Mechanism Design: Engineering Outcomes

You wake up on a Tuesday morning in January 1993 with a problem that no one in the federal government has ever solved cleanly.

You are an advisor to the Federal Communications Commission, and Congress has just handed you an unprecedented mandate. The Omnibus Budget Reconciliation Act has authorized the FCC to auction off licenses for the electromagnetic spectrum for the first time in American history. Hundreds of licenses. Billions of dollars in potential value. The airwaves that will carry the next generation of cellular communications are sitting unclaimed, and it is your job to figure out how to hand them out.

The old system is a disaster you know intimately. For decades, the FCC awarded spectrum licenses through comparative hearings: applicants submitted lengthy proposals, bureaucrats evaluated them, and licenses went to whoever made the most compelling case on paper. The process took years, cost millions in legal fees, and produced outcomes that had almost nothing to do with who could use the spectrum most productively. By the late 1980s, the agency had switched to lotteries, hoping randomness would at least be faster. It was faster. It was also catastrophically wasteful. Lottery winners with no intention of ever building a radio network sold their licenses within weeks to the highest bidder, pocketing millions for doing nothing. One winner of a cellular license in 1989 had no telecommunications experience whatsoever and flipped the license for a profit exceeding $40 million.

Now auctions are on the table. But here is where your Tuesday morning gets complicated.

Two camps have formed inside the building. The first camp argues for simplicity: run a simultaneous sealed-bid auction, highest bidder wins each license, collect the money, go home. Clean, fast, legally defensible. The second camp, led by a group of academic economists including Paul Milgrom and Robert Wilson from Stanford, is proposing something that has never been tried at this scale. They call it a simultaneous multiple-round ascending auction: dozens of licenses open for bidding at the same time, with multiple rounds of bids, prices rising gradually, and bidders allowed to switch between licenses as information is revealed.

The economists insist their design solves a specific problem: spectrum licenses are not independent objects. A cellular company that wins licenses in Chicago and Detroit but loses the license in Indianapolis ends up with a network full of gaps. The value of any single license depends on which other licenses the same bidder wins. The simple auction ignores this completely.

But the complex auction has never been run. Not once. Not anywhere. If it fails, it fails publicly, expensively, and in front of Congress.

You have to recommend one approach by end of week.