Home / Blogs

An Economic Analysis of Domain Name Policy - Part I

“The Root Server is a Scarce Resource” is the focus of part one 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.

We begin our analysis of domain name policy with a brief excursion into economics. Economics cannot answer all of the questions raised by domain name policy. First, domain name policy must answer to the discipline of network engineering. A useful domain name system must work, and the functionality, scalability, reliability, and stability of the system are determined by the soundness of its engineering. Second, domain name policy must answer to public policy. The Internet is a global network of networks, and Internet policy is answerable to a variety of constituencies, including national governments, the operators of the ccTLDs, Internet Service Providers, information providers, end users of the Internet, and many others. ICANN is a nonprofit corporation required by law to serve the public interest. An economic analysis of root service can answer some, but not all, of the fundamental questions that are raised by domain name policy.

Nonetheless, an economic analysis of domain name policy has an important, indeed a crucial role to play. The Internet has been mythologized, and the image of the Internet as a separate realm, somehow exempt from legal regulation and the operation of market forces is still a powerful and compelling ideal in the minds of many. Although this romantic picture may have an element of truth, there is much to be learned by stepping back and looking at root service as an ordinary service, provided by an ordinary organization, subject to the familiar laws of supply and demand. How can the provision of that service be organized so as to provide the greatest benefit of the public? How can the root be put to its highest and best use?

Our answers to these questions begin with a basic truth about the DNS: The Root Is a Scarce Resource

The Root Server System Is an Economically Scarce Resource

The root server system is scarce in the economic sense, because root service is not free. To provide root service, there must be root servers (computers that provide root service to the Internet). The root servers (i.e. the computers) are not free. The software that runs the root servers is not free—it took human labor to create that software. The root servers must be maintained, and the labor that does the maintenance is not free. Therefore, root service is scarce in the economic sense. This is not altered by the fact that root server operators donate their hardware and services to the system—the resources that they are donating are scarce in the economic sense.

This same point can be made by examining the capacity of the root server system. The precise capacity of the root server system is not documented and probably the upper limit has not been tested. We shall assume that a conservative upper limit on the current capacity of the system is from 1,000 to 10,000 TLDs. There is demand for TLDs, and the supply is, at least temporarily, limited by the capacity of the root system. It follows from the definition of economic scarcity that TLDs are scarce, and therefore, if TLDs were sold (as we suggest they ought to be), they will command a price.

So far our discussion has assumed a static root server system, but the hardware and software that provide root service is not fixed in stone; therefore, the potential upper limit on the supply of TLDs is not inherently static. Even if the existing root server system can only support thousands of TLDs, a new and improved root server system might be able to support millions or even billions of TLDs. But this does not change the fact that root service is a scarce resource in the economic sense. Upgrading the root server system would be costly. The fact that additional root service capacity would be costly is sufficient to establish the conclusion that root service is scarce in the economic sense.

The Root Name Space Is an Economically Scarce Resource

There is another reason why the root is a scarce resource. The root name space is itself scarce. Names are simply strings of characters. If there were no upper limit on the length of a name string, then the name space as a logical construct would be infinite. Even if the limit were 10, the number of logically possible names would be in the quadrillions—so many that there are no conceivable circumstances under which the supply would be exhausted.

Nonetheless, the name space is a scarce economic resource. The economic scarcity of the root flows from the differential value of different names. For example, “.com” is more valuable than “.kjd- 7xx9-a,” and this is true despite the fact that either string will do equally well at the job of connecting an end user to a server on the net.

There are two reasons to believe that different names have different values. The first reason is that the market prices different names differently. The second reason is that different names have different useful characteristics; another way of putting this second point is to say that different names have different functional utilities.

Different Names Have Different Prices

Consider the first point. Different names have different prices. This can be demonstrated in two ways. First, experience in the SLD markets demonstrates that different names have different economic values. ICANN has imposed uniform pricing across strings by contract in the primary market for SLDs registered in .com, .net, and .org. If you go to a particular registrar, that registrar will charge you the same price for “auto.com” (if it were available) as for “7skj989_2.com” (if it were available). There is, however, a secondary market for domain names in these gTLDs, and on that secondary market some SLD name strings command much higher prices than others. For example, in 1997 the SLD “business.com” was sold for $150,000.00. And taste.co.uk was sold for £110,000. Moreover, price differentials are allowed for some ccTLDs. The ccTLD for Tuvalu (.tv) markets different names at differential prices. It is beyond dispute that different SLDs sell for different prices in both the secondary and primary markets.

So far, we have been discussing SLDs. Would there be differential prices for TLDs if there were a market? That the answer to this question is “yes,” can be established by two arguments. First, there is simply no account of the functional differences between SLDs and TLDs that would predict massive price difference for the former and no price differences for the latter. Second, a simple thought experiment establishes that prices would indeed be different. Imagine, for example, that ICANN conducted an auction for the following two TLD strings: “.sex” and “.8ki3_d.” We doubt that any reader believes these two strings would receive equal bids—assuming the winning bidder was allowed to charge market prices for registration with their TLD. It is obvious that .sex would command a relatively high price, and the other string would either attract no bids or bids at a much lower level.

Different Names Have Different Functional Utilities

There is a good reason why the market prices different strings differently. Not all names are created equal. Some names are more useful than others. Why? Because names have semantic and syntactic properties that affect their usability. We have identified five such properties, although we suspect that there are more.


One reason why some names are more valuable than others has to do with the way end users interact with the DNS. We have learned that you frequently can get to a useful domain on the basis of a guess. Thus, if I want to go to the website for the International Business Machines Corporation, I am fairly likely to get there by guessing “www.ibm.com.” If I want to go to a website that provides products, services, or information relating to automobiles, it is likely that entering “www.auto.com” or “www.car.com” or “www.automobile. com” will do the trick.

The examples in the previous paragraph are SLDs, but the same thing is likely to be true of TLDs. It is no accident that the new TLDs created by ICANN all have names that facilitate guessing; an obvious example is .museum. If the root were expanded and there were many new TLDs, the guessing that takes place in .com TLD would likely take place in the root. For this reason, guessable names are likely to be more valuable than names that cannot be guessed.


Some names are more memorable than others. For example, www.amazon.com is not a guessable name for an online book merchant, but it is a memorable name. Strings that are memorable, such as .biz or .web, are more valuable than strings that are difficult to remember. Memorability can be a function of either the syntactic or semantic properties of string. The string “ZZZ” might be memorable, although it lacks semantic meaning. Likewise, the string “perspicaciousness” might be very difficult to remember, even though it does have semantic meaning. The string “8k” is very short, but might not be memorable. The string “remember-nine-eleven” might be memorable although it is quite long by domain name standards.


Yet another reason for differential utility is branding. Some strings are more valuable than others because they are more easily associated with particular firms or brands. Thus, if IBM were to operate a proprietary TLD, the string .ibm would have greater utility than .9ks. IBM is a contraction for International Business Machines, but branding does not depend on semantic meaning. Toyota is an effective brand name, even though it had no prior semantic meaning in English.

It might be argued that branding utility can only create differential value if cybersquatting is permitted, but this is not the case. If we assume a perfect mechanism for prevention of cybersquatting on trademarks, branding will still create differential value, assuming a limited number of slots for new TLDs. Suppose, for multinational corporations (such as IBM, Sony, or Nike) would be willing to bid for a slot and the string naturally associated with their brand names, even if no other party would be allowed to bid on that string.


Meaningful strings are strings that have semantic meaning in some natural language. Thus “air,” “aero,” and “luft”—all are meaningful strings. All else being equal, meaningful strings are likely to be more memorable or guessable or brandable, and hence more valuable. Not all meanings are created equal, however. Some meaningful strings are likely to have low value, because of negative connotations. Thus “dour,” “prissy,” and “putrid” are likely to be low value strings—although they are meaningful, memorable, and relatively easy to enter. On the other hand, some strings have positive connotations. Thus, “awesome,” “excellent,” “cool,” or perhaps “phat” are likely to be higher value strings.


Strings that are easy to enter have more value than strings that are difficulty to enter. Thus “fj” might be a more valuable string than “zp,” because of the placement of these characters on the standard qwerty keyboard. Shorter strings are more easily entered than longer strings, and familiar strings are more easily entered than strange strings. Because of the pervasive use of QWERTY keyboards, some strings produce fewer entry errors than others simply because a touch typist is more likely to strike the keys correctly.


Undoubtedly our list of factors that create differential functional utility for strings is incomplete. But guessability, memorability, branding, meaningfulness and enterability are sufficient to establish our thesis. Different strings have different functional utilities. When this functional utility is combined with our prior point about differential prices, the support for our conclusion is very powerful indeed. The name space is a scarce resource, because each string is unique and different strings have different values. This point is reinforced by comparing the name space with another resource: land. One might argue that land is not scarce. There is plenty of unused land. If someone needs land, they could be allocated a parcel from the unused stock (such as in the Gobi desert or Antarctica). But from the fact that the unused supply of land is enormous, it does not follow that land is not a scarce resource in the economic sense. Land is valuable. Although some land is so cheap that it is virtually free, many parcels of land command extremely high prices. That is because different parcels of land have different functional utilities, based on a combination of factors, such as location, flatness, soil quality, and so forth. Although there are hundreds of millions of acres of unused land, land is a scarce resource in the economic sense. As with land, so with names.

By CircleID Reporter

CircleID’s internal staff reporting on news tips and developing stories. Do you have information the professional Internet community should be aware of? Contact us.

Visit Page

Filed Under


Doctor J  –  May 27, 2004 4:46 PM

This is an excellent analysis of the underlying factors explaining the economics of TLD names, and it has some interesting implications for improving our understanding of the economics of SLD names, as well as related public policy issues.

Some important, closely related, economic issues the authors don’t mention are the problems of the “tragedy of the commons”, and the problem of externalities, both of which are adversely affecting the current system of TLD names.

For example, in creating the TLDs of .COM, .NET and .ORG, it should have been easy for users to “guess” where a particular site will be found, since commercial sites were supposed to have names that end with .COM, network-service provider sites would have names that end with .NET, schools would have names that end with .EDU, and other non-profit and personal sites would have names that end with .ORG, etc.

Unfortunately, this system has become heavily polluted with inconsistencies, reducing the benefits to all participants (both sites and users).

Because the actual names being deployed are failing to follow this conceptual schema (and there are not adequate rules to force participants to follow the system) the total economic benefits from the current TLD system are not being maximized.

One obvious consequence is that artificial scarcity has been created, because so many different types of sites are all trying to crowd into the same limited TLD spaces (especially .COM). Nowadays, even universities (who presumably know better) are ending their sites with .COM, rather than .EDU, creating additional scarcity in the .COM space, and making it harder for users to “guess” where a school’s website will be located.

This is similar to what happens when everyone tries to graze their sheep in the particular pasture that happens to be the most convenient, or happens to have the nicest grass—it becomes overcrowded, and everyone is worse off than if some of the farmers were willing (or willing to be paid) to graze their sheep in a less desirable location.

Putting too many different types of sites into the same TLD space also reduces the user benefits cited by the authors. Users cannot accurately “guess” or predict where a particular site will be located when the TLDs lack predictable meaning.

Proliferating more gTLDs will not solve this problem, unless the gTLD expansion is combined with powerful mechanisms to overcome the tragedy of the commons and space pollution problems.

Site operators need incentives (or need to be forced) to only use TLDs for their stated purpose; thereby ensuring that other site operators operating in that “neighborhood” don’t suffer from pollution/confusion.

Consider an extreme example: assume a new TLD is created called .KIDS.  The value of this TLD would be greatly diminished, or destroyed, if porno site operators were allowed to operate in this new TLD space. Someone going to FUN.KIDS wouldn’t be able to predict if they are going to see simulated child pornography, or a site with all sorts of fun things for children to enjoy.

Proliferating more gTLDs won’t solve the current problems unless this effort is part of a much broader, more sophisticated policy solution.

Comment Title:

  Notify me of follow-up comments

We encourage you to post comments and engage in discussions that advance this post through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can report it using the link at the end of each comment. Views expressed in the comments do not represent those of CircleID. For more information on our comment policy, see Codes of Conduct.

CircleID Newsletter The Weekly Wrap

More and more professionals are choosing to publish critical posts on CircleID from all corners of the Internet industry. If you find it hard to keep up daily, consider subscribing to our weekly digest. We will provide you a convenient summary report once a week sent directly to your inbox. It's a quick and easy read.

I make a point of reading CircleID. There is no getting around the utility of knowing what thoughtful people are thinking and saying about our industry.

Co-designer of the TCP/IP Protocols & the Architecture of the Internet



IPv4 Markets

Sponsored byIPv4.Global


Sponsored byVerisign

Domain Names

Sponsored byVerisign

New TLDs

Sponsored byRadix

Threat Intelligence

Sponsored byWhoisXML API

Brand Protection

Sponsored byCSC