What are the US Spectrum Auctions for 5G Teaching Us About Auction Efficiency?

Abstract

This paper seeks to assess the efficiency of the first 5G spectrum auctions held in the United States. In 2018 and 2019, the Federal Communications Commission (FCC) administered auctions 101 and 102 to assign spectrum in the 28 GHz and 24 GHz bands, respectively. Auction 101 sold spectrum on the 28 GHz band for Upper Microwave Flexible Use Service licenses and raised a total of about $700 million. Auction 102 assigned spectrum on the 24 GHz band for Upper Microwave Flexible Use Services. It raised a total of close to $2 billion. In July 2020, the FCC auctioned the 3.5 GHz band for Priority Access Licensees, raising over $4 billion. I implement an assessment of the efficiency of an auction based on its observed bidding data during the clock rounds of the auctions, while relying on revealed preference theory. Data are available at the FCC’s Auctions website. Bidding data are organised by geographical area and for each geographical area, demand point sets for each round, for each bidder. I rely on the application of the well-known Generalised Axiom of Revealed Preference or GARP, and less well-known Homothetic Axiom of Revealed Preference or HARP, to construct two indexes, Afriat Efficiency Index (AEI) and Homothetic Efficiency Index (HEI) to calculate measurements of rational bidding associated with each bidder. The US relied on an auction format, which uses demand revelation but allows the bidders to state the price at which they would demand a number of units of the same item (that is, blocks or licenses). In essence, the auction format for auctions 102 and 105 followed a multiple round format with an ascending clock not entirely driven by the auctioneer, whereby a bid is best described as a stepwise (ascending or descending) demand function revealed on the price range set by the auctioneer for the round. The difficulty in dealing with the data is that the bidding language of these auctions allows a bidder to express its demand for the items being sold in a service area by means of a stepwise function (either ascending or descending). The challenge at this stage is to suitably adapt the data to the usual data set format assumed by GARP or HARP. Using an efficiency approach, I assess each auction at two levels: assessment of efficiency through utility maximisation via AEI and an assessment of efficiency through homotheticity via HEI. Auctioneers, in particular, and spectrum authorities, in general, are concerned with the economic aspects of their spectrum assignment processes. The most important issue that pertains the results of the auction is the answer to the question, was the spectrum assigned to those bidders who can make the best use of it? Answering the question puts us on the grounds of auction efficiency. Devising methods to use the auction bidding data to answer such question has been a long-standing question. GARP and HARP are two revealed preference approaches that do not require us to adopt a parametrized utility function to represent bidder’s preferences. Those methods use the observed data and define a rigorous consistency test on the data.