Skip to main content

AI-Enabled Market Design: Lessons Learned from Radio Spectrum Reallocation

Kevin Leyton-Brown

University of British Columbia

Abstract: Over 13 months in 2016—17 the US Federal Communications Commission conducted an “incentive auction” to repurpose radio spectrum from broadcast television to wireless internet. In the end, the auction grossed $19.8 billion USD, $10.05 billion USD of which was paid to 175 broadcasters for voluntarily relinquishing their licenses across 14 UHF channels. Stations that continued broadcasting were assigned potentially new channels to fit as densely as possible into the channels that remained. The government netted more than $7 billion USD (used to pay down the national debt) after covering costs (including retuning).

Compared to typical market design settings, the auction design was particularly unconstrained, with flexibility in the definitions of participants’ property rights, the goods to be traded, their quantities, and the outcomes the market should seek to achieve. Computational tractability was also a first-order concern. The talk will begin by surveying the adopted design, which drew on both artificial intelligence techniques and novel economic ideas to strike an appealing balance between these constraints. However, the design was also extremely complex, and inevitably, the time constraints of working towards a real auction prevented thorough consideration of every potential variation of this design.

The second part of the talk will ask how computationally intensive simulation techniques can be used to investigate the potential impact of counterfactual changes to an auction design, answering these questions in the context of the incentive auction. First, we investigate computational issues, considering the feasibility checker that was used to determine whether a bidder’s price offer had to “freeze”, asking whether it would have mattered if we had invested less effort in building a strong feasibility checker. Second, we consider economic elements of the auction design, attempting to understand the impact of the choice to repack the VHF band (which nevertheless remained at its previous size after the auction) as well as the UHF band (which did not); and the effect of bid scoring (did Myerson-style scoring increase revenue as intended, and likewise did interference-based scoring increase efficiency)?

Time permitting, the talk will conclude with some ill-advised, rampant speculation about how recent advances in AI will impact the field of market design.