Article - Issue 6, November 2000

Spectrum auctions

Professor Martin Cave

Download the article (411 KB)

Recent events have vividly illustrated the importance to the economy of the radio spectrum. The British Government, auctioning five licences for third generation mobile, was astonished to find them yielding a revenue of £22.5 bn. Britain was following in the wake of other countries, notably the United States – although a five-year series of auctions there had only generated £14 bn. Finance ministers in Europe began to lick their lips. Other commentators, such as Nicholas Negroponte of the MIT Media Lab, expressed concern that the high licence charges would feed into consumer prices and, by blighting the spread of wireless Internet applications, impoverish future generations.

The idea of allocating spectrum through a market process surfaced over 50 years ago in relation to broadcasting. At that time it was dismissed out of hand, largely on the basis that the airwaves were a public resource which governments should allocate in the public interest. This is understandable, but inconsistent with the normal operating principles of a market economy where inputs in the productive process are normally distributed by a price mechanism. Auctions are just another means of employing the price mechanism, normally used in the allocation of a fixed number of objects – fish on a dock, old masters, oil extraction licences and, more recently, spectrum lots.

The point about a market system is that it creates an environment in which scarce resources end up in the ownership of the people who value them most highly. Households and firms demonstrate this every day through their purchases and sales of goods and services and of labour. Alternative resource allocation mechanisms, in which some central agent decides who gets what, have generally failed.

In one sense, the focus of most of the recent discussion, on licence assignment, misses the point. This is because licences are assigned to particular operators within the context of a prior allocation of spectrum across major types of usage. If this spectrum allocation is made suboptimally (as is almost certainly the case), then spectrum assigned for some purposes, such as mobile telephony, is needlessly scarce. One of the main culprits here is the military. In Italy, two thirds of available frequency bands are occupied by the Ministry of Defence at zero cost. In the UK the situation is broadly the same but an administrative charge is levied for the spectrum, the defence budget being simultaneously increased by the equivalent amount. In principle, the Ministry of Defence retains the budget increase even if it releases spectrum, thus saving the administrative charge.

Auctions vs Beauty Contests

When spectrum is allocated through an administrative process, the government typically conducts what is called a beauty contest: it designates the spectrum lots which are to be made available, indicates the criteria under which it will make the assignment, and invites proposals from interested parties. Typically, the successful applicants pay an administrative price designed to meet engineering and administrative costs (though, as we shall see, the price can be set much higher). The licences are then assigned.

There are two major problems. The first concerns the integrity of the process: because beauty is in the eye of the beholder, many possible decisions can be justified. This creates opportunities for favouritism or corruption – which often takes the form in the EU Single Market of a preference for national firms. Secondly, there is a problem of information asymmetry. A government sincerely implementing a beauty contest has to discriminate among proposals submitted by firms, each of which will want to demonstrate its suitability through optimistic projections of revenue, take-up, etc. How can a group of non-commercial public servants, even guided by consultants, make these judgements?

The crucial difference with an auction is that the winning bidder has to back its projections with hard cash. Thus, bidders have every incentive to make realistic estimates. Subject to problems described below, this will have the effect of assigning spectrum to the firm which can use it most effectively. It is this property – the efficient assignment of spectrum – rather than maximising revenue, which is the real pay-off to a properly designed spectrum auction.

Indeed, maximising revenue for its own sake is antithetical to the efficiency which competition can bring. Consider, for example, a situation in which three mobile licences were up for auction. The three licences probably have the highest value in the hands of a single operator, which would then have the monopoly of the service. This would obviously be an undesirable outcome. It can, however, be simply remedied through appropriate auction design – in this case, a rule that an operator can only control one licence.

The auction rules could similarly be set to achieve other objectives. For example, if there is a desire to bring new entrants into the market, some licences can be reserved for them, or they might receive special benefits in the auction process (for example, by adding a notional monetary sum to their bids).

Other policy objectives can be built into an auction process by imposing appropriate licence conditions. Suppose it is an objective of policy to ensure a fast network roll-out, perhaps faster than a licensee would choose to go in its private interests. Then a licence could be allocated subject to roll-out conditions. This would presumably reduce firms’ willingness to pay, but the government must have believed this to be worthwhile, as otherwise it should not have imposed the condition.

Thus auctions are an unusually flexible means of allocating resources in ways designed to achieve a range of policy objectives. They also have the great benefit of transparency. If the auction rules are explicit, potential bidders know in advance the basis upon which they are competing. This is not only efficient, because it encourages bidders to participate, but also equitable, because they are treated equally.

Designing spectrum auctions

Much effort over the past five years has gone into designing efficient ways to allocate spectrum. One of the recognised problems with tender or auctioning procedures is the ‘winner’s curse’ – the risk that an auction system will end up putting the spectrum in the hands of the firm which most overestimated its value. In reaction to this, sophisticated bidders will tend to ‘aim off’ in rather unpredictable ways, creating the risk of misassignment.

In auctioning something like spectrum, which firms are valuing in a commercial rather than a sentimental way, using broadly the same techniques, the best way to assuage anxieties about the ‘winner’s curse’ is to put as much information as possible into the market place. This can be done by holding an ascending auction (known as an English auction), rather than a sealed bid process. Firms see what other firms are bidding and this gives them confidence. The operation of this process can be seen in the UK UMTS auction, where final revenues were five or six times expectations, as operators learnt from each other. (The alternative explanation, that the operators had a fit of collective hysteria, is not really plausible.)

However, open auctions are subject to collusion, which may both reduce revenues and lead to inefficient assignments. In the UK auction, thirteen companies were bidding for five licences, one reserved for an entrant (see Figure 1). In the subsequent Dutch UMTS auction, after withdrawals, six companies (five of them incumbents) were bidding for five licences. Revenues were, on a proportionate basis, one fifth those in the UK. The US PCS auction of 1998 threw up some interesting examples of attempted collusion. Many lots of regional spectrum were being allocated simultaneously. An East Coast operator (E) might resent a West Coast operator’s (W) bid on lot 453 in its territory. It could signal this by bidding for a lot coveted by W an amount of $120,000,453. ‘You stay out of Boston or I’ll go into San Francisco’. The US regulator later dealt with this by requiring bids to be made in round numbers.

In other European auctions, operators have dealt with the competition ‘problem’ by forming consortia beforehand. If governments cannot cope with this using their competition laws, it might make sense to apply a very high reserve price, in the knowledge that not all the licences may go in the first round. A rather less satisfactory solution is the French UMTS proposal, which involves a beauty contest at an administrative price calculated to yield about half UK revenues.

Does the auction of spectrum raise the price of services?

As well as attracting complaints from the industry, the British UMTS auction has been attacked as an ‘economically unsustainable’ tax on internet technology, on the assumption that the high licences will raise prices to consumers, delay the roll-out of infrastructure and discourage operators’ ability to meet social obligations.

The contrary argument is that spectrum auctions merely transfer profits from firms to the government without influencing the prices consumers pay. To follow this argument, consider how the firm decides what to bid for a licence. It will know from the rules of the auction process how many competitors will be licensed and hence how many firms will be competing to provide service. By forming a conjecture about how the competitive process will play out, it can estimate what revenues over and above capital and operating costs it will earn. On this basis, it can calculate the maximum a licence would be worth to it and judge its bidding accordingly.

Suppose that now it has been successful. It has paid for the licence, either up-front in full or through a binding commitment to pay in instalments. As far as the firm and its competitors are concerned, the licence fee is an irrevocable or sunk cost. If the spectrum is transferred to another owner, it must be used for the same purpose and becomes a sunk cost for that owner. When deciding how to set prices, the firm rationally only takes account of its own forward looking costs and revenues and the likely behaviour of other firms. Since the licence fee is a sunk cost for all firms, it falls out of the pricing equation for all of them. Hence the size of the licence fee does not affect prices.

Various attempts have been made to refute this argument, in ascending order of credibility. The first is that even though firms should logically maximise their profits on the basis of the comparison of forward looking costs and forward looking revenues, managers in all the licensed firms will act ‘as if’ the licence fee were a forward looking cost, and raise prices accordingly. The obvious flaw in this argument is that if all the licensees except one behaved in this way, the last licensee could increase its profits by undercutting the others and stealing their business.

Secondly, it is suggested that operators burdened by substantial sunk licence costs will feel themselves more vulnerable to bankruptcy, and hence more reluctant to engage in price wars and more inclined to act collusively. But this is open to the same objection as above.

Finally, it is suggested that large licence payments erode the finances of operators, requiring them to assume debts thus reducing their credit rating. The effect is to increase the costs of borrowing for all purposes, and hence to raise forward looking costs.

This is a genuine risk. High licence payments do not in any way influence the systematic risk associated with a mobile phone company (the degree to which its returns are linked to returns to the stock market as a whole). There is, however, potentially an increase in bankruptcy risk, which may have the effect of increasing the cost of capital. This has been shown by the decision of rating agencies, such as Standard & Poor, to review the credit rating of major companies such as BT and Deutsche Telekom, which have invested heavily in 3G licences. The likely effect is a small increase in the cost of debt.

Short-term financing problems can be relieved by staging spectrum or licence payments. But this should not go too far. If it does, the down-payment is, in effect, merely the purchase of an option, and the body issuing spectrum becomes effectively an issuer of financial instruments. The Federal Communications Commission in the United States got itself into this position with one of its licence auctions.

The roll-out of UMTS assignments

Finally, Table 1 shows the progress in the assignment of UMTS licences within Europe.

The table shows a mixed picture. Finland and Spain have already made their awards through beauty contests. In Spain, France Telecom has challenged the outcome in the Courts; this is a natural response by unsuccessful applicants to the discretionary and subjective nature of beauty contests. The success of the UK auction has elicited responses from other governments and bidders. Bidders have responded by quitting the contests or by forming alliances. Thus the recent German auction attracted only half the number of participants of the UK auction, even though revenue on a per capita basis was the same. There may, in some cases, be the same number of bidders as of licences. The resulting allocation of licences to what may be inefficient operators and reduction in fees paid will deprive consumers of opportunities and governments of revenue without bringing prices down. The lesson to be learnt is better auction design and rigorous implementation of competition policy.

Lessons for the future

The fact that relatively small quantities of spectrum, assigned to designated purposes, are so valuable, is an indictment of the extreme inefficiency of the current allocation of spectrum across alternative uses. The absence of monetary valuations of spectrum in the past has concealed a waste of resource on an alarming scale. The correct lesson to draw from the higher valuations is that more spectrum should be reallocated from less valuable to more valuable uses. Inevitably, the necessary process of clearing and reallocation takes time, especially if it is done within a framework of international agreements presided over by the International Telecommunications Union, a body hampered by international differences in priorities and not known for speed of manoeuvre.

However, progress is being made. One of the key drivers in favour of the switch off of analogue television broadcasting is that it will free spectrum which can be used for a variety of purposes, including mobile communications. By allowing more operators into that market, prices would fall and service levels improve. Ideally, spectrum assignments and allocations should sort themselves out so that the marginal value of spectrum is the same – equally low in all uses. This can be furthered most directly by allowing the option of continuous reallocation through the greatest possible degree of spectrum trading. This activity is – regrettably – currently outlawed in the European Union, but is likely soon to be legalised in part of general communications reforms.

Martin Cave is Professor of Economics and Vice-Principal at Brunel University. He specialises in regulation and competition policy with a focus on the communications industry. He is an adviser to OFTEL, a member of the DTI’s Spectrum Management Advisory Group and a member of the Competition Commission.

[Top of the page]