In all the hustle and bustle, it rather blew by that my friends Dr. Gregory Rose and Mark Lloyd have written this analysis of ten years of FCC spectrum auction data.
Summary — FCC auctions turn out to be great ways for incumbents to exclude new entrants and to bilk the government. They do not yield the promised efficiencies of distribution or even maximize revenue to the government. There are ways to improve the process, but the FCC open ascending auction systems just about ensures that a collection of incumbents can keep out any genuinely disruptive competitors and collude to minimize revenue to the government and maintain the status quo.
The Report is dense, but important. The literature supporting spectrum auctions before the U.S. adopted them in 1993 (and made them mandatory ofr most allocations in 1997) rested on theoretical models and analysis of relatively simple and straight forward open ascending auctions like wine auctions or estate auctions. While individual auctions have been subject to academic study (for example, the “cursed” “C Block” auction of 1994), no one until now has undertaken to do a thorough survey of the last ten years of FCC spectrum auctions to see if the system actually works as advertised.
In theory, auctions distribute licenses in an efficient and competitive manner becuase the parties bidding price the value of the license based on expected return. Auctions also supposedly have the salutory effect of creating a new (one time) revenue source for the government. This last has led me to say repeatedly that “spectrum auctions are the crack cocaine of public policy.” Once governments get a hit of revenue, they are willing to throw all other considerations out the window and sell off pieces of critical infrastructure.
Auction supporters like to point to the big revenue auctions as proof that the auction system works. But the Center for American Progress (CAP) report shows that these big auctions are the exception, rather than the rule, and that over time the constant use of auctions to distribute spectrum rights produces a classic game dynamic where a relatively few players develop well understood rules for communication and collusion to distribute licenses between themselves for far less than the anticipated “market price.” The fact that auctions take place over multiple rounds and are “open” (i.e., everyone can see what everyone else does) allows players to punish mavericks or new players by bidding up licenses these participants need or simply outbidding them to exclude them from the market.
As discussed by London School of Economics professor and auction guru Paul Klemperer, this experience to some degree tracks what happened in the European “3G” spectrum auctions (see the paper “How (Not) to Run An Auction”) on Klepmerer’s site. UK went first, and could get away with an open auction because it takes at least one auction for participants to establish the rules for collusion within the auction structure. Subsequen open auctions for spectrum licenses in Europe yielded relatively little return, however. Klemperer freely acknowledges that some of this is attributable to the change in the financial environrment between 2000 (when the UK auction took place) and 2001 (when the other European auctions took place), he makes a good case that allowing open auctions encourages serious game playing by incumbents at the expense of new entrants, consumers, and government revenue projections.
The Rose & Lloyd Paper certainly supports the basic findings suggested by Klemperer in the European experience. But the Rose & Lloyd Paper goes further in asking a fundamental question — can we structure actions to eliminate this game playing or is it intrinsic to using auctions to distribute licenses? Klemperer clearly thinks that auction design can defeat, or at least minimize game playing and, in any event, it is better than using comparative hearings or other method of distributing licenses.
But I will go further than either Klemperer or Rose & Lloyd (being, after all, unfettered by the need to limit myself to the empircal and launch myself into the theoretical). Do we need auctions at all? Klemperer may be right that auctions are better than comparatve hearings (aka “beauty contests”) — an opinion disputed on occassion by such luminaries as Nicholas Negroponte — but that does not leave us with auctions as the only means of distributing spectrum access.
Regular readers will recognize where I’m going here, of course. Why not forget auctions for awhile and focus on promoting use and development of wireless services via unlicensed (or at least non-exclusive) access? Proponents of auctions and treating spectrum as property (such as the redoubtable Thomas Hazlett generally respond that property rights and auctions create the most efficient distrbution and efficient use of spectrum, and that while “spectrum commons” may provide some utility in certain cases (since after five years, they have finally given up arguing that it has no utility) they will inevitably suffer from a “tragedy of the commons.”
But the empirical evdence doesn’t stack up. Spectrum auctions have not produced the efficiencies imagined by these same theorists in the 1980s and 1990s. When confronted by this empircal evidence, they usually respond with the sort of religious dogmatism one sees when you ask Imams why rejecting western science and culture in favor of Wahabi fundamentalism has failed to produce the promised paradise for the faithful; i.e., that servants of the great Satan (here, the FCC) keep opposing them and that the faithful haven’t been fanatical enough in following the propertizing prescriptions of the Holy Writ.
I’m not ready yet to proclaim the ascendancy of the Supercommons and pronounce the real ultimate “true way of spectrum management,” I think the record certainly justifies giving the spectrum commons approach its turn at bat and puting spectrum auctions on hold for a little while. So here’s my suggestion — why not stop having spectrum auctions for the next ten years? (Although I suppose we are stuck with the AWS auction scheduled for August 9.) After we’ve collected ten years of empirical data on distributing spectrum non-exclusively to compare with the ten years of auctions and the preceeding 60 years of “beauty contests” and lotteries, we can start to make some intelligent decisions based on facts rather than just theoretical models.
Just a thought, but I have a curious penchant for empircal data and pragmatism. I recognize this make me a mutant in policy land among the Policy Wonk Barbies (“Reality is hard! Let’s do policy!”), but I gotta be me.
Stay tuned . . . .
