18 May 07: Ofcom delays auction of 2.6GHz, 2010MHz and 2290MHz spectrum bands
In December 2006 Ofcom issued proposals to award licenses for three spectrum bands: the 2.6 GHz, the 2010 MHz, and the 2290 MHz bands. Ofcom identified four main services for which the bands might be used: (a) advanced mobile telephony services using 3G technologies; (b) broadband wireless services using WiMAX standards; (c) mobile multimedia services; and (d) programme making and special events (PMSE) services, primarily for digital video applications. Ofcom's intention is that the awarded licenses will be tradable. The 2.6 GHz and the 2010 MHz bands are to be awarded together in a single auction, and the 2290 MHz band will be auctioned separately.
One novel feature of the proposed 2.6 GHz and the 2010 MHz bands auction is that the division of the spectrum into 'paired' and 'unpaired' blocks (mobile services require paired blocks of spectrum, whereas Wimax operators require unpaired spectrum), is to be largely determined by the auction process itself. Another is the complex auction design chosen. Ofcom's proposed auction design is a modified version of a combinatorial 'clock-proxy' auction with three stages: a clock stage, a proxy-like "best-and-final-offers" stage, and an assignment stage.
Combinatorial clock-proxy auctions have been proposed in the recent auction literature, but they have never been used in practice, nor even subject to comprehensive laboratory testing. Other types of multi-round combinatorial auction, on the other hand, have been tested in laboratory settings, and multi-stage Anglo-Dutch auctions are currently being used to sell virtual power plant capacity in Denmark. Ofcom's consultation document makes little or no reference to such alternative auction designs, however.
Market Analysis submitted a paper commenting on the auction design, and many of our arguments were echoed in other responses to Ofcom's consultation. Some of the most important issues raised by Ofcom's auction design are:
1. The proposed clock stage "activity" rules prevent bidders from switching demand between paired and unpaired blocks of spectrum. A number of respondents argued that this was an unnecessary restriction which might lead to an inefficient allocation. Less restrictive rules could be used instead which would allow bidders to more fully express their preferences as the auction proceeds.
2. The proposed "best-and-final-offers" stage pricing rule does not fully specify how auction prices are determined from bids, and neither does it eliminate incentives for insincere bidding and manipulation. An incompletely specified pricing rule makes it difficult for bidders to predict the consequences of their bids, and is more likely to result in an inefficient allocation by creating uncertainty and distorting bidding incentives. Market Analysis argued that consideration should be given to replacing it with a single-round pay-as-bid auction, which determines prices unambiguously and is well-understood by bidders.
3. The rules which link the clock stage to the "best-and-final-offers" stage also appear to give rise to incentives for price manipulation, and potentially conflict with the dual objectives of price discovery (in the clock stage) and efficiency (in the proxy stage). These require further consideration.
4. The proposed assignment stage is a single-round pay-as-bid auction, which is likely to lead to inefficiency. It should either be replaced by a simultaneous multi-round action (SMRA) in which much more information is revealed to bidders, or dropped altogether in favour of a 38 clock auction in the first stage.
In light of the responses it received, Ofcom announced on 17 May that it would carry out further analysis and consultation before taking any decisions. Accordingly, Ofcom now intends to set out updated proposals in a further discussion document by the end of July 2007.
Given the importance of the auction for the 2.6 GHz and the 2010 MHz bands - the UK's largest single release of radio spectrum to date - Market Analysis has suggested that a number of alternative auction designs be given further consideration, and that they be subject to laboratory testing. Variants of Ofcom's proposed auction design should also be tested, since auction theory alone is not sufficient to determine the merits and demerits of the detailed auction rules in combinatorial settings. Given that Ofcom is proposing to use very similar auction designs for at least two other spectrum bands, the argument for laboratory testing is made even stronger.
The Market Analysis report on the proposed auction design can be found in "Reports and Submissions" on the Pubications section of our website.
For Ofcom's consultation document and related material see: http://www.ofcom.org.uk/radiocomms/spectrumawards/