To Improve Wireless Networks, Auction the Airwaves
Airwaves are a limited resource. There are only so many frequencies over which information can travel, and once they’re packed full, any new transmissions will interfere with the old ones. That’s why supposedly “unlimited” wireless plans typically throttle users who watch too many videos.
The Federal Communications Commission regulates U.S. wireless spectrum the way zoning boards regulate land use. Different frequencies are earmarked for specific purposes and licensed to individual companies.
Television broadcasters have historically held a lot of that turf. But as more and more people watch TV and other media on phones, tablets and streaming hookups, it makes sense to shift some spectrum from broadcast TV to mobile wireless.
In theory, transferring those spectrum rights would be easy. Mobile applications are worth more than many broadcast channels, so in principle, market-wise telecommunications companies could just call up broadcasters and offer to buy the space they need.
But it’s not that simple. Broadcasters typically hold local licenses; big mobile carriers want national networks. And broadcasters have incentives to hold out for absurdly high prices when they know that their licenses will be part of a big spectrum-acquisition program.
And there’s a second problem: For the assembled spectrum to be useful for mobile, it has to be contiguous and at the same frequencies across the country. This means that broadcasters who don’t sell may have to be moved to different channels, 1 and there’s a risk that leftover broadcast stations could interfere with each other after being rearranged.
So instead of being a straightforward exchange, the task of repurposing spectrum from broadcast to mobile is so knotty as to seem almost impossible. But the key word is “almost.”
A team of economists and computer scientists unraveled the knot earlier this year, as three of them – my colleagues Kevin Leyton-Brown, 2 Paul Milgrom and Ilya Segal – recently reported in the journal Proceedings of the National Academy of Sciences.
The centerpiece was a two-sided bidding process called the “Incentive Auction” that ran for 13 months beginning in March 2016.
First, broadcasters participated in a “descending clock” auction: The FCC started off by quoting high prices for broadcast licenses, at which most broadcasters would be willing to sell. Then the regulators slowly lowered the offered prices. As the prices declined, broadcasters who weren’t willing to sell at the new prices would leave the auction.
Before lowering any broadcaster’s price, an algorithm would check whether that broadcaster could be reassigned to an interference-free channel if it were to exit. 3 If no appropriate reassignment could be found, that station’s price would be frozen. The other prices would continue ticking down until they, too, had to be frozen, eventually determining the total cost of assembling a targeted amount of spectrum from broadcasters.
Then the action shifted to mobile carriers, who bid against each other to buy the spectrum that broadcasters were willing to sell at their last clock prices. If this auction didn’t generate enough revenue to cover the broadcast spectrum buyback (along with other costs), then the target was reduced so that frozen prices could be thawed, and the broadcaster clock was restarted. 4
It took a number of rounds, but it worked. The auction wrapped up in January 2017, recovering 84 megahertz of spectrum – the equivalent of 14 TV channels nationwide – for more than $19 billion. Roughly $10 billion of that went to broadcasters who sold their licenses, leaving a healthy surplus for the U.S. Treasury.
That’s pretty big victory for all involved – not to mention for the public, which will get wireless service with fewer indoor dead spots and better coverage in rural areas. It’s also a beautiful illustration of how economics and computer science can be used to build well-functioning marketplaces when the invisible hand doesn’t suffice on its own.
When the right ideas come together, even complex markets can run like clockwork.
Channel reassignment, called “repacking,” is itself a serious undertaking. Akin to changing the zoning code, it required redefining the structure of broadcast spectrum property rights and was also costly to implement.
Disclosure: Leyton-Brown chairs a professional organization (ACM SIGecom) for which I serve on a committee.
From a computational perspective, the general problem of checking interference is quite difficult. Thus, a team of computer scientists focused on problems that arose in FCC simulations, and used artificial intelligence techniques to train interference-checking algorithms specially adapted to the Incentive Auction setting.
With a lower target – equivalently, more channels left to broadcasters – there was additional space for rearrangement without interference, so the prices that had previously been frozen could be unfrozen and lowered further.
To contact the editor responsible for this story:
Jonathan Landman at firstname.lastname@example.org