A fortnight ago we discussed the effect of infrastructure evolution on the race between traditional telecommunications infrastructures such as ATM and TCP/IP-based infrastructures such as Ethernet. The author concluded that Ethernet held the overall advantage but not in all important attributes. This week’s column will focus on the role of economic forces, including economies of scale and economies of scope, in determining the outcome of the “Great Communications Race.” We again will find these forces tip the scales in favour of data networks, which will probably prevail over circuit-switched networks in most, but not all, market segments." But first, a little background will set the stage for our discussion.
When the US Department of Justice brought action against AT&T that eventually caused the local exchange business to be divested from the long distance business, enormous amounts of time and energy were devoted to trying to prove that there were or were not sufficient economies of scope to justify allowing AT&T to remain intact.
Economies of scope occur when it is less costly to produce two or more products together than separately. Economies of scope provide the glue that keeps most multi-product businesses together; economies of scope explain why there are no canned soup stores or telephone companies that limit service to call forwarding.
In 1984, there were not, according to the US Department of Justice, sufficient economies of scope to justify keeping local and long distance exchange services bound together in a single company. Neither academic research nor experience revealed strong economies of scope.
Economies of scale are similar to economies of scope except that the cost advantage is the consequence of producing more of the same product. Large-scale assembly processes can lower unit costs because economies of scale kick in.
The shape of most industries in market economies results directly from a handful of basic economic factors – two of the most important are economies of scale and scope.
Other important economic forces that are affecting the communications race are the relative costs of alternative technologies, the labour markets that support alternative technologies, and the opportunities to price in ways that capture the value of the information being transported. The last point is important enough to devote next week’s column entirely to the subject of pricing.
Each of these factors handicap one or another network infrastructure. Such matters can greatly influence the race.
Economies of Scale and Scope in Communications
Both circuit switched networks and TCP/IP-based data networks exhibit economies of scale and scope. But data networks will pack the bigger economic punch in economies of both scale and scope.
As we work through the discussion of these economies, we will need to differentiate between wired and wireless networks because each has different attributes with respect to both economies of scale and scope. Since both wired and wireless networks can operate using TCP/IP and circuit switched technology, we must evaluate each separately.
Scale – The communications physical infrastructure, fibre-optic cables, radio spectrum, and other media, all have limited capacities governed by the medium itself and the associated transmitters, receivers and processors. Economies of scale accrue more easily to the technologies that can continuously fill the available capacity more efficiently and more cost effectively. Data networks are more adept at seeking out and using available capacity than are circuit switched networks. The reasons are complex and technical but they boil down to a couple of characteristics of data communications: 1) TCP/IP-based networks can crowd more packets of data into more pathways than can circuit switched networks; and 2) TCP/IP-based networks can handle a greater variety of packet sizes and shapes than can circuit switched networks using relatively less data processing. This all results in data networks having smaller variable costs of expansion (and more rapidly declining marginal costs) than does the PSTN over most of the usual ranges of bandwidth. In plain terms, more people can share the same communications pipes in, and it is easier to add additional data traffic to, the TCP/IP-based network. This will be true in both wired and wireless networks.
Advantage: TCP/IP networks.
Wireless communications including satellite and terrestrial have limited economies of scale primarily due to spectrum and transmission constraints. 1 Squeezing more megabits of bandwidth from a slice of the electromagnetic spectrum in free space is relatively expensive. 2 In terrestrial wireless systems, base stations must be replicated as cells are split and in satellite systems, new satellites must be sent to spin around the earth as transponder capacities are reached.
Wired data networks have the advantage over wireless networks when comparing scale economies. 3
Scale advantage: wired TCP/IP networks.
Scope – Economies of scope occur when additional applications (products, etc.) can be added to others at relatively low cost. In communications, a variety of services and applications (teleconferencing, videoconferencing, streaming broadcast video, large data transfers, etc.) require widely varying bandwidth. Circuit switched data networks, originally designed to interconnect voice grade telephone networks (such as ATM) operate with small packets. Assembling small packets into large cohesive files (as is required for streaming video, for example) requires intensive data processing. This is likely to increase the cost of adding the very applications that offer the greatest opportunity for future growth, and reduces the respective economies of scope. 4
Advantage: TCP/IP-based networks.
Wireless communications including satellite have economies of scope that differ in a fundamental way from the “wired” networks. Wireless networks add a dimension to scope that wired networks cannot: services that have value due to mobility enhance the scope of services available in the wireless world. Navigation systems that use satellite for example can offer a scope of services that cannot be matched by terrestrial wireless or wired networks. Ubiquitous Internet access allows the development of products that are significantly differentiated from Internet access tethered to a wire. Information products that have value at a particular location can be delivered in a mobile network with greater variety than in a wired network. Thus while scale economies work against wireless networks, scope economies work in their favour by adding mobility to the scope portfolio of possibilities.
Scope Advantage: wireless TCP/IP-based networks.
Greater economies of scale or scope in one network over another do not necessarily imply an overall cost advantage (a lower incremental or marginal cost does not guarantee a lower total cost); however, these economies can determine the winner in a competitive market. If, for example, a broadband ISP or Cable TV company can cover its basic infrastructure costs with entertainment programming or Internet access, the small incremental cost of adding voice communications may allow it to undercut a telephone company not having such scope advantages. Likewise, a company having a large scale of business can take on extra business at a lower cost than can a smaller company. It is an incremental, rather than total, cost advantage that determines whether a company succeeds in individual markets; total cost determines whether the company will succeed at all.
If history provides a lesson from the harnessing of energy during the industrial revolution, the information revolution will offer services that evolve from predominantly utility to predominantly entertainment. 5 Indeed this is already the case. Short Message Service (SMS) is dominated by games. Internet services pass more bits of data in the form of entertainment (albeit mostly of the seedy sort) than in the form of more practical information. HDTV, DVD and other video media are prevalent today. What this suggests is that, over time, digital communications will be dominated by recreation and entertainment rather than by a need-to-know. Information that is needed for more mundane and practical uses, and voice conversations that require little bandwidth, will be economical by-products of broadband communications. Thus, the voice communications market segment, which generates most of the communications revenue for the PSTN today, may be the canned soup store of tomorrow. This, of course, is the greatest fear of incumbent telephone companies particularly in countries that depend heavily on government-owned utilities for revenue.
Data networks that can cover most or all of their costs with revenues from entertainment and other high value broadband services, and can offer mobile and fixed network voice services at small incremental cost likely will win the race.
Advantage: Wired TCP/IP networks for high bandwidth applications; wireless TCP/IP networks for mobility applications.
Relative total costs
Incremental costs win sprints but not marathons. Marathons are won by having the lowest total cost for the largest scope of, and highest value, services.
PriMetrica, Inc., in performing economic analysis and business cases for new communication networks, has calculated that the total cost of most new networks is significantly lower for Ethernet over MPLS (EoMPLS) and similar IP-based networks than for ATM or other technologies typically used in the PSTN. This may change as the cost and capabilities of equipment changes and as network designs evolve, but the trend is strongly in favor of TCP/IP-based networks. Both wireless and wired networks can interoperate using these IP technologies. 6
The total cost advantage goes to TCP/IP-based networks. Today’s PSTN has a potential cost advantage only when the cost of the existing infrastructure is considered sunk. As existing infrastructure is displaced by its progeny, networks designed to take full advantage of TCP/IP likely will prevail.
Long run advantage: TCP/IP-based networks.
Concentrated versus Dispersed Network Engineering Skills
Circuit switched networks have been designed, built and maintained by telecommunications engineers, most of whom were trained by and worked for the local or long distance telephone companies. Computer engineers and persons skilled in designing and maintaining data networks are everywhere computers are found. Data Communications equipment providers sponsor training and testing programs that are widely available. It is not difficult, for example, to locate local labour familiar with Ethernet systems; it is more difficult to find local talent experienced in maintaining local telephone switches.
Looked at another way, there are fewer than 10,000 telephone companies in the world comprising the PSTN. There are over 100,000 data networks comprising the Internet. The engineering expertise is much more dispersed in the latter than in the former. An important advantage of TCP/IP-based networks is the limited dependence on carriers’ maintenance personnel.
The disbursement of expertise matters in a couple of important ways. Increasingly, businesses and some households are deriving benefits from integrating their computers and data networks with telecommunications. This requires integrating both hardware and software on or near one’s premises. Local expertise can more effectively work with Ethernet-oriented technology than with ATM-oriented technology, for example, to perform customized integration. Both business and residential customers are demanding increasing control over the features and functions of their telecommunications services. Control and customization requires a certain amount of expertise. That expertise is more readily available for simple data network technologies than for the technologies that make up the PSTN. As the Internet evolves, IP expertise will grow commensurately and likely will give IP-based networks the advantage in all forms of communications.
The economics of networks clearly favours the dominance of TCP/IP networks (and their offspring) over the more traditional PSTN technologies. It is highly likely that both wired and wireless TCP/IP-based networks will co-exist – the latter taking greatest advantage of mobility, the former leveraging its exceptional capabilities in managing high bandwidths.
Next week we will take up the question of pricing. Some new twists in the race are in store. 1. A subtle but important technical point pertains to the “variable” cost of wireless networks: as large changes in traffic are accommodated in wireless networks, the variable costs would include the costs of adding large lumps of capacity such as launching additional satellites. See the economics literature on “Capacity Cost” as it is defined in the U.S. telecommunications industry, for example. The author can provide references on request. 2Fiber optic cable uses the electromagnetic spectrum to separate communications channels within a strand of fiber however the spectrum constraint is not as critical since the spectrum used in one fiber strand does not interfere with that in another and additional fiber strands are available at a small incremental cost. 3 There is another implication of scale economies that is perhaps more important than a network’s efficiency of transporting, storing and processing binary data. Voice services and other narrowband applications don’t benefit much from the scale economies. It is the broadband services such as video and high quality audio that benefit most and this has greater impact on scope economies than on scale economies. Demand for bandwidth is the primary cause of higher variable costs in most communications infrastructures. As more broadband services dominate the use of communications infrastructure, voice traffic becomes a less significant contributor to bandwidth demand. Since circuit switched networks evolved optimized for voice applications, they are at an inherent disadvantage in the race of the future. 4 Frame Relay, unlike ATM, uses variable sized packets and in principle could circumvent this disadvantage. However, for technical and economic reasons that are beyond the scope of this paper, Frame Relay is not likely to be used in an integrated global network because of difficulties extending virtual circuits from one carrier’s network to another carrier’s network. This severely limits economies of scale and scope. The early benefits of the transition from animal power to using other energy sources such as electric power and fossil fuels saved time by working for us. They later drove devices and machines that entertained us. 5 The early benefits of the transition from animal power to using other energy sources such as electric power and fossil fuels saved time by working for us. They later drove devices and machines that entertained us. 6IP operates at layer 3 in the OSI stack. EoMPLS is essentially a layer 2 network. We use the term “IP network” more broadly than the strict OSI designation.