#127 1994 · US Federal Communications Commission (design: Paul Milgrom, Robert Wilson, Preston McAfee) · Telecommunications regulation / public asset allocation
Selling one radio license at a time made bidders guess blind, so economists made every license sell at once instead
the problem
A government needed to sell interdependent assets whose value to any bidder depended on which other assets that same bidder also won
background
Before 1994, the US government gave away broadcast and wireless spectrum for free, first through opaque, lawsuit-prone 'comparative hearings' judging applicants on vague public-interest criteria, and after 1983 through a pure lottery — a system so disconnected from actual value that winners included speculators and, notoriously, actor Ernest Borgnine as part of a spectrum-holding syndicate, while billions of dollars in public asset value went to whoever's name got drawn.
When Congress authorized the FCC to auction spectrum instead, a plain sequential or sealed-bid auction looked like the obvious fix — until economists pointed out the real problem: spectrum licenses across different regions were worth different amounts to a bidder depending on which OTHER regional licenses that same bidder also won, since a national carrier needed contiguous coverage, not just any single license cheap. Bidding license by license in sequence meant a company had to guess at the value of licenses it hadn't won yet while locking in commitments on the ones already sold — a structural trap known as the exposure problem.
what everyone would do
Auction the licenses one at a time or through sealed bids, the standard way governments had sold everything from timber rights to bonds. It fails here because spectrum licenses are interdependent: a license's value to a national carrier depends on which other regional licenses that same bidder also wins, so selling them in sequence forces a bidder to commit to early licenses before knowing the price or outcome of the later ones they actually need to complete a usable network.
what they saw
Milgrom, Wilson and McAfee saw that the problem wasn't how to price any single license, it was that bidders couldn't express what they actually valued — a specific combination of licenses — without seeing the emerging prices of all of them together. The fix wasn't a cleverer single-item auction format, it was letting every related license be bid on at once, so the auction itself could reveal the package prices bidders needed before anyone had to commit.
the move
Economists Paul Milgrom and Robert Wilson, joined by Preston McAfee, designed the simultaneous multiple-round (SMR) auction: rather than auctioning licenses one at a time, the FCC ran a large set of related licenses through repeated bidding rounds all at once, with the current high bid on every license visible after each round. Bidders could watch prices develop across an entire region simultaneously and shift their bids toward the actual combination of licenses they needed as the true relative value became visible round by round, instead of committing blind to one license before knowing the price of the others.
why it works
Running every related license through the same round simultaneously lets bidders watch the current high bid on every license after each round, so a company can see how the price of a license it wants is developing at the same time as the licenses it would need alongside it. Because bidders can shift their bids toward the actual combination that makes sense as true relative values become visible, nobody has to lock in a purchase before knowing whether the complementary licenses they need will still be affordable — which is exactly the exposure problem a sequential sale creates. Prices converge toward each bidder's real, package-level valuation instead of a defensive guess, which is why the first auction alone raised $617 million from spectrum previously given away for free.
the payoff
The first SMR auction in July 1994 sold 10 licenses over 47 rounds for $617 million — a windfall from spectrum the government had given away for free since 1927 — and the format went on to structure 87 subsequent FCC spectrum auctions raising more than $60 billion for the federal government.
where it breaks
The design depends on there being genuine interdependence between the items being sold — for assets with no meaningful complementarity, a simpler sequential or sealed-bid auction works fine and simultaneous rounds add unnecessary complexity and time. It also requires bidders sophisticated enough to process publicly visible round-by-round pricing across many items at once, and a market thick enough that multiple serious bidders are actually competing for overlapping combinations; with too few bidders or too little genuine substitutability between packages, revealing prices in real time mostly just helps the few participants coordinate rather than compete.
what came after
The simultaneous multiple-round auction became the de facto global standard for spectrum sales and was later adapted for allocating gas stations, airport landing slots, telephone numbers, fishing quotas and emissions permits; Milgrom and Wilson were awarded the 2020 Nobel Memorial Prize in Economic Sciences explicitly for auction theory and this real-world design, and the work won the Golden Goose Award for federally-funded research with outsized practical payoff.
references
- [1]The Spectrum Auction: How Economists Saved the DayPriceonomics, 2015priceonomics.com
- [2]2014: Auction DesignGolden Goose Award, 2014goldengooseaward.org