“Competitive Bidding for new gTLDs” is the focus of part three of a three-part series based on a study prepared by Karl M. Manheim, Professor of Law at Loyola Law School and Lawrence B. Solum, Professor of Law at University of San Diego. Special thanks and credit to Hastings Communications and Entertainment Law Journal, Vol. 25, p. 317, 2004.

When new radio frequencies become available for commercial use, federal law requires that licenses be auctioned off to the highest qualified bidder. The FCC does a reasonably good job in designing and conducting spectrum auctions. They are often familiar in format, not much different than found for consumer goods on eBay. In other cases, such as with “Simultaneous Multiple-Round” or “combinatorial bidding,” the auction design is fairly complex. Because of complexity in these cases, the FCC sponsors periodic conferences on auction theory and seminars on auction mechanics for potential bidders.

Compared to spectrum auctions, we believe gTLD auctions will be relatively simple, both in concept and operation. Nonetheless, we think actual auction design should be worked out by ICANN to assure compatibility with technical standards and to maximize economic efficiency. ICANN can draw on the FCC’s operational experience with auctions as well as a large body of literature on auction theory and design. Although we describe an auction model below, we do not propose that ICANN accept it based on our analysis alone. Rather, our purpose here is to lay out a framework, describing what should be auctioned and roughly how. The actual process of auction design should include input from economists who specialize in auction theory; indeed, ICANN should either hire a staff economist or develop a consulting relationship with an economist specializing in auction design.

What Should Be Auctioned?

Unlike the allocation of radio licenses or telephone numbers, where the relevant space is identified prior to issuance, the domain name space is constrained only by the allowable character set and string length. Thus, allocation of gTLDs involves selection both of domain names and sponsors or operators. The former is analogous to the radio frequency band or telephone number set; the latter are the licensees. FCC spectrum auctions, until now at least, have determined only licensees. Determination of the frequencies to be awarded (the “auction inventory”), or their permitted uses, is subject to pre-auction administrative processes. Adapting this practice to gTLDs means that, prior to auction, ICANN would first determine the domain name or names to be added. An auction would then select the operators of the newly approved gTLDs. For example, ICANN might decide to add .sex to the name space, and then hold an auction solely to determine the registry operator. While this mechanism might facilitate gTLD expansion, and would certainly produce revenue, it does not address the principal nature of inefficiency now encumbering the gTLD selection process.

It is unrealistic to expect ICANN to rationally determine which gTLDs should be added to the root. There are few if any objective selection criteria. Does a gTLD need to be pronounceable or have semantic meaning? Does it need to be descriptive? These questions, while sensible in the abstract, are not germane to the question at hand any more than whether nouns and adjectives are eligible to become gTLDs. In one notorious case, the ICANN Board selected .aero over .air because it felt the latter was a public resource.

The principle at play during the first expansion round in 2000 was “proof-of-concept” domains. As useful as that might sound for evaluating many aspects of expansion, including root operation, allocation methods, and even individual applicants, it is not a meaningful tool in distinguishing among possible gTLDs. In fact, there may be no rational policy choices. Regulatory decisions on which gTLDs to add are inevitably arbitrary, or simply favor particular interest groups. The highly engineered grid of gTLD assignments that marks the current domain name space does not necessarily measure or meet the needs of the Internet community.

Why not let the market decide? If .air has greater utility to the Internet community than .aero, why shouldn’t it be added? Why not add both? To be sure, some users may enter .air when they are looking for .aero. And some domain holders might be induced to protect their trademarks or investments by buying domains in multiple gTLDs (so-called “defensive registrations”). But, those inefficiencies occur now, and are presumably reflected in the value of the gTLDs themselves. In short, the auction process should be structured so the question of which gTLDs to add is itself determined by the market. This is likely to yield a higher and better use of the name space than would be achievable by any bureaucracy. The experience of the FCC with “flexible spectrum” allocation strongly supports economic theory on this point.

Closely related to the question of which gTLDs should be added is the issue of how many should be authorized at any time. Here ICANN must play an important role. Integrity of the root and domain name system is its raison d’?tre. Unlimited or too-rapid expansion could overwhelm both the infrastructure and the markets that have developed based on existing practices. It could also undermine the goals to be achieved by competitive bidding in the first place. Instead, ICANN should make an informed judgment, based on technical and public policy factors, on how fast to expand the domain name space. But their decisions must be transparent, so as to avoid any suspicion that stakeholders are manipulating the auction process so as to maintain scarcity or protect incumbents. In expanding radio bands or telephone numbers, the FCC is faced with real-world technical constraints. ICANN needs to articulate similarly objective criteria for any decision reached on how many gTLDs to add.

In this respect, “proof-of-concept” is a prudent policy. We think ICANN can legitimately limit the first round of gTLD auctions to a technically and administratively manageable number. If for no other reason, the model and economics need to be tested. There were 44 serious applicants for new gTLDs in December, 2000, each anteing up the $50,000 application fee. Several more applicants have emerged since then. We believe this number ? roughly 50 ? provides a suitable lodestar figure for ICANN to consider. Presumably, the root server system can support this number of new gTLDs. And if ICANN’s cumbersome selection process were replaced by an auction, no significant administrative burdens would be encountered.

What should be the duration of the right, purchased at auction, to operate a gTLD? Should it be like a spectrum license which has a prescribed term (e.g., 8 years for broadcast licenses), and renewable thereafter? Or should the right be perpetual? One advantage licensing has over outright sale is the oversight it forces at renewal time. ICANN’s agreement with Verisign for operation of the .com, .org, and .net registries was for eight years, with no mention of renewal. We think this is a suitable term, but that a renewal expectancy will enhance the price paid and capital investment of new gTLDs. The right should also be revocable upon material breach of an agreement with ICANN for the stable operation of the gTLD. So long as the rules are transparent and term security is worked out in advanced, the auction should function smoothly.

One objection to an auction for new gTLDs is that incumbent registry owners get a free ride. The question is whether the existing gTLDs should also be auctioned, either in the first round or when their agreements with ICANN expire. Otherwise, the argument goes, they would have an advantage over competitor gTLDs because their acquisition costs were much lower. The half-dozen existing commercial gTLDs could be auctioned as part of the expansion, either by ICANN (perhaps with delayed transfer to reflect extant agreements) or by the private operators as part of a double-sided auction.

For a variety of economic and practical reasons, we think incumbent registry owners should be grandfathered in, at least in the current round. First, they are already likely to take a significant hit with large-scale expansion of the root through loss of monopoly power and rents. Second, their pricing structure for SLDs, often for long terms, was set in reliance on making annual payments to ICANN rather than paying a capitalized up-front purchase price. Finally, the incumbent gTLDs have enormous influence on ICANN’s decisionmaking process. That reality of “regulatory capture” cannot be ignored in restructuring gTLD expansion policies. If incumbent expectancies were suddenly unsettled, it is less likely that the necessary paradigm shift would be approved.

In sum, we propose that ICANN authorize a first-round auction of 50 new gTLDs, for renewable terms where both the specific gTLDs and their operators would be selected by competitive bidding.

Sample Auction

Assume, for the sake of illustration, that 100 bidders meet eligibility standards and participate in the auction.482 When bidding starts, each participant can offer one or more gTLDs together with a bid amount.483 For instance, a Hollywood entrepreneur might make an opening bid of $250,000 for the right to operate .movie. Some bidders may be content to bid on gTLDs already on the bid list, rather than proffer their own. Other bidders will have a unique claim to a gTLD. For instance, if IBM proposes .ibm, no other bidder would be qualified to operate it. One might expect IBM to offer the reserve price ($50,000) and not a cent more, confident that no one else can outbid it. But $50,000 for .ibm might not be one of the top 50 bids, in which case, that gTLD would not be approved.

Bidding would take place in public over the Internet, much the same way that other on-line auctions are conducted. Between the open and close of bidding, all bids would be listed, showing rank, gTLD, amount, and bidder. It would hypothetically look something like this (only six shown):

Relative ranking in the top 50 is unimportant, as each will win the right to be added to the root. Accordingly, bids in this group are likely to be close to one another. The real battle will be waged at the boundary, between the 50th and 51st ranked bids. As the auction nears close, Nokia and Afilias will both raise their bids, trying to preemptively outflank the other, as well as bids higher up the ladder. Anyone who has lost an eBay auction can appreciate how seasoned bidders have perfected winning strategies, such as last-second stealth maneuvers. Indeed, bidding software helps those who are serious about the enterprise.484

This is an auction within an auction. For instance, if Hustler and Playboy both bid for .sex, only the higher of the two can succeed, even if they are both among the top 50 bids. Subjecting both gTLDs and operators to competitive bidding serves two salient purposes. First, it works as an anti-collusion mechanism. Competitors are unlikely to conspire to make low bids, since that could exclude both from the winning pool. Second, the auction identifies the 50 most valued gTLDs, not merely those who most want to operate registries. This not only maximizes revenue to ICANN, it puts the domain name space to the highest and best use, as measured by market players. To facilitate this assessment, the auction should remain open long enough (perhaps 30 days) so that participants can obtain feedback from potential customers and other constituents.

The most notable result of an auction along these lines is that the successful gTLDs are not likely to be ones that ICANN would select under existing policies and practices. Therein lies a principal purpose in preferring markets for the allocation of private goods.