The little things that matter

Why do the same frequencies, offered through very similarly structured auction processes at roughly the same time fetch rather different prices? Differences between the countries that awarded these frequencies may explain some of it, but as is generally the case with auctions, the details of the design matter as well, as this article explains.

The next generation

A decade ago, the award of 3G spectrum across Europe started the mobile data revolution that brought us smart phones and iPads, created the promise of working untethered from one’s desk and of social networking on the move. In order to realise the full potential of these developments, more spectrum is required, and the current wave of awards of 4G spectrum across Europe is the basis for new technologies (such as ‘Long-Term Evolution’, or LTE) that will provide the capacity needed in order to make more data-intensive services available to a greater group of users. The new spectrum lies mainly in the 2.6GHz band, and the 800MHz band that has been freed as a result of the move from analogue to digital terrestrial transmission (the so-called ‘digital dividend’). While the longer wavelengths of the digital dividend spectrum allow for more efficient propagation over long distances and around obstacles, the higher bandwidths of the 2.6GHz spectrum allow for faster, more concentrated data transmission to a number of simultaneous users, making it an ideal band for providing speed and capacity in major urban areas.

Spain and Italy have recently awarded their 4G spectrum licences via auctions (in July and September 2011 respectively). Both countries opted for selling licences in different frequency bands, including both the 2.6GHz and 800MHz bands, simultaneously, and used relatively standard “Simultaneous Multiple Round Ascending” (SMRA) auction formats.5 A brief description of the two award processes is provided in the box to the right. The same format was used in Germany in 2008.

However, even though the Italian and Spanish auctions took place at almost the same time, and both used a process that was very similar to the one used in the German auction, there were substantial differences in auction prices of 4G spectrum in these different awards: Italy achieved the highest prices (on a per-MHz-per-capita basis) for 800MHz spectrum in Europe so far, raising 10% more than the German auction did for the 800MHz licences. In Spain 800MHz licences sold for significantly less than in both Italy and Germany, with final auction prices being more than 40% lower than in Italy for spectrum in the digital dividend, and around half of the Italian prices in the 2.6GHz band.6

Why were prices so different?

The large differences in prices between the Spanish and the Italian award are likely to be the result of a number of factors.

First, differences in on-going annual spectrum fees need to be taken into account when comparing prices; the winning bids in the auction are only part of the overall price paid by winning bidders for spectrum, and higher on-going fees can be expected to be reflected in lower auction prices. This means that a simple comparison of auction prices may not show the full picture.

Second, differences in licence conditions and potential coverage obligations will affect bidders’ willingness to pay, as they will take into account the cost of such obligations when assessing the value of a licence. The relative burden of coverage obligations is very much dependent on factors such as the specific speed and service requirements, speed of roll out requirements and demographic and topological factors that influence build costs. It would seem that the Spanish licences have a much more ambitious coverage obligation, with a technologically neutral joint obligation on all bidders winning at least 2x10MHz of 800MHz spectrum to provide access speeds of at least 30 Mbps to 98% of the national population and 90% of the population in small towns with less than 5000 inhabitants by 2020 relative to Italy where 800MHz licensees have a seemingly less onerous service and coverage requirement to provide download speeds of 2Mbps to 90% of the population in municipalities with less than 3000 inhabitants, having to cover 75% of a pre-defined list of municipalities associated with the specific 800MHz frequency block won within a span of 5 years.

Third, reserve prices may also have an important role to play, in particular in cases where competition in the auction is limited. The Italian auction had substantially higher reserve prices, which already exceeded the final prices achieved in Spain.

Most obviously, the fact that only three of the four existing national mobile operators took part in the Spanish auction, in conjunction with tight spectrum caps that ultimately meant that there was excess supply of spectrum below 1GHz, is one of the main factors explaining the difference in the price achieved for the digital dividend spectrum. Relaxing these caps might have produced more competition in the auction and higher prices (and also resulted in less spectrum remaining unsold). However, loser spectrum caps also run the risk of more asymmetric outcomes, which could endanger competition in the downstream market.

The 2.6GHz conundrum

Perhaps more intriguing is the question why competition from the regional operators for spectrum in the 2.6GHz band has not made up for the fact that one of the four national players did not take part in the Spanish auction. Or, more specifically, why the price of national licences in the 2.6GHz band was not driven up by the competition that existed between the regional players and the national operator who decided to switch from bidding for one of the national 2x5MHz licences to assembling a national footprint by buying the corresponding amount of spectrum in each of the regions. Indeed, the total price for 2x5MHz of 2.6GHz spectrum in all regions – which should be equivalent to a national 2x5MHz licence, was higher (at €29.3 million) than the price of 2x10MHz lots offered on a national basis, the most expensive of which sold for €22.9m, or the price of the 2x10MHz lots aggregated across all regions at €24.3 million (including the reserve price for the unsold lot in Extremadura). The price per MHz per person of the smaller 2x5MHz regional lots thus averaged at 6.3 € cents exceeding the average 2.6GHz price achieved in Italy.

The answer to this question lies in the fact that switching between a national licence and the corresponding amount of spectrum acquired on a regional basis is not easy under the rules of the game. Equally, switching from a 2x5MHz lot to a 2x10MHz lot is not possible for an operator who is bidding up to its spectrum cap. Only if such switching were simple and straightforward would a combination of regional licences be a good substitute for a national licence, and would competition take place between those interested in a national footprint and those wishing to acquire frequencies only in a few regions.

And out to the regions …

With only 2x55MHz in the 2.6GHz band being available on a national basis, and three national operators being allowed to acquire 2x20MHz in this band each, the only way in which all of these big players could end up winning the full amount allowed for by the cap was for one of the bidders to try and assemble a national footprint by acquiring licences in all regions.

Vodafone was the first bidder to take the plunge, moving from bidding on a 2x5MHz national licence to bidding on the corresponding regional licences. Even though competition from regional operators turned out to be strong – perhaps unexpectedly so – the number of options for Vodafone were limited:

Because Vodafone was standing high bidder on 2x15MHz of spectrum on a national basis, it could not switch from bidding for 2x5MHz lots to 2x10MHz lots in the regions, as this would have meant a breach of the spectrum cap. In order to bid for 2x10MHz regional lots, Vodafone would have had to withdraw one of its standing high bids on a national 2x5MHz lot. This was permissible under the rules, but it would have exposed Vodafone to the risk of having to pay a potentially hefty withdrawal penalty.
Because Vodafone was standing high bidder on some of the regional 2x5MHz lots, it could not switch back to bidding for a national 2x5MHz licence without breaching its spectrum cap. Again, withdrawing from the standing high bids on the regional 2x5MHz lots would have been an option, but at the risk of having to pay withdrawal penalties.
By contrast, regional bidders such as ONO, Jazz Telecom and smaller regional operators were able to move between the regional 2x10MHz and 2x5MHz lots in response to changing prices. This meant that prices not only went up in regions with strong regional operators, but also in regions where competition might otherwise not have been strong.

The best of both worlds

The reason for using auctions for the award of spectrum is of course not to raise money – whilst realising a fair value of the spectrum given to operators for the public is clearly important, there are other considerations that come into play. The decision to award regional licences in Spain was driven by the desire to promote regional competition, and in this respect the auction format was effective.

There is still a question whether it might have been possible to sustain stronger competition in the auction through slight tweaks in the rules. For example, offering all spectrum in chunks of 2x5MHz would have avoided a situation in which regional 2x10MHz blocks effectively sold for less than the 2x5MHz blocks acquired by Vodafone. However, such a strategy would have exposed the regional operators to aggregation risks – the risk of winning only one block of 2x5MHz even though effectively two blocks might have been required for a viable business – and thereby discouraged participation. Similarly, offering all the lots on a regional basis would have facilitated switching, ensuring that all lots would be subject to competition from regional operators. However, this would have exposed national operators to aggregation risks when bidding on regional licences to obtain national coverage, as such operators could have failed to obtain licences in some of the regions, and could have resulted in an insufficient amount of spectrum being allocated as national licences.

Relaxing spectrum caps on 2.6GHz spectrum might similarly have allowed for easier switching between regional and national lots (subject to any constraints arising from the activity rules that are necessary in open bidding processes to provide an incentive to bidders to reveal their demand rather than hide in the grass). However, this would again have created the risk of a more asymmetric outcome, and perhaps one in which regional operators could have been squeezed out.

A general lesson would seem to be that mixing regional and national lots can create problems in an SMRA format, where standing high bids limit the ability to switch between combinations of regional lots and national lots. Alternative auction formats – such as clock auctions – might perform better in this regard (though such formats have of course other known weaknesses). It is clear, however, that auction design matters, and that one might have to be prepared to trade off different policy objectives. With a number of multi-band and 800MHz auctions approaching throughout Europe, it will be interesting to observe how other award designs will perform.

1. Staged activity requirements allow bidders to begin the auction with only part of the bidding activity they ultimately intend to use, and is subsequently tightened as the auction progress. [↩]
2. Because of their existing spectrum holdings in the 900MHz band, the spectrum cap limited France Telecom and Vodafone to bidding for at most 2x10MHz of sub-1GHz spectrum, and Telefónica to 2x15MHz. This maximum demand from existing national operators of 2x35MHz contrasted with an availability of 2×39.8MHz. This resulted in the smaller lot of 2×4.8MHz in the 900MHz band went unsold. [↩]
3. Telecom Italia and Vodafone won 2x5MHz of paired 1800MHz spectrum along with 2x15MHz of paired 2.6GHz spectrum, while Wind won 2x20MHz of paired 2.6GHz spectrum. [↩]
4. 2x5MHz of paired 1800MHz spectrum along with 2x15MHz of paired 2.6GHz spectrum. [↩]
5. In very general terms, in an SMRA multiple lots are auctioned off simultaneously over a number of rounds, and the auction closes on all lots simultaneously when no further bids are received. Bidders place bids on individual lots, potentially subject to some constraints, which allows them to switch between lots in response to changes in prices. The open bidding provides bidders with some information about the valuation of others, which helps reducing so-called ‘common value uncertainty’ and avoid the winner’s curse. [↩]
6. Prices throughout this article only refer to the headline prices paid at the end of each auction. They do not include any annual licence fees set by the regulators as these licence fees are not always made publicly available or may be based on aspects such as the financial performance of the operators. Sufficiently high ongoing fees may have depressed the headline price paid during the auction as bidders are likely to consider the lifetime cost of spectrum when making decisions during an auction. [↩]