EPUBs are an experimental feature, and may not work in all readers.
The Economics of Layered NetworksSkip other details (including permanent urls, DOI, citation information)
This work is protected by copyright and may be linked to without seeking permission. Permission must be received for subsequent distribution in print or electronically. Please contact firstname.lastname@example.org for more information. :
For more information, read Michigan Publishing's access and usage policy.
Presented at MIT Workshop on Internet Economics March 1995
1. Problem Statement
The creation of a National Information Infrastructure (NII) will require large investments in network facilities and services, computing hardware and software, information appliances, training, and other areas. The investment required to provide all businesses and households with broadband access to the NII is estimated to be in the hundreds of billions of dollars; large investments in the other components of the NII will also be required. In The National Information Infrastructure: Agenda for Action, the Clinton Administration states: "The private sector will lead the deployment of the NII."
What architectural and economic framework should guide the private sector's investment and deployment decisions? Is the Open Data Network (ODN) described by the NRENAISSANCE Committee  an appropriate economic framework for the communications infrastructure that will support NII applications? A key component of the ODN is the unbundled bearer service, defined to be "an abstract bit-level transport service" available at different qualities of service appropriate for the range of NII applications. The Committee states that "The bearer services are not part of the ODN unless they can be priced separately from the higher level services...". (page 52). The rationale for this requirement is that it "is in the interest of free entry at various levels." (page 52).
What effect will the unbundled bearer service proposed by the NRENAISSANCE Committee have on the investment incentives of network providers in the private sector? Will carriers that invest in long-lived assets be given a fair opportunity to recover their costs? This paper provides a preliminary discussion of the economics of an unbundled bearer service.
Convergence and Emerging Competition
Technological advances are rapidly blurring traditional industry boundaries and enabling competition between firms that did not previously compete with one another. For example, cable TV providers in the U.K. (some of which are now partly owned by U.S. telecommunications firms) have been allowed to offer telephone service to their subscribers since 1981, and currently serve more than 500,000 homes.  Numerous telephone companies in the U.S. are currently testing the delivery of video services to households over their networks. In addition, new firms have recently entered markets that were not subject to major inroads in the past. Competitive Access Providers (CAPs) have begun to serve business customers in the central business districts of many large cities in competition with Local Exchange Carriers (LECs), and Direct Broadcast Satellite services have begun to compete with cable providers. In sum, new entrants are using new technologies to compete with incumbents, and incumbents in previously separate industries are beginning to compete with one another.
Theoretical Approaches to the Economics of Pricing under Differing Market Structures.
In a pure monopoly, a variety of price structures may be consistent with cost recovery, and the firm (or regulators) may be able to select price structures that promote political or social goals such as universal service or unbundling of raw transport. In a competitive market, this flexibility may not exist. Under perfect competition (which assumes no barriers to entry and many small firms), the price per unit will be equal to the cost per unit (where costs are defined to include the opportunity costs of all resources, including capital, that are used in production). There is no pricing flexibility. When neither of these pure market forms exist, economic theory does not provide any general conclusions regarding equilibrium price structures or industry boundaries. While substantial progress has been made in developing Game Theory and its application to oligopoly,  no completely general results on pricing are available. This is particularly true in the dynamic context where interdependencies between current and future decisions are explicitly considered.
Some theoretical work in this area is summarized in Shapiro.  An important result in game theory asserts that no general rules can be developed. "The best known result about repeated games is the well-known "folk theorem." This theorem asserts that if the game is repeated infinitely often and players are sufficiently patient, then "virtually anything" is an equilibrium outcome.  Modeling based on the specific features of the telecommunications industry may therefore be a more promising research strategy.
The economic analysis of competition among Network Service Providers (NSPs) is further complicated by the presence of externalities and excess capacity. Call externalities arise because every communication involves at least two parties: the originator(s) and the receiver(s). Benefits (possibly negative) are obtained by all participants in a call, but usually only one of the participants is billed for the call. A decision by one person to call another can generate an uncompensated benefit for the called party, creating a call externality. Network externalities arise because the private benefit to any one individual of joining a network, as measured by the value he places on communicating with others, is not equal to the social benefits of his joining the network, which would include the benefits to all other subscribers of communicating with him. Again, the subscription decision creates benefits that are not compensated through the market mechanism. It has been argued that the prices chosen by competitive markets are not economically efficient (in the sense of maximizing aggregate consumer and producer benefits) when externalities are present. 
It has also been argued that "Industries with network externalities exhibit positive critical mass - i.e., networks of small sizes are not observed at any price."  The consequent need to build large networks, together with the high cost of network construction (estimated by some to be $13,000 - $18,000 per mile for cable systems),  implies the need for large investments in long-lived facilities. The major cost of constructing fiber optic links is in the trenching and labor cost of installation. The cost of the fiber is a relatively small proportion of the total cost of construction and installation. It is therefore common practice to install "excess" fiber. According to the FCC,  between 40% and 50% of the fiber installed by the typical interexchange carriers is "dark;" the lasers and electronics required for transmission are not in place. The comparable number for the Major Local Operating Companies is between 50% and 80%. The presence of excess capacity in one important input is a further complicating factor affecting equilibrium prices and industry structure.
To summarize: a full economic model of the networking infrastructure that supports the NII would need to account for at least the following features:
- oligopolistic competition among a few large companies that invest in the underlying physical communications infrastructure
- network and call externalities at the virtual network level
- large sunk costs and excess capacity in underlying transmission links.
An analysis of the optimal industry structure is well beyond the scope of this paper and is a promising area for future research. This paper focuses on implications of two of these issues (oligopolistic competition and sunk cost with excess capacity) for industry structure and the unbundled bearer service. The Internet is used as a case study to illustrate trends, and provides a factual background for future analyses. This paper provides a description of the current Internet industry structure, current Internet approaches to bundling/unbundling and reselling. some recent examples of the difficulties raised by resale in other communications markets, and the increasing use of long term contracts between customers and their service providers. The implications for the unbundled bearer service are drawn.
The Internet is a virtual network that is built on top of facilities and services provided by telecommunications carriers. Until recently, Internet Service Providers (ISPs) located routers at their network nodes, and interconnected these nodes (redundantly) with point-to-point private lines leased from telecommunications companies. More recently, some ISPs have been moving from a private line infrastructure to fast packet services such as Frame Relay, Switched Multimegabit Data Service (SMDS) and Asynchronous Transfer Mode (ATM) service. Specifically, among the providers with national backbones:
- PSI runs its IP services over its Frame Relay network, which is run over its ATM network, which in turn is run over point-to-point circuits leased from five carriers.
- AlterNet runs part of its IP network over an ATM backbone leased from MFS and Wiltel.
- ANS' backbone consists of DS3 links leased from MCI.
- SprintLink's backbone consists of its own DS3 facilities.
- CERFnet, a regional network based in San Diego, uses SMDS service obtained from Pacific Bell to connect its backbone nodes together. 
These examples reveal a variety of technical approaches to the provision of IP transport. They also show different degrees of vertical integration, with Sprint the most integrated and AlterNet the least integrated ISP in the group listed above. The variety of organizational forms in use raises the following question: can ISPs with varying degrees of integration coexist in an industry equilibrium, or are there definite cost advantages that will lead to only one kind of firm surviving in equilibrium? The answer to this question hinges on the relative cost structures of integrated and unintegrated firms. The costs of integrated firms depend on the costs of producing the underlying transport fabric on which IP transport rides. The cost structures of unintegrated firms are determined in large part by the prices they pay for transport services (such as ATM and DS3 services) obtained from telecommunications carriers. These prices, in turn, are determined by market forces. More generally, the layered structure of data communications services leads to a recursive relationship in which the cost structure of services provided in any layer is determined by prices charged by providers one layer below. In this layered structure, a logical starting point for analysis is the lowest layer: the point-to-point links on which a variety of fast packet services ride.
Competition at the Bottom Layer
For illustrative purposes, consider a common set of services underlying the Internet today. At the very bottom of the hierarchy, physical resources are used to construct the links and switches or multiplexers that create point-to-point channels. In the emerging digital environment, Time Division Multiplexing (TDM) in the digital telephone system creates the channels out of very long-lived inputs (including fiber optic cables). The sunk costs are substantial.
There are at least four network service providers with national fiber optic networks that serve all major city-pairs.  Each of these providers has invested in the fiber and electronics required to deliver point-to-point channel services, and, as was stated earlier, each has substantial excess capacity. The cost structure of providing channels include: high sunk costs of construction, the costs of lasers and electronics needed to light the fiber, costs of switching, high costs of customer acquisition (marketing and sales), costs associated with turning service on and off (provisioning, credit checks, setting up a billing account, etc.), costs of maintaining and monitoring the network to assure that customer expectations for service are met, costs of terminating customers, and general administrative costs. The incremental cost of carrying traffic is zero, as long as there is excess capacity.
If all owners of national fiber optic facilities produced an undifferentiated product (point to-point channels) and competed solely on price, economic theory predicts that they would soon go out of business: "With equally efficient firms, constant marginal costs, and homogeneous products, the only Nash equilibrium in prices, i.e., Bertrand equilibrium, is for each firm to price at marginal cost."  If this theory is correct,  firms could recover the one-time costs of service activation and de-activation through a non-recurring service charge, and recover ongoing customer support costs by billing for assistance, but they would not be able to recover their sunk cost. Industry leaders seem to be aware of this possibility. Recently, John Malone, CEO of TCI, stated: " ... we'll end up with a much lower marginal cost structure and that will allow us to underprice our competitors." 
The history of leased line prices in recent years does reveal a strong downward trend in prices. According to Business Week,  private line prices have fallen by 80% between 1989 and 1994, and this is consistent with Bertrand competition. During the same period there has been a dramatic increase in the use of term and volume discounts. AT & T offers customers a standard month-to-month tariff for T1 service and charges a non-recurring fee, a fixed monthly fee, and a monthly rate per mile. Customers who are willing to sign a 5 year contract and commit to spending $1 million per month are offered a discount of 57% off the standard month to month rates. Smaller discounts apply to customers who choose shorter terms and lower commitment volumes: a one year term commitment to spend $2000 per month obtains a discount of 18%. The overall trend towards lower prices masks a more complex reality. ÒThere [sic] are two types of tariffs: 'front of the book' rates, which are paid by smaller and uninformed large customers, and 'back of the book' rates, which are offered to the customers who are ready to defect to another carrier and to customers who know enough to ask for them. The 'front of the book' rates continue their relentless 5 to 7 percent annual increases."  In 1994, AT & T filed over 1200 special contracts, and MCI filed over 400. 
There are some theoretical approaches that address the issues discussed above. Williamson's discussion of nonstandard commercial contracting  as a means for sharing risk between the producer and consumers is relevant to the term commitments described above. In addition to risk reduction, long term contracts reduce customer churn, which often ranges from 20% to 50% per year in competitive telecommunications markets.  As service activation and termination costs can be high, reduction of churn can be an effective cost saving measure.
There appears to be an empirical trend toward term/volume commitments that encourage consumers of private line services to establish an exclusive, long term relationship with a single carrier. There is little published information on long distance fast packet prices. According to one source, none of the long distance carriers or enhanced service providers (e.g. CompuServe) tariff their Frame Relay offerings.  Some intraLATA tariffs filed by Local Exchange Carriers do offer term and volume (per PVC) discounts, and the economic forces that give rise to term/volume commitments for private lines have probably resulted in term/volume commitments for long distance fast packet services as well.
Competition among Internet Service Providers.
The effect of term/volume commitments in private lines and fast packet services affects the cost structures of ISPs who do not own their own transport infrastructure. It may be expected that large ISPs who lease their transport infrastructures will sign multi-year contracts, possibly on an exclusive basis, with a single carrier. These providers will then have sunk costs, as they will have minimum payments due for a fixed period to their carriers. Competition at this level will then be similar to competition at the lower level, and we may expect to see term/volume contracts emerge in the Internet. A quick survey of Internet sites shows this to be the case. For example, in January 1995, AlterNet offered customers with a T1 connection a 10% discount if they committed to a two year term. Global Connect, an ISP in Virginia, offers customers an annual rate that is 10 times the monthly rate, amounting to a 17% discount for a one year commitment. There are many other examples of this sort.
The Internet is beginning to resemble the private line market in one other aspect: prices are increasingly being viewed as proprietary. ISPs that used to post prices on their ftp or web servers now ask potential customers to call for quotes. Presumably, prices are determined after negotiation. This development mirrors the practice of long distance carriers to use special contracts that are not offered on an open basis at the "front of the book," but are hidden at the back.
Economics of Resale
Kellogg, Thorne and Huber  describe the history of the FCC's decision on resale and shared use. Noam  analyzed the impact of competition between common carriers and contract carriers (including systems integrators and resellers), and concluded that common carriage cannot survive the competitive struggle. Recent events lend some credence to this view. According to one recent study, "resold long distance services will constitute an increasing portion of the total switched services revenue in coming years, growing at a compound annual growth rate of 31% from 1993 to 1995.... The number is expected to rise to $11.6 billion, or 19.2% of the estimated total switched services market in 1995."  The growth of resale in the cellular market suggests that there are equally attractive resale opportunities in this market.  In the Internet, some ISPs charge resellers a higher price than they charge their own customers.  Other ISPs, such as SprintLink, make no distinction between resellers and end users. Facilities-based carriers have had a rocky relationship with resellers, and courts have often been resorted to by both carriers and resellers. 
The pricing model that is emerging appears to resemble rental arrangements in the real estate market. In the New York area, low quality hotel rooms are available for about $20 per hour. Far better hotel rooms are available for $200 per day (which is a large discount off 24 times $20). Roomy apartments are available for monthly rentals at much less than 30 days times $200/day. And $6000 per month can be used to buy luxury apartments with a 30 year mortgage. Term commitments are rewarded in the real estate market, where sunk costs and excess capacity are (currently) quite common. The telecommunications industry appears to be moving in the same direction. Contracts are not limited to five year terms; MFS and SNET recently signed a 20 year contract under which MFS will lease fiber from SNET,  and Bell Atlantic has a 25 year contract with the Pentagon.  The term structure of contracts is an important area for empirical and theoretical research.
Implications for Unbundled Bearer Services
Unbundled bearer services have much in common with common carriage: both approaches facilitate greater competition at higher layers (in content, with common carriage, and in enhanced services of all types with the bearer service). The dilemma facing policy makers is that, if Noam is right, competition in an undifferentiated commodity at the lower level may not be feasible. In his words: "The long term result might be a gradual disinvestment in networks, the reestablishment of monopoly, or price cartels and oligopolistic pricing."  Thus policies promoting competition in the provision of unbundled bearer services among owners of physical networks may ultimately fail. In the Internet, for example, some service providers charge resellers a higher price than they charge end users. The market may be moving towards contract carriage based on term/volume commitments and increasing efforts at differentiation, and away from the ideal of an unbundled bearer service. Should unbundled bearer services be aligned with this trend by being defined as a spectrum of term/volume contracts? The competitive mode that is emerging is quite complex, and the effects of unbundling in this environment are hard to predict.
This paper does not suggest specific architectures or policies for the emerging NII. It identifies some difficult economic problems that may need to be addressed. These are the familiar ones related to resale and interconnection, with the added complication of competition among multiple owners of geographically co-extensive physical networks. This paper has provided references to recent developments in telecommunications markets, and identified strands in the economics literature that are relevant to the central issues raised by the bearer service.
There is an urgent need for a clearer economic analysis of these issues, and it is critical that the analysis pay close attention to the realities of competition and evolving competitive strategy. Three specific areas appear particularly promising:
- Empirical analysis of evolving price structures that quantifies the movement from pricing by the minute (the original Message Toll Service) to pricing by the decade (contract tariffs).
- Game theoretic models of competition in long term contracts with sunk costs.
- Experimental approaches to network economics. 
Jiong Gong (email@example.com) is a Member, Technical Staff and Padmanabhan Srinagesh (firstname.lastname@example.org) is a Senior Scientist with NII Initiatives, Bell Communications Research Inc.
19. See FCC Docket 93-197, Report and Order, 1.12.95, page 8 for statistics on AT & T's churn in the business market and Cable Franchise Growth in the U.K by David Ko and Michael Gell, mimeo, University of Oxford, UK for churn in the UK cable market.
25. For example, Alternet's policy in January 1995 was: "Because wholesale customers use more of our backbone facilities and because they also place greater demands on our staff, we charge more for our wholesale services".
26. For example, "Oregon Jury Decides against AT & T in Reseller Case". Telecommunications Reports, July 4, 1994, p 34, and "AT & T sues Reseller for Unauthorized Trademark Use", Telecommunications Reports, November 7, 1994, p 26.