Minds + Machines comments on the latest economic report

[Antony Van Couvering <avc@xxxxxxxxxxxxxxxxxxxx>]

Thanks once again to ICANN for the opportunity to comment.  These comments 
refer to the Economic Study, Phase Two, by Michael L. Katz Gregory L. Rosston 
Theresa Sullivan.

My comments on the second phase of the ICANN-commissioned economic study by 
Katz, Rosston, and Sullivan are in two parts:  the value of forward-looking 
economic studies in general; second, the methodology, information, assertions, 
and conclusions in the study itself. 

ICANN is conducting these studies because it must: ICANN is required to do so 
under its obligations under the Affirmation of Commitments.  Though the AoC is 
a fine statement of a political accommodation and working principles for 
Internet governance, and it does require ICANN to conduct an economic study, 
that requirement is no less silly, and no-one should be fooled into thinking 
that such studies are useful for making policy. On the contrary: no-one at 
ICANN, the GAC, the Department of Commerce, or any other group or person 
involved in the new gTLD process, including the authors of the numerous studies 
ICANN has commissioned, has presented or even referred to any evidence that 
economic studies have any value or validity for understanding the effects of 
new gTLDs, a field with many unknowns.  Even in thoroughly studied areas such 
as the economy of the United States, the very best minds in the field have 
shown their utter incapacity for predicting events or even large-scale trends, 
such as the meltdown of the U.S. housing market, or the effect of stimulus 
spending or tax relief on the unemployment rate.  It should be no surprise, 
therefore, that economists' predictions about new gTLDs, which will create an 
entirely new market, are without merit. The only solid prediction resulting 
from the AoC mandate, sadly, is that ICANN would spend millions of dollars to 
produce studies that rival augury or phrenology in their accuracy. 

Which is not to say that the authors of the latest study actually produce any 
hypotheses that can be tested, or provide much guidance for policy makers. 
Rather, the study says, essentially, that new gTLDs might produce some 
benefits, or the might produce some harms, or that it might be an admixture of 
the two. 

There is one area, though, where the study takes a stab at a testable 
prediction, when it says: "[W]e find that additional generic, unrestricted TLDs 
using the Latin alphabet would be unlikely to provide significant additional 
competition for .com." The evidence adduced for this conclusion, however, is 
deeply flawed, in several respects:

1. It uses information that is over ten years old (the introduction of .info 
and .biz), even though there is information that is less than a few months old 
that is far more to the point (the introduction of .co in Colombia, and the IDN 
version of .rf in the Russian Federation).  The comparison of these new (and 
successful) TLDs to an existing (successful) TLD like .com would have been more 
useful. 

2. It assumes, without justification or explanation, that all Latin-character 
unrestricted gTLDs have the same appeal, and are semantically undifferentiated. 
 On page 8, the study says, "If past gTLDs have relieved name scarcity, then 
one might, for example, see Moespizza.com for a restaurant in New York and 
Moespizza.biz for a restaurant in San Francisco."  To highlight how unjustified 
this assumption is, let me rephrase this: "If past gTLDs relieved name 
scarcity, then one might, for example, see Moespizza.goodname for a restaurant 
in New York and Moespizza.badname in San Francisco."  With no disrespect to the 
operators of .biz, it is an uncomfortable, faddish abbreviation, and is widely 
recognized as such. What makes the authors suppose that new gTLD applicants 
will choose uncompetitive TLD names, when applicants have been unanimous in 
their determination to pick strings that have strong appeal?  

3. The authors say (p. 3) that "by studying a small sample of generic words, 
and we do not find evidence that scarcity of generic second-level domain names 
is a pervasive problem; in a high percentage of cases studied, generic terms 
are unregistered or unused on several different gTLDs."  This is a flawed 
method of trying to understand historical trends, let alone predict the future. 
 The sample size is far too small, *as the authors themselves admit*.  I 
understand first-hand the hard work involved in putting together a truly useful 
sample size, but hard work is not an excuse for an unscientific sample. 

4. It commits a bad mistake by using zone files to try to understand 
registration patterns, because zone files tell you nothing about names not 
registered for reasons other than being previously registered. For instance, 
names may be reserved, and in fact of the five strings they examined, several 
TLDs reserved these names, tainting their results.

A few additional comments might be helpful to the Board as they consider this 
issue:

- Each new economic study costs quite a bit of money; at some point, these 
costs will need to be passed on to registrants, and will impact ICANN's ability 
to undertake other worthy efforts, such as providing assistance to underfunded 
applicants.

- Each of previous gTLDs chosen in previous rounds were chosen by ICANN. The 
TLD mentioned before as not the best, .biz, was not the first choice of 
Neustar.  The new round will finally allow people who understand their 
customers and markets to choose the best names and market them to customer they 
know.  The authors of the study do not mention this essential point. 

I appreciate the opportunity to comment.

Antony Van Couvering
CEO, Minds + Machines