doczalacznik_Formaty aukcji i przyklady ich zastosowania na
download 
Reklamacja Transkrypt
zalacznik_Formaty aukcji i przyklady ich zastosowania na
Załącznik do stanowiska: „Uwagi P4 Sp. z o.o. do projektu rozporządzenia w sprawie przetargu, aukcji oraz konkursu na rezerwację częstotliwości lub zasobów orbitralnych (projekt z dnia 15 lutego 2013 r.) („Rozporządzenie”)” Annex: Best Practice auction design - Case studies 1. Spectrum auctions need not necessarily be complex, but their detailed design is always important. Even small changes in auction rules can significantly affect outcomes, benefiting some parties at the expense of others and potentially lead to inefficient outcomes. Getting the design right is therefore crucial for achieving the underlying policy objectives. 2. However, there is no single design that is unambiguously best. The right format depends on the frequencies that are on offer, their candidate uses, market conditions and policy objectives. It is also important to understand that there is a close interrelationship between the way in which the available spectrum is packaged into lots, and the best auction format. Dividing up the spectrum into small lots can be great for efficiency. It allows bidders to assemble the frequencies that meet their specific needs. This means that different types of bidders, using different technologies and pursuing different business models, can compete on a level playing field. However, this only works if the auction format does not expose them to severe aggregation risks – the fear of ending up with only a subset of the lots that a bidder needs in order to run a viable business. This shows that there is a host of factors that need to be taken into account when designing a particular auction. As they may differ from award to award it is essential for UKE (as any other regulator) to remain flexible in terms of choice of format it wants to adopt for any one award. There is no one size fits all in auction design. 3. A number of different formats have been used for spectrum auctions in the past. Most auctions are conducted as either a single-round sealed bid process or a simultaneous ascending-bid process. 4. In a sealed-bid process, bidders submit one or more bids in a single round. The auctioneer then determines winners and prices based on the bids received. Simple sealed bids where a single item is going to the highest bidder are a typical example, and the format extends easily to cases where there are multiple lots but each bidder can win only one. However, allowing bidders to assemble different lots makes matters more complex, and we will discuss the sealed-bid combinatorial auction as an example of a sealed-bid process that might be used in this case. This format consists of a single round of bidding in which bidders can make bids for any combination of lots subject to underlying spectrum caps and relevant reserve prices. This format was implemented for the Nigerian FWA auction, the UK 412-414MHz/422-424MHz auction, the Irish 26GHz band auction and the Portuguese 3.4-3.8GHz band auction. 5. In open ascending-bid processes bidding is organised in rounds. In each round, bidders can bid on one or more items simultaneously, and prices for items where demand from bidders exceeds supply continue to rise. Bidding closes when certain pre-defined criteria are met (and sometimes the auctioneer may have the right to call a ‘final round’). Two auction formats are prevalent in this category: • The simultaneous multi-round ascending (SMRA) auction and its variants. The 'standard SMRA' was first developed and used by the US FCC and subsequently adopted by many other regulators. Bidders submit multiple bids in each round for individual lots and can shift their demand between lots over successive rounds, subject to certain activity rules. This format was used recently for a number of spectrum awards in Hong Kong (e.g. the 2.6GHz award in 2009), in Germany (2010), Belgium (2011) and Spain (2011), amongst others. A variant of the SMRA that has become popular with Scandinavian regulators is the ‘SMRA with augmented switching’. This format was used by Norway (2008), Sweden (2008) and Finland (2009) for their 2.6GHz awards. The format allows bidders greater flexibility to switch across substitutable lots than the standard SMRA, thus reducing fragmentation risks for bidders seeking contiguous spectrum. • The ‘combinatorial clock auction’ (CCA). This format provides a practical design for package bidding with either specific or generic lots over multiple rounds. The CCA, developed by Ofcom, the UK regulator, for the UK 10-40GHz, L-band and 2.6GHz awards eliminates aggregation risks for bidders. There seems to be a trend away from the standard SMRA auction format in Europe. Many European regulators have adopted the CCA for recent awards in the 800MHz and 2.6GHz bands (e.g. Dutch 2.6GHz auction (2010) and 800MHz auction (2012), Danish 2.6GHz auction (2010) and 800MHz auction (2012), Austrian 2.6GHz auction (2010), Irish 800MHz, 900MHz, 1.8GHz and 2.1GHz auction (2012) and the UK 800MHz and 2.6GHz auction (2013). 6. Another format that has occasionally been used in spectrum awards (e.g. in the 2005 GSM auction in Trinidad and Tobago, or – in modified form – in the recent Indian 3G and BWA auctions) is the simple clock auction. As in the CCA, bidding takes place over a number of clock rounds but is not followed by a supplementary round, and bidders simply pay the clock prices at the end of the auction. However, this format is more prevalent in other contexts such as the sale of permits in the context of emission trading schemes. 7. Although there are package-bid SMRA designs (such as the SMRA with hierarchical packages, used in the US 700MHz auction), these have not been widely implemented for spectrum auctions and are likely to be unduly complex to implement. We therefore do not consider these alternatives. 8. In the following subsections, we describe the most prevalent formats in more detail, highlighting the key rules that are needed to make them work (although their concrete implementation may vary) and identifying their relative merits. A.1 Simultaneous ascending-bid auction formats A.1.1 Standard SMRA 9. This is the most common and widely used version of the SMRA, pioneered by the US FCC in the 1990s. The Bundesnetzagentur, the German regulator, used this format to award spectrum in the 800MHz, 1.8GHz, 2.1GHz and 2.6GHz band in 2010, and it was also used recently in Spain (multi-band award) and Belgium (3G and 2.6GHz awards). Mechanics of the auction 10. As with all SMRA formats, the standard SMRA takes place over multiple rounds, with a number of lots sold simultaneously and the auction closing on all lots simultaneously when there is no new bid activity (i.e. when no new bids are received and no waivers are played). In an SMRA auction, bidders bid for frequency blocks, each of which we call a lot. The auction consists of a number of rounds. Bidders may submit a bid in each round. Bidders are allowed to bid for as many lots as they wish, subject to any caps on their bidding defined before the commencement of the auction, and subject to activity rules that provide incentives for bidders to reveal their demand in order to aide price discovery. 11. When a round is scheduled, the auctioneer announces a price for each specific lot. This can be a fixed amount, a menu of bid amount options or a minimum amount. In the case of fixed amounts, bidders then specify whether they wish to place a bid for a lot at the specified bid amount. Otherwise they select the bid amount they would like to bid from the menu or set the bid amount freely subject to the current minimum amount. 12. At the end of the round, the highest bid on each lot (submitted in that or any previous round) becomes the standing high bid on the lot (a tiebreaking rule is applied where there is more than one bid at the highest bid amount), and the bidder that submitted this bid is declared the ‘standing high bidder’ on that lot. 13. In the following rounds, prices are increased only on the specific lots that received at least one new bid in the previous round. Standing high bidders may be out-bid by other bidders submitting higher bids. Bidders may also be allowed (depending on the auction rules) to withdraw their standing high bids if they wish to switch their demand to a different lot. There are usually tight limits on the number of withdrawals allowed and often also financial penalties if a withdrawal then leads to a lot not being sold. Allowing for the withdrawal of bids mitigates aggregation risks and the risk of a fragmented outcome (if specific lots are used). However, allowing bidders to withdraw bids might also allow for strategic bidding that may distort the auction outcome. In order to provide flexibility to bidders and also discourage collusive behaviour, withdrawals can be permitted subject to penalties, or limited to a specified number of withdrawals allowed per bidder. 14. A standard eligibility-points based activity rule is normally used to determine the bidding rights for participants during the course of the auction. The rule essentially ensures that bidders cannot increase demand as prices rise. Such behaviour would be in violation of standard economic theory, and allowing bidders to ‘hide’ their demand would undermine the effectiveness of open multi-round formats in assisting price discovery, but in order to provide some flexibility the activity rule often includes stagedactivity requirements to control the pace of the auction and to allow bidders to remain flexible at the start of the auction. 15. In order for a round to constitute the last round of the auction, the activity requirement must be at 100% and the round must close with no new bids being placed, no withdrawals being made for any lots and no waivers being submitted. 16. When the auction ends, each lot is awarded to the bidder that holds the standing high bid on the lot at the price of its standing high bid. Advantages and disadvantages of the SMRA auction format 17. The SMRA auction format has a number of positive qualities: • In its basic form, this auction format is relatively simple, although it can become more complicated by the introduction of withdrawals and a staged activity requirement. • Introducing a staged activity requirement allows bidders the flexibility to hold back a proportion of demand and corresponding expenditure until after a certain amount of information regarding prices is revealed. • As bidders are only liable for lots on which they are standing high bidder at any given time and can only become liable for further lots where they place new bids, there is a high degree of clarity and certainty as to the level of expenditure committed by the bidder at all times during the auction. Together with simplicity, this feature may encourage the participation of interested parties that might be deterred as a result of fixed budgets or limited resources to devote to a bidding strategy. 18. However, the standard SMRA format has a number of quite significant drawbacks: • The greatest problem associated with SMRA auctions is that this auction format is poorly suited to deal with aggregation risks when spectrum is offered in smaller blocks. In a standard SMRA auction, bidders bidding on a combination of lots may be exposed to the risk of ending up being the standing high bidder for some but not all of the lots on which they wished to win. In this case, bidders may be ‘stranded’ on a subset of the combination of lots they wanted. A predefined packaging approach may mitigate this as bidders do not run the risk of ending up on only part of the frequencies they need. However, if these larger lots are not consistent with bidders’ demand profiles and are thus predefined incorrectly, an efficient outcome cannot be achieved. • Aggregation risks may be somewhat mitigated by introducing the possibility of withdrawing bids. However, unless penalties are applied, allowing withdrawal of bids may create incentives for strategic bidding and subsequent withdrawal. Conversely, where penalties on withdrawals are applied, bidders may still be subject to a cost for withdrawing bids from unwanted lots. • There is a risk of lots inefficiently going unsold following withdrawal of bids by standing high bidders. For example, a bidder might withdraw a standing high bid at a point in the auction when all other bidders that might want the lot have already lost their eligibility to bid for that lot. • Where bidders have an interest in specific lots, this can facilitate a collusive outcome where these operators do not bid on one another’s currently held lots and vice versa. A typical SMRA has a high degree of transparency and it is very easy to formulate gaming strategies aimed at reducing competition and trying to establish tacitly collusive arrangements. SMRAs in both the US and Canada with regional licence structures have been plagued by this problem. It is possible to limit transparency to reduce this problem (e.g. by not revealing the identity of the standing high bidders), but this makes the problem of aggregation risks worse as bidders have less information to assess the chances of being stranded as the highest bidder. 19. Aggregation risks can be addressed in an SMRA by following a predefined packaging approach. Under this approach, bands are split into larger blocks of spectrum. This, however, predetermines the outcome and if the efficient allocation is actually different to the pre-determined split, the outcome will be inefficient. Case study 1: The European 3G auctions 20. The standard SMRA was the format of choice for the wave of 3G auctions in Europe in the 2000s. In these auctions, bidders were normally only allowed to bid on a single lot, so there were no aggregation risks. Under these circumstances, the SMRA generally performed well (although in the case of Switzerland, for example, bidders realised that they would benefit from avoiding competition, which lead to a process of consolidation that eventually meant that the auction ended after a single round with lots going at reserve price). 21. In the UK, five licences were to be awarded with one being reserved for an entrant. This created significant interest from new players leading to strong competition in the award. In contrast, in the Netherlands no spectrum was reserved for an entrant. This meant that new players did not see themselves winning spectrum in the auction, so they signed deals with existing operators to use their networks instead. This meant that only the five incumbents and one weak new entrant competed for the five licences. In this situation, it might have been preferable not to disclose the number of participants in the auction and provide no information about aggregate demand during the bidding rounds.1 22. In Germany, a more flexible approach was adopted, in which bidders could bid for multiple paired 5MHz blocks (subject to certain caps). This meant that the number of operators was determined endogenously. The auction produced six winners – but two of them with unviable business cases, leading to the spectrum being returned and re-auctioned in 2010. A.1.2 SMRA with augmented switching 23. This is a variation of the SMRA that aims to reduce fragmentation risks when specific lots are sold. This auction format has been pioneered in Norway and Sweden. As with the standard SMRA, bidding takes place over multiple rounds, with all lots sold simultaneously and the auction closing on all lots simultaneously when there is no new bidder activity. 24. This auction format would use the same eligibility points-based activity rule as described for the standard format. However, because of the much 1 For more information about the European 3G auctions, see Klemperer, 2001, "How (not) to run auctions: The European 3G telecom auctions", available online at: http://economics.ouls.ox.ac.uk/11928/1/hownot.pdf. greater importance of switching, more care needs to be given to the setting the eligibility point weights than for a simple SMRA. Mechanics of the auction 25. In the first round, bidders place an initial set of bids on lots. The activity of each bidder in the first round is equal to the sum of the eligibility points associated to the lots on which the bidder bids in the first round, and cannot exceed the bidder’s initial eligibility. 26. At the end of each round, each bid is ranked according to its value. Bids of equal value are given a ranking order by the tie-breaking method. The bid ranked first is designated the standing high bid on the lot, and the bidder who submitted the bid is the standing high bidder on the lot. 27. Standing high bids can be withdrawn (subject to the activity rules described below) provided they are switched to bids on other lots. In the case that the standing high bidder on a lot withdraws its standing high bid in a round, and no new or raised bids are received for that lot in the same round, the bid previously ranked second will become the standing high bid, and the bidder who submitted that bid will become the standing high bidder.2 This means that bids on lots that are not the standing high bid in any given round can become the standing high bid in the further course of the auction if all bidders with higher bids on the lot switch their bids to other lots. In this case, the bid is said to have become re-activated. 28. Because any bid made during the auction can effectively be reactivated (unless it has been withdrawn), bidders are committed to all their bids made throughout the auction, regardless of whether they are standing high bidder. At the end of the auction, bidders will only win the lots on which they are standing high bidder. However, they may become the standing high bidder as a result of withdrawals by other bidders. 29. The activity of each bidder in the first round will determine the number of committed bids and the level of eligibility associated with these committed bids. 30. In subsequent rounds, bidders can maintain, raise or switch their bids during the auction in accordance with the activity rules. Bidders can potentially change their set of bids in each round. However, the activity rules ensure that the eligibility points associated with the set of bids of each bidder remains constant throughout the auction. 31. The level of activity in each round after the first is calculated as: 2 If no other bids had been made previously on a lot on which a standing high bid is withdrawn, the lot reverts to the auctioneer. • the eligibility associated with lots on which the bidder is raising a bid; plus • the eligibility associated with lots on which the bidder is leaving a standing high bid unchanged; plus • the eligibility associated with lots on which the bidder is placing a new bid; minus • the eligibility associated with lots from which the bidder is withdrawing an existing bid (in order to switch it to a lot on which it does not hold a bid). 32. In any round after the first, the following activity rules apply: • the eligibility of the bidder is calculated as its activity in the previous round plus the eligibility associated with lots for which the bidder’s bids have been reactivated (and become the standing high bid) due to withdrawals by other bidders;3 • the activity of the bidder cannot exceed its eligibility in that round; and • the eligibility associated with lots from which the bidder is withdrawing bids must be exactly equal to the eligibility associated to lots on which the bidder places new bids. 33. The auction may only close when there is no bidding activity (i.e. no bids are raised or switched) and no waivers are used. The auctioneer may also bring the auction to an end by calling a last and final round, during which bidders can submit their final bids. As with the standard SMRA, when the auction closes, each lot is assigned to the highest bidder, at the amount of their final bid for that lot. Advantages and disadvantages of the SMRA format with augmented switching 34. 3 Relative to the standard SMRA, the augmented switching format reduces fragmentation risks for bidders (especially if compared with a standard SMRA with no withdrawals). Allowing bidders to switch all their bids without any associated penalties allows them to move their full sets of bids in order to target contiguous spectrum more effectively. This means that where bidders want a number of lots in a contiguous frequency block, The rules may include a provision for a minimum level of eligibility to which the eligibility of a bidder is increased if bids are reactivated in order to ensure that a bidder who had lost eligibility before having one or more of their bids reactivated has enough eligibility to bid for sufficiently large combinations of lots to win usable spectrum. they can switch around in response to price differences and pursue substitution strategies. 35. However, SMRA format with augmented switching has many of the original problems of the traditional SMRA and some additional ones as well. The main shortcomings of this auction format are as follows: • The bidding process is complex. The switching rules associated with this format are complex and not very intuitive. In particular, when bidders are at a stage where they have dropped eligibility relative to their initial eligibility, it may be difficult for bidders to understand the limitations on what bids they can switch and what bids they can raise. Bidding in an auction with an SMRA format with augmented switching becomes particularly difficult when bidders can bid on a large number of lots, and when lots have different numbers of eligibility points associated with them, as there are numerous combinations of withdrawals and bids but only a few options conform with the auction rules. • It is difficult for bidders to bid within their budget constraints. As all bids may be re-activated at any time during the auction due to withdrawals submitted by other bidders, it is difficult for bidders to contract demand in response to price increases in a manner that truly reflects their budget constraints. This is illustrated in the following example: Suppose two lots are being auctioned, A and B and that a bidder may be willing to bid 16 for winning both lots A and B, and 10 for winning a single lot (either A or B), with a budget constraint of 16. Straightforward bidding would allow such a bidder to bid for two lots until the price of a lot exceeded 8, at which point it would reduce demand to one lot, and cease bidding on any lots once the price per lot exceeded 10. Now consider the SMRA format with augmented switching. Assume that the bid amounts for A and B have reached 8, and that the bidder has raised its bids up to this amount. In subsequent rounds, the bidder will need to reduce its demand, as bidding for both A and B would not be within the bidder’s budget. Suppose that the bidder then raises its bid for A alone, and that bid amounts for both lots go up to 10. The bidder is willing to acquire lot A at this price, which also is within its budget constraint. However, if in the following round the bidder’s bid for B is re-activated, then the bidder is liable for a total amount of 18 (10 from its bid for lot A plus 8 from its bid for lot B). In this case, the bidder would be over its budget constraint. If bidders have flexible budget constraints, they may be prepared to risk such situations. However, bidders with fixed budget constraints may have to drop out at an earlier stage in order to ensure that it does not end up in such a position. In the example discussed, the bidder would have had to drop out from the auction when the bid amounts reached 8 for each lot, even though the bidder had a valuation of 10 for one of those lots alone. • The SMRA with augmented switching has been proposed on many occasions for awards where bidders might be exposed to the risk of undesirable assignment outcomes. However, while the risk of fragmentation is mitigated with this variant of the SMRA format, it is not removed altogether. • The aggregation risks are just as much of a problem as with the standard SMRA. There is still a danger that at the end of the auction, a bidder is left with some, but not all, of its target lots and is unable to exit cleanly. 36. Overall, although the SMRA with augmented switching has desirable features and may effectively help to mitigate bidder fragmentation risks, we do not consider this to be the best way to address such risks within an SMRA format. It does not get the heart of the problem with the SMRA: that there are standing high bidders until such time as they are overbid by someone else and that this creates aggregation risk. The risk of inefficient fragmentation may better be addressed by offering generic lots which are then assigned either by operator agreement (with regulatory intervention as a back-stop, as in the case of the German multi-band auction), or through a separate bidding stage for specific frequency positions subject to allowing only contiguous assignments. Case study 2: Small details matter - The Finnish 2.6GHz auction (2009) 37. For the award of 2.6GHz spectrum, FICORA had chosen simply to borrow the format and rules previously used by PTS for the 2008 Swedish 2.5GHz auction. The rules as initially drafted, contained some small, but significant, omissions. These were fixed in the case of the Swedish auction. FICORA, on the other hand, decided that it was not important to address these omissions. 38. The rules governing price increases in each round were drafted such that the price of a lot would be set to the level of the standing high bid plus a bid increment. Hence, if a standing high bid is cancelled and a lower priced bid is promoted to standing high bid, the new bid increment would be based on the new standing high bid, not the previous one. This means that prices for lots can drop in the auction. 39. The rules allowed a bidder to cancel a bid on a lot and then return to the same lot later, potentially at a much lower price. This means the auction can end up in an infinite cycle in which bidders bid on certain lots, then cancel their bids on these lots and bid on other lots and then bid back on the initial lots when their price has decreased to reserve. Not unexpectedly, the auction finished with prices for 2.6GHz paired spectrum close to reserve. The Finnish auction (2009) achieved a price per MHz per population of € 0.003 which is much lower than the €0.159 achieved with the "fixed" version of the auction format in Sweden (2008).4 A.1.3 Combinatorial clock auction 40. The combinatorial clock auction (CCA) is a practical package bidding format, pioneered by Ofcom in the United Kingdom. It is an adaptation of the clock auction format combining the multiple-round process typical for a clock auction, where a bidder effectively makes a package bid in each round, with a supplementary round in which bidders can to express their valuations for an arbitrarily wide range of packages. 41. The CCA format is very flexible and can be adapted to cope with any situation where bidders face varying aggregation risks relating to lots either within or across bands. With the introduction of package bidding, this format is best suited in combination with a flexible packaging approach as bidders will be guaranteed to win their full packages at a price at or below their bid for this package. 42. The auction itself would consist of two stages, namely a principal stage and an assignment stage. 43. In the principal stage there are two phases of bidding, which determine the identity of the winning bidders and the number of generic lots awarded to each bidder in each category: • Primary rounds. The first phase consists of primary rounds, which follow a clock auction format. Bidders make a single bid each round for a package of lots across the various categories in response to a set of prices (one for each category) set by the auctioneer. The primary bid rounds continue until there is no excess demand in any category. • Supplementary round. This is a single round sealed bid process, in which bidders have the opportunity to make multiple, mutually exclusive bids for packages of lots across categories, subject to constraints created by their primary bids. Once the round is completed, the auctioneer identifies the highest value combination of bids that can be accommodated and prices for each winning bid. 44. 4 The assignment stage determines the actual frequencies in the various bands and at the potentially different times that are assigned to the winning bidders from the principal stage. Bidders participate in parallel See Richard Marsden, Eimear Sexton and Arisa Siong, June 2010, ‘Fixed or flexible? – A survey of 2.6GHz awards’, DotEcon discussion paper 10/01, available online at: http://www.dotecon.com/assets/images/dp1001.pdf. sealed bid auctions, one each for each category, and can make ‘assignment bids’ for particular ranges of frequencies compatible with the number of lots that they won in the principal stage. The auctioneer then identifies the highest value combination of bids that can be accommodated. The final price to be paid by each winner is the sum of the base price and any additional prices from the assignment stage. Mechanics 45. The primary rounds follow a ‘clock auction’ format. In the first primary round, the auctioneer sets a price for each category of lot. The bidder then states its demand for each category of lot based on these prices (subject, where applicable, to bidding constraints). 46. If there is excess demand for any category of lot, the auctioneer will schedule another round. For this round, the price set by the auctioneer for categories of lots that had excess demand in the previous round will be increased for the scheduled round in line with some pricing rule set out before the commencement of the auction. The price set by the auctioneer for categories of lots that did not have excess demand in the previous round will generally be unchanged for the scheduled round. 47. A standard eligibility-points based activity rule is used to determine the bidding rights for participants during the course of the primary rounds. Such an activity rule allows bidders to switch demand between different categories during the course of the primary rounds. 48. During the scheduled round, bidders state their demand for each category of lot based on these new round prices. This process continues until there is a round during which the aggregate demand for each category of lot can be met by the number of lots in that category. When the primary bid rounds end, the auction will progress to the supplementary round. 49. The supplementary round is a one-off further round of bidding following the primary bid rounds which provides an opportunity for bidders to: • express their full value for the package that they were bidding on at the end of the primary bid rounds; and • to bid for packages of lots that they were eligible to bid for in the primary bid rounds but that they did not bid for. This allows bidders to place multiple bids that they did not have the opportunity to place during the primary rounds. Also, they may not have reached their full valuation for their most preferred package in the primary rounds given round prices, and there may be other packages that they have a lower value for but would still like to win at certain prices. Bids in this round are subject to constraints; bids will be subject to a minimum and in some cases a maximum based on the bidder’s primary bids. These constraints depend on the specific rules of the auction (which are quite complex and not explained here). The intention of the constraints is to provide incentives for truthful bidding throughout the auction. 50. At the end of the supplementary round, winners and prices are determined. Bids considered include all bids submitted in the primary bid rounds and all valid bids submitted in the supplementary round. The combination of bids (at most one from each bidder) with maximum value that fits into the available supply is selected as the winning combination. 51. The next step is to determine prices to be paid by winning bids. The details of the pricing rule for CCAs to date are somewhat complex, but the basic principle is that the price paid is determined by opportunity costs, not by what bidders actually bid.5 These auctions use a generalised notion of opportunity cost that ensure that each winning bid pays at least its opportunity cost, but also each and every group of winning bidders collectively pays at least its opportunity cost. The effect of this rule is that winning bidders pay not what they bid, but the smallest amount that, if they had bid that amount instead, they would still have won.6 52. At the end of the supplementary round, the number of generic lots assigned to each bidder is determined. However, the actual frequency blocks corresponding to each lot awarded are assigned to winning bidders in a follow-up assignment process. 53. The assignment stage typically consists of a single round where winners in the auction are presented with the different feasible frequency ranges they could be awarded given their winnings. This stage provides winners the opportunity to express their relative preferences (if any) as to which assignment option they are awarded. Winners can do this by placing a simple unconstrained bid for one or more of the feasible assignment options they have been presented with. 54. After the assignment round closes, the location of winners within the band is determined by choosing the set of frequency assignments among winners that maximises the total winning bids. Prices to be paid by winners are determined using a similar pricing rules as in the principal stage. 5 For an explanation of the detailed algorithm to find winners and prices in a CCA, see http://www.dotecon.com/assets/images/dp0701.pdf. 6 This form of pricing rule provides good incentives for bidders to bid close to their true value. Whilst it is theoretically possible to reduce the amount paid by shading down bids in some cases, in most practical applications it is very difficult for bidders to assess the implications of bidding less than their true value for the risk of losing, as they do not know the bids of other bidders. The incentive to bid close to their true value, together with the winner determination step that optimises the outcome given the bids received, should lead to an efficient outcome. Advantages and disadvantages of the CCA 55. Given that the CCA auction format allows bidders to express demand for different combinations of lots, and each bid is only considered in its entirety (i.e. bidders will not be awarded only a subset of the package they bid for), this format allows bidder to express their full value for lot packages without facing aggregation risks. 56. The CCA provides an effective way of dealing with bidder aggregation and fragmentation risks by allowing for package bidding. As mentioned above, bidders may be awarded a package they bid for, but not any subset of that package for which they did not place a bid. Given that bidders are not exposed to being awarded only a subset of the lots they bid upon, therefore, they can safely express their full synergy values for different combinations of lots. 57. Another advantage of combinatorial auctions is that, provided that appropriate rules have been designed, package bid auctions may also support a more complete expression of bidder demand and provide less opportunity for strategic behaviour. This simplifies bidding, as bidders can focus on expressing their valuations rather than the implications of other strategic actions. This increases the likelihood of an efficient outcome. 58. The CCA format is very flexible and can be adapted to cope with a multitude of situations where bidders are competing for different amounts of spectrum and want to deploy different services and technologies. 59. Although the format can accommodate either specific or generic lots, its advantages are clearest in situations where lots are organised into a small number of categories, with each category containing generic lots that can be treated as identical for the purposes of the auction: • With large numbers of lots in different categories, the number of package options that a bidder may value is likely to be very large. This creates complexity for bidders and may also cause problems for auction implementation when the number of possible lot combinations becomes unmanageable. • There may be different technologies that bidders could deploy in a given band and the traditional approach of packaging spectrum into fixed size lots in anticipation of a particular technology and business model being used may be inappropriate. The CCA provides a highly flexible approach in this regard. • By removing the additional complexity to bidders of placing bids in order to target specific blocks (particularly when trying to ensure contiguity of multiple lots), the bidding process becomes easier and bidders can focus on the size of the package they wish to acquire during the principal stage, and consider only its feasible assignment options in the assignment round. 60. One drawback of the generic approach is that bidders cannot express any preferences for specific frequencies in their primary and supplementary bids. Because they can only express their valuation for generic lots, bidders might find it difficult to assess how much to bid if the difference in the value of specific lots within each generic lot category is large, because of the uncertainty about the specific frequencies that they will eventually win. If this problem is severe, there is the risk of an inefficient allocation. Therefore, it is important to define lot categories such that lots within each generic lot category are sufficiently substitutable so that the value difference across the different conceivable assignments that successful bidders may obtain are small. 61. Another drawback of auctions with package bidding, including the case where there are generic lots and particularly where a second price rule is employed is that, depending on the number of lots, such package bidding may introduce complexity for bidders. Under such circumstances, the outcome, even if efficient, may not be as transparent to bidders and observers as a standard SMRA, owing to the complexity of the mechanism used to identify winning bids and prices. Case study 3: Creating competition - Danish 2.6GHz CCA 62. Demand reduction is a potential strategy in an SMRA (and a standard clock auction) to keep prices down. Incentives for demand reduction tend to be strongest in any multi-round format that uses a pay-your-bid approach.7 This has a particular impact on bidders who expect that they will ultimately have to settle for a small number of lots (weaker incumbents or new entrants) as any attempt to compete for additional spectrum would ultimately increase the price they pay for the lots they do win. This will lead to inefficient outcomes where a weaker bidder might ultimately have been able to acquire additional spectrum, but refrains from competing for it for fear of having to drop back and pay more than necessary as a result of having tried. 63. Demand reduction incentives are much weaker in the case of the standard CCA, as the price a bidder pays is, in most cases, unaffected by its own bids. Consequently, weaker bidders can keep bidding for larger packages 7 Engelbrecht-Wiggans and Kahn (1998) and Brusco and Lopomo (2002) have highlighted this problem in many SMRA formats (see Engelbrecht-Wiggans, R. and Kahn, C.M. (1998), “Multi-Unit Auctions with Uniform Prices”, Economic Theory, 12, pp.227-258; Brusco, S. and Lopomo, G. (2002), “Collusion via Signalling in Simultaneous Ascending Bid Auctions with Heterogeneous Objects, With and Without Complementarities”, Review of Economic Studies, 69, pp.407-436). until they no longer provide value, as these bids do not affect what they pay if they later drop back to smaller packages. 64. The recent Danish 2.6GHz auction provides a good example of how the CCA may be more effective than the SMRA in protecting awards from the effects of demand reduction. This award took place one year after a similar award in neighbouring Sweden and had a similar cast of bidders. Specifically, in both auctions, the only bidders for paired spectrum were the four incumbents, with aggregate demand of 16 lots (4 lots of 2x5MHz each) against 14 lots supply. Sweden used an SMRA format, which ended with healthy prices after H3G eventually dropped back from 4 to 2 lots. 65. In Denmark, H3G was again likely to be the marginal bidder. Had Denmark used an SMRA format, a good strategy for H3G would likely have been to drop immediately to 2 lots, allowing all bidders to win spectrum at the reserve price; this would be a sensible approach in the light of the Swedish outcome. The assumption here is that H3G could predict with a high degree of certainty that it would be the marginal bidder, and that therefore its willingness to pay for an additional 4th and 3rd lot would determine the prices for the 1st and 2nd lot they would most likely end up winning. In these circumstances, they could achieve a much better outcome by dropping demand immediately rather than bidding up to their marginal value for the 4th and 3rd lots. 66. Instead of using an SMRA, Denmark adopted a standard CCA. As a result, H3G was able to bid up to its maximum willingness to pay for 4 lots, thus testing out whether it was indeed the marginal bidder, before dropping back to 2 lots without any price penalty. There was no price penalty because the price for the package of 2 lots for which H3G eventually settled was unrelated to its package bids for 4 lots, as it was determined based on the opportunity cost imposed on other bidders when allocating these two lots to H3G. This was the reserve price as bidders were allowed to win at most 4 lots and hence the two lots for which H3G eventually settled were uncontested by the other bidders. Together, the other bidders were only able to win 12 lots together meaning that H3G knew that they can always get two lots at the reserve price once they decide to drop back.8 The other bidders were, however, obliged to pay the full opportunity cost of denying H3G its preferred allocation of 4 lots, and revenues were similar to Sweden. The average price per MHz per 8 See NITA’s website for the details of the award: http://en.itst.dk/spectrum-equipment/Auctionsand-calls-for-tenders/2-5-ghz/public-consultation-over-draft-2-5-ghz-auction-documents. population achieved in Denmark was €0.175 compared to €0.159 in Sweden.9 1.1 Sealed bid combinatorial auction 67. DotEcon first implemented a sealed bid combinatorial auction for the award of fixed wireless access (FWA) spectrum in Nigeria in 2002. This format was also used for the award of the 412-414 MHz paired with 422424MHz band in the UK (2006), for the award of spectrum in the 26GHz band in Ireland (2008) and the award of spectrum in the 3.4-3.8GHz band in Portugal (2009). 68. The sealed bid combinatorial auction is, in effect, a CCA without the primary rounds. Bidders can place bids on as many different combinations of packages as they wish, but these bids are collected in a single round with no bidder having visibility of the other bids made. These bids are constrained only by underlying spectrum caps and a minimum of the relevant reserve prices. Unlike our three previous candidate auction formats, the sealed bid combinatorial auction does not provide for price discovery. Instead, bidders have only one opportunity to submit their best bids for the lots auctioned, and the winning bids and bidders are determined on the basis of just one round of bidding. Mechanics of the auction 69. A sealed-bid combinatorial auction consists of a single round where bidders are invited to submit their final bids for specified lot packages (combinations) that they would like to win. Each bid is exclusive, meaning that at most one bid from any bidder will be accepted, and is only considered in its entirety (so bidders may be awarded the whole package they bid for, but not any subset of lots they did not place a bid for). 70. The winner determination process is essentially the same as for the combinatorial clock auction. The winning bids are the set of bids amongst all bids received that maximise the total of winning bids, subject to no more lots than are available being sold. Prices are determined using a generalised notion of opportunity cost. This pricing rule means that bidders have good incentives to bid close to value. 71. As with the CCA auction format, the auction may be structured so that bidders bid for generic lots, and the assignment of particular frequencies to winners may be determined in a follow-up assignment process. The 9 See Richard Marsden, Eimear Sexton and Arisa Siong, June 2010, ‘Fixed or flexible? – A survey of 2.6GHz awards’, DotEcon discussion paper 10/01, available online at: http://www.dotecon.com/assets/images/dp1001.pdf. assignment stage may be designed so that bidders can express their preferences over alternative (feasible) spectrum frequency plans. Advantages and disadvantages of the sealed bid combinatorial auction 72. The sealed bid combinatorial auction format offers many of the same efficiency advantages as the CCA relative to SMRA formats: • By allowing package bidding, it removes bidder aggregation risks entirely. • If generic lots are used, and the assignment of particular frequencies is determined in an assignment stage, it can support an efficient assignment of frequencies from a spectrum management perspective. 73. The sealed bid combinatorial auction has some additional advantages over the CCA: • The process is quick to implement, as it requires just one round to determine the winning bidders (and potentially a further round to determine assignment of specific lots if specific frequencies are assigned in a follow-up assignment stage). • The bidding process is simplified (e.g. bidding can be completed on paper forms delivered in sealed envelopes, or by electronic data files), thus reducing the costs to both bidders and the auctioneer. • This format is the least vulnerable to strategic behaviour, especially tacit collusion as bidders cannot observe each other’s behaviour over multiple rounds. Further, concerns about predatory bidding are also eased because entrants know that strong bidders do not have the opportunity to revise their business case during the auction in order to out-bid them. 74. As with the CCA, the process may be more complex and seem less transparent than the traditional SMRA auction format. In addition, the sealed bid combinatorial auction format has the disadvantage that it has no price discovery mechanism. Multi-round auctions, whereby bidders are able to submit and raise their bids over a number of rounds, allow bidders to process the information made available at the end of each round in order to update their valuations.
