ICANN has taken another crack at the question of the economics of launching new top-level domains (TLDs). The first report that the group commissioned on the subject was greeted by a loud and unhappy uproar. Now we have the preliminary draft of a new report, this one by professors Katz, Rosston, and Sullivan. It is insightful and analytic, but the final version needs to consider the theoretical and empirical issues outlined below.

Theoretical

1. Advantages of using a signaling framework

• Puts into focus the areas of unmet needs for new TLD signals/messages by new registry applicants and registrants. TLDs such as .com, .tel, and .me have strong signaling value propositions. For example, .com has practically no substitutes for signaling a global brand. TLDs that signal location include country-code TLDs (ccTLDs) and some proposed TLDs such as .NYC (which signals New York City). TLDs that signal a particular business strategy include .outlet and .eco. The .tel has a strong use differentiation because it signals the brand owner’s alternative contact information, while .me is personal and reassuring, as opposed to the chilly and faceless .name.

• Clarifies the strategic approach that needs to be followed by new TLD registry applicants. For example, to compete with .com, a product differentiation strategy needs to consider established network effects and to recognize that the argument for shorter second-level domain names is not viable. On the other hand, new unmet needs require a strategy for expanding the pie.

• Temporal approval decisions have to take into account the type of TLD signal. Otherwise, there might not be any informational benefits from sequential launches. Without a signaling framework, a multitemporal approval mechanism would ignore the reality of first mover advantage (FMA). Consider .green and .eco, two initial substitute-signal applicants. Quite an unfair advantage would accrue to .eco if it were approved first, followed by .green after a considerable wait.

• One cannot perform market power analysis without intuitively knowing what constitutes similarity signals. Numerical measures of substitution effects may not be reliable. For example, no matter what the numbers may say, the signals from .com and .me are intuitively different.

2. Externalities: The report

• Does not identify the sources of domain name externalities so as to work on reducing them.

• Uses a framework more suited for downstream analysis and ignores the possibility of an upstream-produced externality, namely one produced by ICANN.

• Considers trademark infringements and search costs as operating costs, though arguably they are externalities (within the framework of the report).

• Ignores the costs of potential rogue TLDs, whose private benefits outweigh their social value.

3. Instead of adopting a general social-private cost-benefit framework, the report can narrow down the scope of the analysis to, say, search, navigation, companies, and registries.

4. The report proposes no solution to trademark infringements except establishment of a clearinghouse. It ignores the benefits of establishing a cooperative regime as a complement to any registry-level trademark solution. An effective trademark regime can only be reached and implemented through negotiations.

5. The report ignores the distinction between defensive and offensive second-level domain registrations. The latter are value adding and thus should not be automatically labeled as a net operating cost.

Empirical

1. Without a signaling framework, the number of registrations of various existing TLDs cannot be used to estimate a TLD’s demand and/or its market power. The lack of registrations by brand owners under certain TLDs can be due to the irrelevance of their signal to the brand name. Hence, I agree with the report’s assessment that registrations of new TLDs under currently underserved signals would increase the cost of infringement rates and/or cybersquatting costs significantly.

2. The economic rationale for a domain registration is that its value must be greater than its cost. Statistical pricing models have been developed that can shed light on the value of keyword-based domain names. Moreover, such models identify statistically significant factors that drive prices for different TLDs and are useful in estimating price-premium variations over time. By contrast, using average and/or median sale prices is practically useless, as prices of various statistically comparable domain names fluctuate at different rates; during the same periods, prices of comparable domain names have not always moved in the same direction nor magnitude.

3. Such statistical models can also be used to estimate cross-price elasticity of demand for purposes of determining market power and competition.