ICANN recently commissioned a report from Power Auctions LLC to investigate the merits of auctioning new Top-Level Domains (TLDs). Below I outline some of the issues related to stakeholder interests and mechanism design. (See related post 1 and post 2.)

Successfully managing the design of an allocation mechanism for new TLDs will entail coordinating functions across various competencies. To rely on a standard auction mechanism for the allocations would be a historic setback for the domain name industry, as successful allocation design is all in the details.

Before developing a mechanism for allocating TLDs, we must compile a list of activities that need to be coordinated, identify the various abilities and areas of knowledge the activities require, and then select a project manager—ICANN or another body—that has these competencies. These will certainly cover a wide range, with technology, marketing, mechanism design, and protection of stakeholder interests all being involved.

Stakeholder Interests

1. Determining the objectives of the allocation process. There are four plausible objectives, and these are not necessarily mutually exclusive:

a. Extracting the highest possible value from the allocations.
b. Protecting trademarks and brand names.
c. Minimizing irrational biding exuberance.¹
d. Resolving potential feature-related conflicts between the domain name community and bidders.

2. Addressing what to do with allocation proceeds. Plausible uses are:

a. Improving Internet infrastructure.
b. Supporting ccTLDs of developing countries.
c. Lowering registration fees.
d. Distributing proceeds to existing domain name owners.
e. Lending support to the Internet Commerce Association (ICA).
f. Supporting the creation of a market mechanism to solve conflict arising from brand use in domain names.

3. Prior to launch, it is imperative that potential bidders know about the allocation mechanism and that they know it will be safe.

Allocation Mechanisms

In theory, the objective is to award each TLD to the entity that values it most. The most obvious way to do this is to use an English Auction, as popularized by eBay. But, in practice, a mechanism design works only if its rules are tailored to specific desirable objectives, and if the designer can predict how participants will actually respond to the rules. Thus, the design we choose must be created specifically for allocating TLDs.

The biggest attraction of Internet auctions is that buyers do not all have to be at the same place to participate. Moreover, they don’t have to place bids at the same time or wait for the last minute to bid.² Nevertheless, live domain auctions, including those for expired domain names, are flourishing. We must decide whether live auctions are desirable for domain name buyers, and whether they are therefore desirable for allocating TLDs in particular.

It is important to get the design right in the first place. The current on-line advertising auctions went through a number of iterations, but allocating TLDs is a onetime process with no room for experimentation. Moreover, when switching costs are too high or, say, the allocation process is too complicated, a new superior mechanism may not be adopted. Ease of use and a transparent allocation process are vital. Google’s³ current auction design, by which sponsored ads are ranked, is not necessarily optimal. In fact it has two undesirable features,⁴ and some alternative designs present clear advantages in certain areas.

The team in charge of allocation design should have a deep understanding of the theoretical design issues, be intimately familiar with the domain industry, and be able to draw on the successes and failures of previous onetime auction design implementations.

¹ Technically this is the “winner’s curse.”
² In live auctions, there are strategic and tactical reasons to bid at the last minute that are not driven by irrational bidding exuberance (“winner’s curse”).
³ Yahoo’s current advertising-allocation auction mechanism falls within the same class of auctions but is not identical.
⁴ Technically, the mechanism does not have equilibrium in dominant strategies and truth-telling is not an equilibrium.