Presented at MIT Workshop on Internet Economics March 1995

Introduction

A year ago, debate was in full force over commercialization of the Internet. At one extreme were the proponents of the classic market approach who revere the genius of the market exchange system. It was time, they said, for the government (the National Science Foundation, or NSF) to step aside and let the free market—that system of entities that produce and exchange goods in markets—determine the future growth and direction of the Internet. At the other extreme were the evangelists who saw the Internet as a great equalizer in the new information economy; they claimed that only strong government intervention and subsidy would allow the Internet's full potential to be realized.

Today, one year later, NSF has stepped away from supporting commodity inter-regional connectivity by ending the services known as the "NSFNET backbone." Non-profit regional networks as well as for-profit Internet service providers now contract directly with national network service providers to achieve the same level of connectivity that they had when the NSFNET backbone was the heart of the Internet. Registrations for domain names and IP addresses by commercial organizations now far exceed those by academic and other non-profit institutions.

Other changes have also taken place. The commercial Internet marketplace is no longer solely the venue of daring commercial entrepreneurs marketing "bleeding edge" technology to equally daring early adopters. Giant corporations such as Time-Warner and Microsoft are entering the Internet-access market to offer commodity connections using market pricing strategies for individual consumers and organizations alike.

The temptation to claim that the classic market proponents indeed have won the debate would appear to be supported by the rhetoric of the press and the hype of commercial advertisements. However, as we have demonstrated in an earlier paper, even a competitive market is not a panacea for allocation of goods and services of all economic characteristics, and, in fact, it is not the economically-efficient solution for large classes of goods.

Economic reality is more complex than either extreme position (market pricing vs. full government subsidy) might suggest. Goods of different economic characteristics exist simultaneously in all real, functioning economies. Efficiency requires different solutions to allocation problems for goods of different economic characteristics. Clearly, private markets for IP connection services are growing, as shown by the increasing number of private entrants into the connections market. But the Internet requires more than a simple connection. All of its components—the technology; the consumers; the suppliers of the myriad forms of electronic, communication, information, and human creativity—are continually changing. The economy based on those components is as complex as they are. To attempt to use a single model to explain and efficiently allocate such a diverse, dynamic, complex set of elements is not just inappropriate, it is impossible.

This paper repeats from the prior paper a statement of the basic economic foundation for understanding the different kinds of goods in an economy and for understanding the trading mechanisms that result in their efficient distribution in society. Specifically, we examine public and private goods and some in between, plus the allocation mechanisms that are appropriate for each. We examine two Internet "goods," neither of which in its initial form can be efficiently allocated through markets, first a widely-used piece of Internet routing software called GateDaemon (or GateD—pronounced 'gate-dee'), and second, (uncongested) electronic networks. We consider Cornell University's implementation of a funding model for GateD based on its economic characteristics as an economic public good. Electronic networks have the same economic characteristics and can be efficiently allocated in the same way. Then, we analyze congested networks—an intermediate form of economic good explained, discussed, and analyzed below. Much study, both economic and technical, is being done on network congestion and its effects. We explore congestion based on its economic characteristics and examine ways to deal with its intermediate nature—one solution for which is clearly the preferable choice under the real-world conditions of the Internet. We conclude with a summary of the model and its extensibility.

Economic Definitions: Public Goods, Private Goods

It is important to recognize that there is more than one kind of economic good in the world. ("Good" is a generic term for something that is exchanged.) It might be a candy bar, a piece of software, a book, or a service. But no matter what it may actually be, the principles presented below still apply.

There are private goods and public goods. An economic public good does not necessarily imply something provided by a government. "NYPD Blue," a TV program broadcast by ABC Television has the economic characteristics of a public good, for example. A public good has a precise economic definition based on the characteristics of the good itself, not the characteristics of the provider. A public good is a good that is non-depletable and non-excludable, as we illustrate in a moment.

Private goods are both depletable and excludable. An example of a private good is a candy bar. When one person purchases a candy bar, there is one fewer bar available for the next person; the number of candy bars is reduced or depleted by one. If a person eats this candy bar, then all others are excluded from receiving its benefits.

A public good, in contrast, is non-depletable. This means that when it is used by one person, what is available to others is not depleted in quality or quantity. A public good is non-excludable. What is available to one is available to all. This means that the use of the good by one person will not exclude others from its benefits in any way. Simply putting an antenna on my roof, turning on my TV set, and watching "NYPD Blue," does not mar your reception: my reception of the signal does not "deplete" the signal available to you; neither does my viewing of the program somehow "exclude" you from doing the same thing at the same time at your home. Public goods are non-excludable.

The most commonly used example of a public good is national defense. National defense is both non-depletable and non-excludable. Everyone in the country, including any newcomer or newborn, is protected simultaneously and to the same degree by the system of national defense (whatever it is). Because your neighbor is protected does not mean you are protected any less. The resource is not depleted by being used by your neighbor; and because the neighbor partakes of its benefits does not mean that others are excluded from the same benefits.

Economic Definition: Externalities

There is a third class of economic goods that falls between pure private goods and pure public goods. These are goods with "externalities." The unintended "spillover" of any good is called an externality. If the spillover is positive (e.g. a research breakthrough opening a new avenue of commerce to all), then it is a positive externality, a benefit; if the spillover is negative (e.g. pollution from automobiles), then it is a negative externality, a cost to society. In some cases, the positive economic spillover may actually be of more benefit than the intended benefit of the good to its original creator, as often occurs with research. But, since they are unplanned and cannot be captured by the creator, externalities may be hard to quantify.

Externalities are themselves of two economic types. Public goods externalities are unintended spillovers that are neither depletable nor excludable. They are themselves "public goods" (or "bads," as with pollution), but since they are not intended by those who create them, they are public good externalities, rather than "plain old" public goods. An example is U.S. national defense that simultaneously "spills over" to protect a nation or group whose protection was not intended by the United States at the time it made its expenditure on defense.

Unlike the case of public good externalities that differ from plain-old public goods only by the intent of their creator, private goods externalities represent a fundamentally different class of economic good: it is a good that is a half-step from a public good toward a private good. Private goods externalities are externalities that are depletable, but not (effectively) excludable. An example of a private goods externality is an underground oil pool that can be accessed from multiple surface locations. The pool is depletable, but not effectively excludable (assuming U.S. law on mineral rights). Under current international law, ocean fishing has the characteristics of an economic private goods externality—unfortunately. Network congestion is another example of a private goods externality—an example that we deal with extensively below.

Principles for Efficient Pricing of Public and Private Goods

Once it is understood that there are both public goods and private goods (among others) in an economy, the next step is to understand how to allocate the different kinds of goods efficiently. A basic economic principle of efficient pricing for either type of good is that price must be equal to "society's" marginal cost, the cost of producing a unit of the good for the next user. For private goods, efficient allocation is achieved through competitively-structured markets in response to supply and demand. Price equates to marginal cost through the "invisible hand" of market forces. The "right" level of output is determined at a quantity that equates to the most efficient output point for the firm, its minimum average cost point, and simultaneously to the marginal cost of the good.

Prices determined in markets are "symmetrical." In other words, the producer receives the same price as the buyer pays. When a private good is allocated through a competitive market at equilibrium, the pricing is efficient as well as symmetric. The price equals the marginal cost of producing the good; in economic terms, this "cost" includes a "normal return on investment:" a sufficient incentive to the private firm to continue production of the good.

A public good is a good that, once produced, is undiminished by being used by one or more users (it is non-depletable) and available to all (it is non-excludable). This means that, by definition, the marginal cost of supplying the public good to the next user is zero. Recall that the requirement for efficient pricing of a public good is the same as for a private good. The efficient price must equal the marginal cost of the good. In the case of a public good, the efficient price is zero!

The "kicker" in the definition of a public good in the paragraph above comes with the words, "once produced." How do you get the good produced if you are "selling" it at a price of zero? To get the good produced in the first place is often quite costly, as it is for "NYPD Blue" (millions of dollars), for national defense (trillions of dollars), or the Internet (indeterminate). The producer must be paid "enough" to produce the good; at the same time, efficient allocation requires that the good be given away. It is impossible to retain an efficient price to the user and deliver a public good through what we normally call a "market." For society to receive full benefit of the good, a different method of pricing must be found.

In the case of a public good, "asymmetric" pricing must be used: the price paid to the producer must be different from the price paid by the consumer. This type of pricing requires formal or informal economic "taxes" to generate enough resources to get the good produced and still be able to provide it to users at a price of zero. For example, for commercial television, the economic "tax" is informally added to the price the consumer pays for the goods advertised; for public television, the "tax" is raised through pledge drives. (And both involve a costly, negative, public good externality—watching the "commercial" through which the tax is raised.) So "economic taxation" does not require involvement by a government agency, as might be one's initial reaction to the concept of a "tax." The method any organization uses to implement asymmetric pricing should be recognized as a form of economic taxation. It is a means of collecting, from whatever sources, funds that can be used for production of a particular public good. The organization collects the tax with one hand, then pays the money to those who produce the public good—e.g., the military industrial complex for national defense—and then with the other hand, provides the good "free" to the country. It is important to recognize that the tax itself cannot be tied to the use of the public good, or it becomes a de facto price, greater than the marginal cost of zero.

What is GateDaemon?

As the first organization to manage the NSFNET, Cornell University had to find a way to connect existing (and future) networks. It therefore created gateway software that it called "GateD" to accomplish the task. GateDaemon or GateD is a modular software program that implements multiple routing protocol families on a Unix-based hardware platform. It is the heart of an IP router. The router is a critical element of a network. It links organizations together as it transfers blocks of information from one place to another. Originally developed to link the early regional networks with the original NSFNET backbone, GateD was designed to listen to different routing protocols and choose the best route for traffic to a given network. Cornell has chosen to make GateD freely available to the worldwide Internet community. NSF (federal tax dollars) supported the initial development of GateD directly and indirectly, making it possible for Cornell to price it in accord with its economic characteristics.

The GateD project continues its history of being the routing arbitration software used between organizations for complex routing situations that cannot be handled by commercial products. It is a de facto reference implementation for many routing protocols. It can be used with independently developed routing protocol implementations in interoperability tests. It is a research and prototyping platform, allowing the community to focus on resolving critical problems such as scaling. It is important to understand that its value comes not only from its functionality at a given point, but also from the continuing development of enhanced functionality, plus maintenance updates and bug-fixes.

A Public Good for the Internet

GateD is an example of an economic public good, both in definition and in practice. Once the software was developed, the cost of making it available to the next user was (and is) close to zero; its marginal cost approximates zero. Efficient resource allocation requires that it be priced at its marginal cost—essentially "given away." Its value to one user is not diminished by making it available to other users. In fact, that value is enhanced—one of many positive externalities that arise from wide use and connectivity of the networks.

The commercial market for IP router boxes is flourishing, thanks in part to GateD. Several organizations (U.S. and non U.S.) use GateD to jump-start and enhance their own development efforts for their networks or for new, value-added router products. These physical routers themselves are private goods: they are depletable and excludable. As part of the conditions of use by those organizations that intend to redistribute the GateD software in their own hardware products, is the requirement to sign a redistribution license stipulating that the enhancements they introduce into the software must be turned back to Cornell for inclusion into the public distribution. Commercial companies that use GateD in their hardware then become part of the community collaboration contributing to the public distribution of the software, rather than simply taking from it.

GateD provides a (financially) low-end entry for organizations wishing to connect into the Internet. It can be installed on an existing Unix platform being used as a file server or bulletin board server. It can link that organization into the Internet without necessitating the purchase of a new (private good) router.

Because of its widespread use both as publicly available software and as the heart of many private good commercial hardware products, it is a powerful deployment mechanism for necessary routing enhancements needed to sustain the current growth of the Internet. For example, its implementation of BGP4 and CIDR will be among the first in the U.S.-portion of the Internet.

In addition, its widespread use has guaranteed a level of interoperability between and among networks whose routers are based on it, or that have tested against it. Here, an unanticipated level and scope of interoperability has developed. GateD has created a worldwide de facto standard, in large part because it is a quality product, and was (and is) priced at zero, i.e., given away.

Implementing an Asymmetric Funding Model

Cornell University's goal for the GateD software project is to provide a freely available "state-of-the-art" software platform to support the most current routing protocols for the Internet community. As noted, initial funding came from the National Science Foundation. Then, as with so many important projects, more and more groups wanted more and more functionality and applied significant pressure to the project, to the point that the demands far exceeded then-currently available resources. Given the financial constraints on universities, Cornell could not and cannot use internal funds to meet these demands. A knee-jerk response might have been to make GateD a commercial product and distribute it "through the market." However, that would violate the rules for efficient allocation of goods with public goods characteristics and would more than proportionally undermine the spillover benefits that are catalyzed by GateD's free distribution.

To keep the GateD software freely available requires classic asymmetric pricing. Instead of "marketing" the software as a product with a licensing or royalty fee, Cornell has created a consortium to raise the necessary funding for GateD. This has answered the needs of the community, while remaining consistent with the economic realities of the good. It permits zero-price to the user, while generating sufficient resources to pay for continuing development. As noted earlier, a funding mechanism had to be found that was independent of the use of the product; that is, it could not represent a de facto price for the software.

Cornell's strategy is to look both to government and to other organizations for funding through lump-sum contributions (i.e. membership fees). It uses no fees tied to "use" or number of implementations. Nor is GateD access tied to membership in the consortium.

NSF and other grant-giving agencies that have funded GateD recognized it as an important element of Internet infrastructure. However, even though direct and spillover benefits far exceed the cost of funding the project, this was not the basis on which NSF did so. Instead, with each new grant, it required a detailed project plan describing technical enhancements that are of value to the U.S. infrastructure effort.

In something of a departure from the pure model for public goods, Cornell has also focused some of its fund-raising efforts on those organizations that benefit directly and indirectly from having GateD freely available. The Cornell GateDaemon Consortium, an international affiliates program, fosters and expands the already successful community collaborations centering on the development of the GateDaemon software. Prospective members for the consortium come from industry, government, and academia. Membership fees—the consortium's equivalent of "taxes"—are collected by Cornell and are used directly to support the development effort.

GateD Consortium: Successful or Not?

There are at least two ways to judge the success of the consortium. Clearly the GateD software satisfies the economic definition of a public good and a public good with public goods externalities. In practice, it is easy to see the economic benefits derived from the availability of GateD and, given enough time and effort, those benefits could be quantified. That Cornell has been able to get funding from the National Science Foundation, itself funded through public taxation, and attract membership in the consortium, a form of private taxation, is further proof that this model can work.

The current challenge for the consortium is to generate enough funding to increase the development staff resources in order to deliver critical enhancements and updates to the community in a timely fashion. This does not happen automatically. It requires considerable effort to convince individual organizations in the Internet community to pay these economic taxes voluntarily. Often, firms do not perceive an economic incentive to join the consortium. Since the software is already available to them for free, many organizations are content to be "free riders," the classic problem of dealing with public goods, especially with public goods externalities, when the providing agency does not have the power to impose taxes. This is the bad news. The good news is that even by free-loading, these organizations, nonetheless, have helped establish GateD as a de facto standard for Internet connectivity—a huge spillover benefit.

As a further compromise to help it overcome the free-rider problem, Cornell has used targeted benefits as direct and measurable incentives for organizations to join the consortium. A firm that bases its private good hardware product on GateD is given the opportunity to attend technical briefings or utilize a technical residency at Cornell, for example. It can quantify these benefits and recognize them as product development or professional development for its staff. Paradoxically, this, in effect, modifies the process: in other words, Cornell has found it necessary to sweeten the consortium with private goods-style benefits to induce organizations to pay "taxes" (greater than the costs of providing those particular benefits) and sufficient to support the continuing development of this public good with its strong public goods externalities.

Electronic Networks as Public Goods

The services of uncongested electronic networks have the same economic characteristics as does GateD. Once a network is fully enabled and up-and-running, its use by the next user implies costs to society that are essentially zero. The network service is neither depletable (up to the point of congestion) nor excludable (by "definition" of the network). Therefore, the price to the user appropriate for efficient allocation of an uncongested network is also zero.

This is the approach that historically has been used with the Internet. Resources, mainly provided from and by the universities (about 97% of the total costs of the Internet) have been cobbled together in response to the catalyst of some government funding (3% of the total) to enable the world-wide Internet to be offered to users "free."

Economics of Congestion

Congestion on an electronic network means that the network is experiencing "performance degradation." It then takes on economic characteristics that are consistent with a private goods externality: it is depletable, but not excludable. Once congestion occurs, a decision point has been reached. Decision-makers now can choose to transform the private goods externality in either direction as shown in Figure 1. They could transform its characteristics from those of a private goods externality back into those of a public good by acting to overcome its depletability—by expanding its bandwidth, for example. Or they could introduce an institutional constraint of some sort to remove the depletability by making the resource excludable (E): "enough" current users would have to be excluded from the network to remove or obviate the congestion. In effect, this is a process of introducing one characteristic of a private good, "excludability," in exchange for the other, "depletability."

The institutional change required to bring about excludability can be achieved either through pricing or through an administrative limitation on entry, or by introducing a combination of the two: an administrative limitation combined with a "white market" through which those with an administrative permit for entry can trade their permit in exchange for financial considerations, to persons who place a higher value on it.

Transforming the private good externality back into a public good, can be done by expanding the capacity of the constrained resource. For example, this can be done for an electronic network by increasing the bandwidth of the network; for a physical highway, it can be done by increasing the number of its lanes.

Figure 1: Economic Characteristics of the GoodFigure 1: Economic Characteristics of the Good

In deciding whether it is preferable to transform the private good externality into an impeded public good or back into a public good, a number of factors must be considered. Perhaps the strongest negative for adopting the option of transforming the resource into an impeded public good is that the capacity will remain constrained at its pre-existing level; the resource will remain scarce in relation to demand, and society will be saddled with the financial and economic costs of that scarcity. Users of the constrained resource will have to pay for the entry they achieve, while others will experience the cost of being foreclosed from any use of the resource.

The major negative in the choice of transforming the resource back into a public good is that it becomes necessary to somehow aggregate funds sufficient to remove the capacity constraint. Society is back at the fundamental conundrum for such resources, the need for asymmetric pricing: sufficient funds must be raised to provide the resource at the necessary level, while the zero-price "charged" to users generates zero revenue. We have just discussed Cornell's approach to doing this, the consortium. In a section below we generalize this approach and identify the institutional form—an entire "industry"—through which economies worldwide have for centuries responded to such conundrums. It is called "The University." But first, we look at an important illustration of what happens if the choice is made to do nothing.

A Retrospective on ARPANET

The Internet of today could face the same fate as the ARPANET of old. The problem with the ARPANET was that once congestion occurred, neither of the above two choices was made. The choice that was made was to do nothing. The ARPANET was permitted to become increasingly congested. As a packet-switching network, it became less and less useful. It was almost continuously overloaded, with increasing numbers of packets being lost with each ratcheting-up of the congestion. It became virtually impossible to get anything beyond a short message through, and often impossible to know whether or not even a short message had gotten through. The performance of the ARPANET became increasingly degraded (its service was depleted more and more) while access became more and more widespread (it remained non-excludable).

This proved to be worst of all worlds. It is a fate—a tragedy—that could befall the Internet, as it befell the commons of old.

Implications

In a market of very rapid technological change subject to very significant economies of scale plus huge positive externalities (most of which have public goods characteristics), the choice among these options is indeed obvious: the first choice is clearly preferable. By definition, the presence of very significant scale economies for a factor implies increases in capacity at much-less-than-proportional increases in cost. Given the history of the Internet, a creature of universities operating under the University Model (discussed further below), it has been well-demonstrated that unconstrained capacity in electronic networks can be, has been, and is being put to very good use.

The University Model and the Worldwide Knowledge Industry

The University Model operates under an incentive structure appropriate to creation of public goods, such as knowledge and innovation. The shorthand statement of the incentive structure is "publish-or-perish"—give your best idea (or breakthrough) away before someone else comes up with it! This is because universities are "in business" to create knowledge. Knowledge is neither depletable nor excludable. Because one party understands the Internet, for example, does not somehow exclude you from understanding the Internet—in fact, quite the opposite is true: the better others understand it, the more readily they can help you understand it. Knowledge of the Internet is not depletable, neither is it excludable: knowledge enriching one party's understanding does not in any way interfere with the similar enrichment of others.

Internet users with good ideas continue to give them away. Among the innovative results are: the Internet itself; CU-SeeMe, innovative, free software from Cornell that enables desktop video conferencing among multiple sites; and most of the tools used to access the Internet today, including Mosaic and Netscape (still available free, despite the public offering of the stock of its founding organization), among others.

The publish-or-perish incentive structure that functions throughout this worldwide "industry" of the universities has demonstrated for centuries its ability to result in innovative new products and services. These did, and are likely to continue to, "blindside" even our richest and most admired private-sector industrialists (as has been the case with the Internet, as reported by Andy Grove, CEO of Intel).

An insight of even greater significance is that private sector firms operating through markets could not have produced the Internet! The Internet has grown up in response to incentives that fit the University Model, not the private sector market model of patents, trade secrets, high prices, destructive rivalry, etc.

Introducing Excludability to an Uncongested Network

If a price or other form of excludability is imposed on an uncongested network either as a "rent seeking activity" or as a means of "removing depletability" the result is an "impeded public good" (it has been made excludable, but non-depletable), rather than a private goods externality (that is depletable, but not excludable). Remember, also, an externality is an unintended result. In the context of rent seeking and in the situations depicted in Figure 1, pricing or other forms of excludability are intentional acts. The characteristics of the public good and its optimal pricing are violated. The result is an impeded public good, rather than a plain-old public good.

Society pays the cost of this violation in more ways than one. A measurable number of users who would have benefited from access if basic principles had been followed, are excluded from access by the price. Those who achieve access must all pay the explicit price, while network capacity remains unchanged. It should be noted, however, that if the price is a flat rate, this is a less virulent way to violate the characteristics of optimal pricing for public goods than do most "rent-seeking" approaches. The user's marginal cost of each use is unaffected: it is still zero.

There is another possible set of characteristics that might emerge: those of a private good. This could occur if the attempt to remove the depletability was unsuccessful. Then the introduction of the attribute of excludability would have "bought" much less than intended: both depletability and excludability would be present—where the preferable solution would be for neither to be present.

Freeways, Rush-Hours and Other Oxymorons

Perhaps an intuitive example from a related "network"—the physical highway or "freeway"—can help clarify the key parameters of the above discussion. Freeways belie their name at rush-hour when they become congested; at that point, their use is no longer free. The arrival of an additional user imposes real costs on all users. Together, all pay the costs of delay, frustration, pollution, etc., associated with the congestion. This provides both meaning and contradiction to the term "rush hour;" at rush hour, traffic is stalled bumper-to-bumper. This represents the quintessential private goods externality. The resource of high-speed highway travel is woefully depleted; drivers are not excluded; congestion worsens.

Incorporating an institutional change that introduces the other attribute of a private good, excludability, in exchange for removal of congestion, transforms the characteristics of the good from those of a private goods externality into those of an impeded public good: depletability has been removed, excludability has been introduced. The benefit of this exchange is that the usefulness of the resource as originally conceived has been reestablished—performance is no longer degraded along the freeway (or over the electronic network). The original capacity of the resource has been retained. But it remains scarce in relation to unconstrained demand, and with pricing, all those who achieve access to the resource must pay this explicit monetary price for the "transformation."

In the case of a freeway, a system of toll-booths could be incorporated as the means of limiting access (providing excludability) to the freeway. If the toll is a price sufficient to foreclose enough potential users so that the carrying-capacity of the road is fully re-established, then in this idealized condition, the above-stated objectives would have been achieved.

An alternative (also idealized) one-step institutional procedure to achieve excludability would be for potential users to bid for access to the freeway, with access provided to all those whose bids exceeded the price necessary to limit access to the carrying-capacity of the freeway. Through this mechanism, persons valuing access most highly would achieve access. Those not willing to pay the market-clearing fee, by definition, value access less than do those willing to pay that fee or more. Congestion would no longer exist, traffic would flow up to speeds equal to the speed limit, and the resource would be optimally allocated (given the existing distribution of income and wealth).

It is important to emphasize that the costs of providing highway carrying-capacity that is less than the unconstrained demand for such capacity, will persist whether or not congestion is manifest or has been obviated. As noted, with a freeway that is free (i.e., non-excluding) the costs are found in congestion delays, the opportunity costs of those delays, the frustration, the pollution, etc. With the introduction of even the optimal institutional adaptation, the costs of unconstrained demand in relation to existing carrying-capacity are no longer manifest in the physical ways just identified; rather they exist in three other ways.

The most obvious cost is the explicit monetary price paid by users (at peak time) for access to that freeway. All actual users of the freeway pay this market-clearing price. Then there is the additional identifiable, measurable cost: the value placed on use of this highway by those potential users who were excluded from access to the no-longer-free, freeway. The economic value lost is equal to the bid price each was willing to pay; it is the value forgone by those excluded from access by the market price placed on access. These costs are experienced while the capacity of the freeway remains at its pre-existing level.

There still is a measurable consumer surplus for those achieving access, however. It is measurable through the administrative mechanism as the difference between the price each successful bidder was willing to pay and the market-clearing price.

Transaction Costs and Investments

Our model is an idealized one, assuming virtually instantaneous information flows and access authorizations. It abstracts the costs of the technologies required to permit these phenomena to go forward. For example, major resource commitments are required to transform a freeway into a toll-road, a crude mechanism for rationing access to the carrying-capacity of the road. Three foreseeable, but often ignored phenomena accompany this crude approach. First is the cost of creating toll-booths and expanding the number of lanes to provide access to the toll-booths. The second is the requirement that traffic stop at the toll-booth in order to pay the toll, thus slowing throughput for the highway. The third is the congestion that usually builds up in the immediate vicinity of the toll-booths, despite the expansion of the roadway to provide for multiple toll-booths. In other words, this crude mechanism for reducing congestion often causes congestion, lowers the carrying-capacity of the road, while greatly increasing the costs of its construction as a direct result of the toll-booths themselves.

Similar costs accompany attempts to price access to networks. The hardware, software, and administrative paraphernalia can exceed the cost of the original network—and may even be economically or technologically impractical. Decision-makers must bear these facts in mind when making their choices.

Clarification of Model/Summary

We started this discussion by stating that in a complex economy, public goods exist side by side with private goods, and that externalities are pervasive. No "one-size-fits-all" approach to allocation of goods of such complexity makes sense. Efficient allocation of a good must be driven by the characteristics of the good. When the characteristics of two goods are different, then to be efficient, the approaches to their allocation must be different.

We have presented extensive discussion of two Internet "goods," GateDaemon and service on uncongested networks, each of which is non-excludable and non-depletable. By definition, they are "public goods;" and the marginal cost of supplying these goods is zero. Appropriate allocation for public goods, following the basic economic principle of price equal to marginal cost, is a price to the consumer of "zero." In turn, this implies the need for asymmetric pricing: a way must be found to provide resources to the producer of the good sufficient to keep him or her producing the good. Economic taxes (whether from public or private sources) are used to achieve the latter.

If a public good experiences sufficient performance degradation, its characteristics are transformed into those of an economic private goods externality. It has become depletable, but remains non-excludable. There are three possible approaches to resolving the private goods externality. One option is do nothing; the good can continue to degrade until it is no longer usable and is finally shut down. The second option is to reinvest in the good, using the resources (taxes) acquired through asymmetric pricing to fund the reinvestment. The third option is to impose some form of institutional constraint and transform the good into an impeded public good that is now excludable, but no longer depletable. Under the conditions of option three, the user pays a price for access to the good that is equal to the cost to society of removing the depletability (the congestion). That price is arrived at through a market mechanism with a market-clearing price that is symmetrical: what the user pays is what the institutional authority receives. This feature explains why the private sector so readily adopts this economic form (for movie theaters, CATV, commercially provided software, and myriad other activities).

Asymmetric pricing for unimpeded public goods is most effective when the agent that must deal with the asymmetry has the power to impose taxes (as "Ma Bell" did in its monopoly days). If the agent does not have this power, then in part, it must "beg" or it must create an incentive program—consortium benefits—and invest considerable effort to generate the funds, even to the point of creating a good with mixed characteristics of both private and public goods.

The GateD Consortium is a successful implementation of asymmetric pricing through a strategic alliance of organizations that value GateD. More important than proving that (somewhat modified) asymmetric pricing can work, GateD, as it has been implemented, has proven to be invaluable to the growth of the Internet and the Internet economy through its spillover benefits, especially those of enhanced interoperability of the Internet.

Since the decisions on the appropriate allocation mechanism for public goods, or for private goods externalities are at the heart of the debate over the commercialization of the Internet, it is extremely important to note why there is a "best" option in each case: that of asymmetric pricing for the public good; and for the latter, that of returning the characteristics of the good to those of a public good. In each case we have advanced the multiple reasons above. If the growth of the economy—and the growth of the "Internet" business—is a desirable goal, then it is to everyone's advantage to recognize that the appropriate allocation decisions will avoid the negatives and achieve the positives we have identified.

Conclusions

The competitive market is not the only approach to resource allocation currently being employed, even in the U.S. economy. It should not be considered a panacea—especially in the presence of instances in which there is an approach that is obviously superior for all. The Internet economy has blossomed because a market pricing strategy was not imposed on its development.

There are clear guidelines for appropriate resource allocation approaches in the presence of public goods and in the presence of private goods externalities. There are powerful reasons for following them.

The GateD Consortium and Project is a foundation and a model. It is a prime example of where a little cooperation can result in great benefits for all. It is consistent with a current incentive structure—the "University Model"—that has proven its worth, literally over centuries, in relation to economic public goods. It demonstrates that through the consortium mechanism, the university and not-for-profit sectors of our society can implement today rational, efficient, resource allocation approaches on their own for the Internet.

We do not need to wait for, or rely on, government to do this for us. But we do need to exercise some internal leadership. What institution(s) will step up?


Acknowledgments

Thanks go to research assistant, Matthew Wagner, for his insight. The original paper, "Funding an Internet Public Good: Definition and Example," was presented at INET'94 in Prague, Czech Republic and published in Computer Networks and ISDN Systems 27 (1994) 403-409.


Author Information

Martyne M. Hallgren, Cornell Information Technologies, Cornell University, Ithaca, New York 14853 USA, voice: 1.607.254.8324, fax: 1.607.255.8169, martyne@cornell.edu. Alan K. McAdams, Johnson Graduate School of Management, Cornell University, Ithaca, NY 14853 USA, voice: 1.607.255.6443, fax: 1.607.254.4590, akm3@cornell.edu.