revenues. These fees were added on to the price that the winning bidder had to pay. The other participant of the auction submitted a significantly lower bid of only 25 250 SEK. One factor that might have had an impact on the low bids is the limited usability of the spectrum, which caused the reservation price to be 1 SEK13. The license is currently unused in Sweden, and therefore there is no need for a more in- depth analysis of it.

Telia AB and Net4Mobility AB, which is a joint venture between Telenor and Tele2, share the 1800MHz spectrum, both having seven blocks each out of the total of 14 blocks. In 2017 one of Net4Mobility's blocks expired which was then auctioned again by the PTS. The interest for this single block was low, a possible reason for this was that the operators did not need more blocks to meet the market demand.

The 1800MHz auction had a reserve price set at 40 million SEK. Hi3G Access AB (Tre) did not know that they were alone in the auction and submitted a bid of approximately 100 million SEK. The reason why they decided to participate in the auction is unclear; one reason might be that they were the only major operator without any access to the 1800MHz spectrum. We do not know if Tre placed a bid which is in line with their truthful valuation. However, the theories based on the first price sealed bid auctions suggest that Tre had no incentive to do so. It can be argued that the block was allocated correctly to the company that valued it the most, as there was only one company that saw it necessary to participate in the auction.

There is no empirical data, which allows us to deduce whether the usage of the 1800MHz spectrum acquired by Tre has been efficient. The accessible data is the coverage map14 provided by Tre, which shows the surface coverage provided. The operator has several spectrum licenses, and the map regarding the surface coverage covers all the licenses. The map indicates that their network is primarily limited to highly populated areas. A possible reason is that five out of seven spectrum licenses Tre15 holds, operate on a higher frequency range. Therefore, the

13See Dnr:08-418 (2018)

14 For fully dynamic coverage map see Täckningskarta Tre Sverige. (n.d.)

15 See Frekvenser för 5G, 4G, GSM och 3G. (n.d.).

(27)

expansion to remote areas of Sweden may not be cost-efficient, as the high range spectrums are utilised efficiently in highly populated areas where data capacity demands are high.

In 2014, PTS conducted a market study for the 450MHz spectrum; they concluded that there was a demand for mobile broadband services which operate under the 1GHz spectrums. The argument for this is that expansion and maintenance of a low-frequency network are relatively more cost-efficient compared to a high-frequency network16. They decided to auction a single license that covers the whole 450MHz spectrum, and the chosen format was second price sealed bid, with a reserve price of 10 million SEK. In this auction, there were two participants: Telia AB and Netett AB (now known and referred to as Net1). Net1 was the current license holder of the spectrum, and they had been developing their network for it since 2009.

Net1 won the auction by placing a bid of approximately 91 million SEK and paid the second- highest bid close to 40 million SEK submitted by Telia, which earned them a surplus of roughly 51 million SEK. As the dominant strategy in the second price is to bid truthfully, we can argue that according to theory, the 91 million SEK was the truthful valuation of the 450MHz license.

However, it remains unknown if the truthful valuation was the correct valuation as the model does not promote price discovery.

Figure 117. Graphical representation of 3G and 4G surface coverage in Sweden during 2017, for a data speed of 10Mbit/s, comparing the effect of Net1’s contribution to the total surface coverage in each county.

16See 16-9255 (2017)

17 For data used in Figure 1 see Appendix B 70,00%

75,00%

80,00%

85,00%

90,00%

95,00%

100,00%

Surface coverage for data transfer services of 10 Mbit/s 3G and 4G excluding NET1 (2017) Surface coverage for data transfer services of 10 Mbit/s 3G and 4G including NET1 (2017)

(28)

Studying the data provided by PTS in their 2018 report18, we found that 88,93% of Sweden was covered with a network that provides speeds up top 10Mbit/s. This figure excludes the coverage provided by Net1, which, when accounted for, increased the coverage to 94,40%. While other operators use multiple different spectrums to reach the coverage target, Net1 only uses the 450MHz. The network developed by Net1 provides coverage in more remote areas due to the strength of the signal, as demonstrated in Figure 1.

We deem that the outcome of the 450MHz auction was efficient. The licence was allocated to a company which has built a network, that has a significant impact on the surface coverage in remote areas of Sweden. Net1's expansion of the 450MHz network has been efficient, and in line with the expectations set by the PTS. The latter underpins our assumption that truthful bids result in positive societal outcomes.

When comparing the two different auction outcomes, we can see that the highest bids for both were very close to each other. The highest bid for 1800MHz was approximately 100 million SEK, which is nine million more than the 450MHz bid. The 9 million SEK difference in winning bids is more likely to be caused by preferences related to the area of application between the spectrums.

The incentive for truthful bidding, limiting collusive behaviour and promoting price discovery should be held as a priority when conduction spectrum auctions. The sealed bid second-price model meets two of these criteria, while falls short on promoting price discovery. The first- price model only limits collusion. The revenues in the formats differ significantly; however, according to the revenue equivalence theorem, the average revenues should be equal. According to theory, the second-price model is the better option for efficient allocation. Based on our limited data, we cannot empirically verify that the second-price outcome has been more efficient than the first-price.

3.2 Multiple object auctions 2003–2018

The PTS allocated 3G19 spectrum rights using a beauty contest, which resulted in very inefficient outcome. One glaring problem with the allocation is the inevitable creation of an oligopoly in the market, which are challenging to regulate. A non-cooperative oligopoly can discover mutual interdependence and collude tacitly by adjusting price levels. Such collusion

18 Dnr: 17-8445 (2018, Mars 29)

19Wessel, J. (2007, June 1). Faktablad 3G i Sverige

(29)

can result in higher end-consumer prices or the blocking of new entrants in the market, as they can set a price level which is less or equal to the average variable cost for a new operator.

Regulating such occurrences is done through antitrust laws. The laws are heavily reliant on showing documentation or proofs of the agreements between the operators, rather than showing the actual outcome of their actions on the market (Carlton & Perloff, 2016)

While it might seem brilliant at first to let the government pick and choose the most suitable candidates to build the network, it created a multitude of problems when the time came to construct the network. The companies selected did not build the network in the timeframe demanded and PTS was left to evaluate if it should issue fines to the operators for not meeting the demands. The operators promised to cover 99.98 per cent of the Swedish population by December 31, 2003. Fairly soon after the beauty contest was conducted, the operators realised that meeting such coverage demands were not feasible and applied for extensions, which PTS denied. As the deadline set by PTS passed, only 65 to 75 per cent of the population was covered.

PTS responded by lowering the quality needed and extended the deadline to February 28, 2005.

In March 2005, PTS conducted a new measurement and found that 86 to 87 per cent of the population was covered, which failed to meet the expectations of PTS. After lowering their demands once more, and giving permission for the companies to use alternative technologies to cover the remaining areas, 90 per cent of Sweden was covered by the end of 2005 (Welin, 2006).

If the fines were to be issued, the companies would have had to raise the prices they charged from the end-users, thus lowering the societal benefit. To add to the inefficiency of the beauty contest, as the telecom operators had not put in significant financial investments into the project, it was easy for them to opt-out from the project at any given time. Such was the case when Orange left the market in 2004, leaving one license unused.

The beauty contest is flawed according to all our evaluation criteria, as the companies do not pay anything for spectrum licenses; therefore, there is no price discovery or truthful bidding.

There is a high possibility of collusion, both from the business colluding with each other and with government representatives. It can be thus theoretically concluded that beauty contest is not an allocation method that should be used when allocating spectrums as it follows none of our critical criteria for efficient allocation. As shown in the research done by Welin (2006), the real-world outcome of the allocation proved to be in line with our theoretical analysis.

(30)

While the beauty contest was only used once in the allocation of spectrum rights, SMRA has been used on three different occasions, as shown in Table 3, Table 4 and Table 5.

Table 3. 2,6 GHz SMRA Auction 200820

Bidder Number of blocks Bid in Millions

Tele2 Sverige AB 8x5 MHz 548,1 SEK

Hi3G Access AB 4x5 MHz 296,6 SEK

TeliaSonera Mobile Networks AB 8x5 MHz 562,45 SEK

Telenor Sverige AB 8x5 MHz 533,05 SEK

Intel Capital Corporation 1x50 MHz 159,25 SEK

Total of winning bids 2 099,45 SEK

Note. A representation of the results of the 2.6GHz spectrum auction conducted by the PTS in 2008.

The table includes all participants who won, the number of blocks and the winning bids.

Table 4. 800MHz SMRA Auction 201121

Bidder Number of blocks Bid in Millions

Hi3G Access AB 2x10 MHz 431 SEK

Net4mobility HB 2x10 MHz 469 SEK

TeliaSonera Mobile Networks AB 2x10 MHz 854 SEK

Total of winning bids 1 754 SEK

Note. A representation of the results of the 800MHz spectrum auction conducted by the PTS in 2011.

The table includes all participants who won, the number of blocks and the winning bids.

Table 5. 700Mhz SMRA Auction 201822

Bidder Number of blocks Bid in Millions

TeliaSonera Mobile Networks AB 2x10 MHz 1 082,657 650SEK

Net4mobility HB 2x5 MHz 720,968 398 SEK

Net4mobility HB 2x5 MHz 720,968 398 SEK

Total of winning bids 2 524,594 446 SEK

Note. A representation of the results of the 700MHz spectrum auction conducted by the PTS in 2018.

The table includes all participants who won, the number of blocks and the winning bids.

20See Dnr: 08-417 (2008, January 17) and Dnr: 08-417 (2008, May 8)

21 See Dnr: 10-10534 (2010, December 13) and Dnr: 10-10534 (2011, April 4)

22 See Dnr: 17-9908 (2018, July 4) and Dnr: 17-9908 (2018, December 14)

Figure

Updating...

References

Related subjects :