Selling the Right to Use a Frequency Nobody Can See
Governments allocate radio spectrum by auction, converting a public resource into licences worth tens of billions. The auction design turned out to matter enormously, and getting it wrong has cost real money.
An Unusual Thing to Own
Radio spectrum is the range of electromagnetic frequencies usable for communication. It is not consumed by use, cannot be manufactured, and is worthless without exclusivity, because two transmitters on the same frequency in the same place interfere and neither works.
So the resource requires an allocation mechanism, and the question of which mechanism dominated telecommunications policy for decades.
Historically licences were assigned by comparative hearings, in which a regulator judged which applicant would serve the public interest best. The process was slow, litigious, and rewarded lobbying rather than valuation.
Lotteries were tried and were worse, since winners frequently had no intention of building anything and simply resold the licence, transferring the entire value to a random recipient.
Why Auctions Won
Auctions address both problems. They allocate to the party valuing the licence most, which is a reasonable proxy for who will use it most productively, and they capture the value for the public rather than for a lottery winner.
American spectrum auctions began in 1994 and have raised well over a hundred billion dollars cumulatively. Similar mechanisms are now used almost everywhere.
The design was developed by economists, and it is one of the clearest cases of auction theory being applied directly to policy with measurable results.
The Design Problem
Spectrum is not sold as one item. It is sold as many licences covering different frequency bands and different geographic areas, and bidders need specific combinations.
A carrier wanting national coverage needs licences across every region. A licence covering one city is worth far less to it than the same licence as part of a national set. This is the complementarity problem, and it produces the exposure risk: a bidder that wins some licences but not the rest may end up paying for a set that is worth less than the sum of its bids.
| Design Problem | Mechanism Response |
|---|---|
| Complementary licences | Simultaneous multiple round bidding |
| Price discovery across items | Repeated rounds with published prices |
| Bidders signalling to each other | Anonymous bidding, bid increment rules |
| Freeing occupied spectrum | Incentive auctions paying incumbents |
The simultaneous multiple round format addresses complementarity by keeping all licences open across repeated rounds, so bidders observe prices developing everywhere and can adjust their aggregation strategy before anything closes.
The auction is not a sale, it is a price discovery process for a set of items whose value to each bidder depends on which of the others they win. That interdependence is what makes the design hard and why early auctions elsewhere went badly.
The Ways It Has Gone Wrong
Several failure modes have appeared in practice, and each produced a design response.
Collusive signalling. In early auctions bidders used the trailing digits of bid amounts to encode market numbers, communicating which licences they wanted others to leave alone. Rules now restrict bid increments and use anonymous bidding to prevent it.
Winner curse and overpayment. European third generation licence auctions around 2000 raised extraordinary sums, and several winners subsequently wrote down the value substantially, having bid on demand forecasts that did not materialise. Raising the most revenue is not the same as the best outcome, since a carrier that overpays has less capital to build the network the licence was meant to enable.
Bidder credits gamed. Programmes offering discounts to small businesses were exploited through structures where a small entity fronted for a large carrier, which produced enforcement actions and rule changes.
The Incentive Auction
The most technically ambitious mechanism addressed a different problem: valuable spectrum already held by broadcasters using it inefficiently.
The incentive auction ran two auctions simultaneously. A reverse auction determined what broadcasters would accept to surrender their licences, and a forward auction determined what carriers would pay for the resulting spectrum. The regulator then repacked remaining broadcasters into a narrower band and cleared the rest.
The design is genuinely novel, since it required solving a repacking problem across thousands of stations while prices were still moving. It cleared spectrum from an incumbent use without expropriation, paying holders to leave voluntarily.
What Happens After the Auction
The licence is not the end of the process. Winners typically face build out obligations requiring coverage of a defined share of the population within set periods, on the reasoning that spectrum should be used rather than hoarded.
Licences are also generally transferable, subject to regulatory approval, which creates a secondary market and means the auction allocation can be corrected later by trade.
Financially, spectrum licences appear on carrier balance sheets as intangible assets, frequently among the largest items there, and are typically treated as indefinite lived and tested for impairment rather than amortised. That treatment means a payment of tens of billions produces no ongoing charge against earnings, which flatters reported profitability relative to the capital deployed.
The Bottom Line
Spectrum auctions replaced regulatory judgement and lotteries with a market mechanism, and they work because they allocate to whoever values the resource most while capturing the value publicly. The design is genuinely difficult because licences are complements rather than substitutes, and the failures have been instructive: signalling between bidders, overpayment against optimistic forecasts, and credit programmes gamed by exactly the parties they excluded. The incentive auction is the most sophisticated version, buying spectrum back from one industry and selling it to another in a single simultaneous process.