The UK’s 4G auction was completed in February and Ofcom published detailed information on the bids made in the auction soon after. I thought it would be interesting to sift through this information in order to bring out the story of what happened in the auction and see if there were any indications of the mobile operators’ wider strategy.

So I put my deerstalker on and went through the data looking for clues. I was lucky with the auction design which let operators switch between a number of different types of spectrum ‘lot’. I’ll explain below how each time an operator jumped between different lot types, they left vital clues behind to unravel the secrets of the auction. However, this is still a work of deduction (though not detective fiction I hasten to add) and although the conclusions might not stand up in a court of law, I hope they make interesting reading.

A brief description of the auction design (skip if you’re already familiar)

Ofcom issued hundreds of pages of documentation about the auction, but I thought it would be useful to condense it into a few points. Ofcom auctioned two basic types of spectrum:

• low frequency spectrum in the 800MHz band, better for penetrating walls and providing good quality reception inside buildings (important because a lot of mobile broadband use takes place in the home or office) but a substantially lower amount of spectrum – 2x30MHz – on offer.
• higher frequency spectrum in the 800MHz band, less good at providing good quality reception inside buildings, but the larger amounts on offer – 2x70MHz  and 1x50MHz – means faster download speeds can be provided than with 800MHz.

The 800MHz band was split into two categories. One category had a coverage obligation – to cover 98% of the population by the end of 2017. The other category was free of any coverage obligation. The 2.6GHz band was split into 4 categories including paired spectrum and unpaired spectrum.

In the auction, participants bid for combinations of spectrum “lots” in the different categories and specified how much in each they wanted. Most of the bidders had caps on how much they could buy in total and in specific bands. This was done to ensure that market would be competitive after the auction.

Bidders could switch between lots of different types, at a fixed rate that stayed the same during the auction.  For example, the one 800MHz lot with the coverage obligation had twice as much spectrum (2x10MHz) as the four 800MHz lots without the obligation (2x5MHz each) and bidders were able to switch from the former to the latter at a rate of 2:1.

At the beginning of the auction, bidders specified how much spectrum they would initially bid on at the reserve price, and this set their eligibility – how much spectrum they could bid on in the next round. Bidders could reduce the amount of spectrum they bid for as the auction progressed (thus reducing their eligibility as the auction progressed). However, bidders were not allowed to increase they amount of spectrum bid on, i.e. bid more than their eligibility.

There were a number of phases to the auction and I focus on the primary bid rounds and the supplementary bids round, which are the most important for determining the winners and giving clues as to their wider strategies.

In the primary rounds, prices increased round by round (and demand fell in response) until the total demand for the spectrum equalled the amount available – there were 52 primary rounds. The supplementary bids round is a single round that follows the primary bid rounds. It gives bidders greater flexibility to express how much they are willing to pay for spectrum, consistent with how they bid in the primary bid rounds.

The overall progression of prices and demand in the auction

Before looking into the detail, I’ve put together some charts to give an overview of how the bidding ran in the primary rounds. The first chart shows how the prices changed round by round for 800MHz and 2.6GHz and the second chart shows the total number of lots demanded in each category.

Evolution of lot prices in the primary rounds

Evolution of total demand for lots in selected classes in the primary rounds

The competition for 800MHz spectrum

In round 16, Everything Everywhere, the UK’s largest operator, becomes the first operator to reduce its demand for the more valuable and strategically significant 800MHz. It then stops bidding entirely on 800MHz spectrum in round 24.

At this stage in the auction EE risks missing out on 800MHz spectrum if the price were to keep rising substantially. However it may be a smart move if it brings the bidding on 800MHz to an end more quickly (and more significantly at a lower price). EE will be able to modify how much it bids in the supplementary bids round, though its room for manoeuvre will be limited. . Moreover, EE has a substantial amount of 1800MHz spectrum, which it is already using to offer 4G services, and it may see this as a good back up if 800MHz becomes too expensive.

H3G is the second major operator to drop out of bidding for 800MHz in round 30. H3G knows it is very likely to win at least one block of 800MHz spectrum, because of Ofcom’s competition rules which limit the amount of 800MHz spectrum that O2 and Vodafone could win to 2x10MHz and because it is better placed than EE which dropped out of the bidding for 800MHz earlier.

The final burst of activity in the two 800MHz categories determines which out of O2 and Vodafone is likely to get the spectrum with the coverage obligation.

Interestingly, before the auction started, O2 argued that the price per MHz in the two 800MHz categories should be the same, whereas Vodafone argued that there should be a discount on the price of the lot with the coverage obligation. If the value of 800MHz spectrum were similar for Vodafone and O2 we should expect O2 to be willing to pay more for the lot with the coverage obligation than Vodafone.

As I said before, bidders could switch between 800MHz lots with and without the coverage obligation at a rate of 2:1. Now, if the extra cost due to the coverage obligation were minimal (e.g. if an operator would have met the coverage targets with or without the obligation) the coverage obligation would be worth twice as much as the lot without – reflecting difference in spectrum between the two lots.

However, the ratio of the starting (reserve) prices is significantly lower at 1.11 because Ofcom was cautious about the cost of the coverage obligation when setting the starting point.

So, excess demand is much greater for the coverage obligation lot early on in the primary rounds because it is relatively cheap compared to the other 800MHz lot. This causes the relative price of the coverage obligation lot to rise and it reaches 2 in round 40. At this point, Vodafone switches to the lot without the obligation and supply equals demand in both 800MHz categories as a result.

The case of the 2.6GHz lots

The bidding on 2.6GHz spectrum is seemingly straightforward, but there’s a twist at the end which any writer of detective fiction would be proud of (OK perhaps I’m exaggerating a little here). Up to round 27, the available 2.6GHz spectrum is more than three times oversubscribed and there is little change in the bids of the major operators. Then, Vodafone cuts its bid for paired 2.6GHz spectrum in half to 2x20MHz, and marginally increases its bid for unpaired spectrum (to 45MHz).

H3G makes a similar move to Vodafone, in round 30, reducing its bid on paired 2.6GHz spectrum by more than half to 2x20MHz and increasing its bid for unpaired spectrum marginally. H3G also drops out of the bidding for 800MHz at this point, suggesting that prices could be nearing its underlying values or that H3G may be close to a budget limit –H3G had bid just under £1.5 billion, though it bid nearly £1.7 billion in the supplementary bids round.

In the next round, 31, it’s O2’s turn to reduce significantly the amount of 2.6GHz spectrum it bid for (both paired and unpaired). This still leaves substantial excess demand for the 2x70MHz of paired 2.6GHz spectrum available – H3G, Niche (BT), O2 and Vodafone are each bidding for 2x20MHz and EE for 2x40MHz.

Similarly there’s also substantial excess demand for the 45MHz of unpaired 2.6GHz spectrum – H3G, Hong Kong Telekom, Niche (BT) and Vodafone each bidding for 45MHz and O2 for 15MHz.

Things move steadily on until EE makes a dramatic grab for the unpaired 2.6GHz spectrum (and stops bidding on the paired spectrum) in round 38. EE is the only one left bidding for the unpaired 2.6GHz spectrum at the end of the primary bid rounds and Vodafone, O2 and Niche are the only bidders remaining for the paired 2.6GHz spectrum.

But, just one more thing, as Columbo might say. I’ve forgotten the supplementary bids stage. The final twist is that the positions at the end of the primary bid rounds are overturned in the supplementary bids round. So the final result is that Niche and Vodafone win the unpaired 2.6GHz spectrum instead of EE, while Vodafone Niche and EE win the paired 2.6GHz spectrum.

Conclusions

Ofcom should be satisfied with how the auction ran. Bidders did respond as economic theory predicts to changes in the relative prices of the different lots in the auction and it showed the importance of having a supplementary round to extract more information about what bidders were willing to pay. There is no clear evidence to suggest that bidders were trying to ‘game’ the auction, i.e. put in spurious bids to trick their competitors (although there are some bids that are more difficult to explain in the two minor categories I haven’t talked about).

The competition proposals did probably affect behaviour in the bidding for 800MHz, although there was still a reasonable amount of bidding activity and the overall amount of money raised, when corrected for population, was similar to other European 800MHz auctions.