The Great Communications Race: Future Pricing Plans

23 May 2003

In the prior article of this series, I examined two important economic forces shaping communications networks of the future: 1) economies of scale and scope and 2) the availability of labor in concentrated as compared to dispersed markets. In this article, I explore how telecommunications prices have evolved as wireless communications grew in importance and how they will evolve again as the Internet develops.

Because pricing models in communications and IT are changing in fundamental ways, I will break the discussion of pricing into two parts: pricing of communications transport through networks, and pricing of the content. This week, I will concentrate on pricing of communications transport through networks. In the next article, I will address the evolution and pricing of content and related products.

Now, a bit of background about price discrimination will be helpful. Price discrimination often is viewed as inherently unfair. After all, our tendency is to believe all persons should be treated equally until a strong case is made to the contrary. Indeed, price discrimination based on the identity of a person or other entity is generally neither acceptable nor economically efficient, but price discrimination that treats entities in similar circumstances similarly can be both efficient and fair.

But price discrimination can be more beneficial than harmful when used judiciously. Monday night at the movies can be priced lower than other movie showings, hotels and rental cars are priced by the day of the week and season of the year, seniors and children are given special discounts, and staying over a Saturday night on vacation can reduce the price of an airline ticket dramatically. These are all examples of beneficial price discrimination.

In general price discrimination increases the profits of the seller as compared to a single price tactic. It also allows a greater variety of products and services to evolve in competitive markets. Price discrimination is a prevalent and generally beneficial practice in market economies.

Today in the Public Switched Telephone Network (the PSTN), voice telephone call prices discriminate effectively among calls having very different values. International-call rates using the PSTN can be as high as several dollars (U.S.) per minute or as low as a few cents per minute depending on the origination and destination of the call. Calls made from home are priced below calls from a business. These prices are not as closely linked to cost as to the value of the call to the consumer. In the telephone service market price discrimination is possible because the service provider “knows” the location of the terminating and originating ends of the voice circuit that is necessary to establish the connection. This information is captured in the billing or pricing software that is used to both set up and charge for the call. The ability to discriminate among calls is inherent in the circuit-switched technology that is used in the PSTN. That is, the technology easily and accurately provides the calls’ origins and destinations, the duration of calls, and the time of day that calls are made. Billing systems capture this information and price according to differential values based on these time and space characteristics of the call.

Wireless telephone networks changed this pricing structure for several reasons. Long-run variable costs are significant in wireless networks.(1) Accordingly, wireless prices for bundles of minutes increase as the size of the bundle grows, but the price per bundled minute declines, thus providing a volume discount of sorts. This discounted price pattern is better aligned with cost causation than is the constant price per minute common to most pricing plans on the traditional cabled PSTN.(2) In the largest bundles of wireless minutes, the price per minute is competitive with fixed network prices. But wireless networks have two advantages over the PSTN: mobility, and handsets that were designed from the outset to display visual information.(3) Thus, mobile wireless calls generally have a higher value per minute than do calls on a fixed network despite having a lower overall quality of voice communications. In addition, features using displays are developing at a more rapid rate in wireless networks than in the PSTN.

In wireless networks, prices discriminate according to several characteristics of a call. Text messages, games, and other short message services are often priced differently than voice calls. Persons using more minutes pay less per minute than do persons using less minutes. Special services that are location-dependent or require higher bandwidth are priced separately from ordinary conversation minutes. Price discrimination is more prevalent in mobile networks than in the PSTN. Because of mobility and the heavy emphasis on visual displays, it is likely that opportunities for price discrimination will grow more rapidly in wireless networks as well.

If the price war were limited to wireless networks and the PSTN (and it is not as we shall discuss shortly), events may have unfolded something like this. Packages of wireless minutes arewould be structured so that the incremental cost of a call within the selected block of minutes (both long-distance and local) would be zero. Calls on the PSTN would need to be comparably priced at the margin. Similarly, features would be bundled with packages of minutes at no incremental price. Many subscribers would buy both wireless service and fixed-line service but other subscribers would buy one or the other. Among those who would buy only one service, wireless would be preferred because of the convenience of mobility.(4) Niche markets would be served with specialized packages that appeal to people who use phones for mixed personal and business purposes, messaging services, games, and so forth. Price discrimination would operate most effectively in the niche markets. Basic voice plans would migrate towards a market-based “commodity” price that discounts average (per-minute) based on the size of the minute package. But this pricing scenario may never be fully realized because of the advent of the Internet Protocol (IP).

IP-based networks presently cannot differentiate among a packet of bits carrying a voice call, a data transmission, or a streaming video.(5) The content, and hence the real value of the communications, hides inside a packet that discloses only its header information while in transit.

At first glance, it seems (incorrectly as I will soon explain) that price discrimination might be diminished to the detriment of revenues in an IP network. Without some ability to discriminate among packets, a single price per packet would prevail. And a constant price per bit (6) could not be structured to reflect both the value of a movie and the value of a voice call. Videos could not be cost-effectively transmitted at voice call rates. Similarly, large data streams could not be priced at voice call rates in a competitive market. The relative values of packets of bits in each situation are vastly different. Simply put: price discrimination cannot be effectively applied to packets of data in a data communications network without some method of segregating packets based on a proxy for the value of the content. VoIP (Voice over IP) is just another assembly of data packets mixed in with packets of e-mail, book orders and streaming video newscasts running through the same digit-pipe unless and until the contents can be inferred and their value captured in a billing system.

But price discrimination is very potent in IP-based networks. To see how price discrimination can be effectively applied in IP networks, we look to the information most available to billing systems: the packet header.

The next generation of the Internet Protocols (e.g., IPv6 or IPng)(7) will cause routers and data switches to manage packets differently based on information contained in the “header” of each packet. While the content of the packet could be examined cost-effectively using the technologies on the foreseeable horizon, the headers could and will be examined to determine a variety of information necessary for routing and other packet management functions. Price discrimination could be implemented based on the generic value of different flavors of packets. Saturday night stays may not apply to packets but staying a few extra milliseconds at a router in the middle of Internet might serve the same purpose.

The header information that will be easily captured or accessed by IP-oriented billing systems falls into three general categories: 1) time sensitive packets (e.g., voice conversations, real-time video streams, and certain scientific monitoring instruments), 2) packets requiring extraordinary security (e.g., confidential data, financial transactions, and sensitive corporate communications), and 3) location-sensitive packets (e.g., locality maps, location-sensitive communications, and emergency services).(8)

The ability to discriminate among packets using header data together with the ability to substitute packets with different characteristics will allow new forms of price discrimination to be effectively implemented. For example, depending on the relative cost of data storage versus prioritized transmission, movies may be viewed in real time using streaming video stored in terminal devices for later viewing. Voice conversations may be purchased in varying qualities or substituted by e-mail. Data packets may be streamed in an unbroken sequence overnight or interjected in quiet periods over the Internet at all times of day. The possibilities are extensive.

By contrast the traditional PSTN would open circuits, physical or virtual in form, upon request. These circuits may have different characteristics and the pricing likely would be based on the characteristics of the circuits rather than the characteristics of co-mingled packets. Even if the PSTN technology were capable of discriminating among packets, the circuit-oriented network is not optimized for packet-level management and associated pricing.

So, the future pricing of telecommunications might go something like this. The content of packets will be inferred by their header data. The value of the content will be captured through price discrimination based on the packets’ characteristics made available by the predominant Internet Protocol. As is the case in the wireless competition, the PSTN will be forced to move towards the market prices so established. The cost structure of the PSTN will be out of line with the market prices and the PSTN will continue to struggle for survival by taking advantage of the (temporarily) favorable low variable cost of using sunk assets. But this situation is not viable in the long run.(9)

A typical future IP header will contain a couple of handfuls of basic fields that can be used in pricing. The origination and destination addresses will allow devices and, with the help of external databases, locations to be uniquely identified. The capacity of the address portion of the headers could allow every communications device to be uniquely identified. In principle, every customer, every device, and every location of a device could have a customized pricing plan. Quality of service fields could be used to charge higher prices for higher priority packets – packets that must be delivered in fewer nanoseconds; packets that must have special security or routing handling, etc.; or packets that require special handling at the edge of the network or in transit through the network.

Thus, four key characteristics will be prominent in communications transport pricing: the priority of the communications (voice conversations in real time will require a reasonably high priority, for example), the required bandwidth (movies will require a lot), the security requirements of the data stream (bank transactions could be more expensive than an ordinary e-mail), and the required accuracy of the data transfer (entertainment may be less expensive per unit of data than transmitting accounting records).

Today pricing plans based on time and/or distance can appear complicated but the potential complexity of transport prices is much greater in IP-based networks of the future. The rewards of price discrimination surely will encourage creative uses of header information to develop new pricing plans. Prices are headed for the headers.

In the next article, I will address the likely future of pricing content and all the devices that will make content useful.


1) Long run variable costs include any necessary costs of splitting cell sites, adding antennas and other infrastructure, expanding back haul facilities, and other costs of relieving congestion in wireless networks as traffic grows.

2) Indeed, there is even more reason for prices per minute to decline with volume in a cabled network than in a wireless network; a higher proportion of costs are fixed and therefore average cost declines dramatically in a cabled network.

3) There is no technical reason why a handset connected to the PSTN could not have the same visual display. Nevertheless, handsets did not and therefore features provided over the PSTN did not develop to use this characteristic. Witness, for example, the need to dial *”something” to activate a feature rather than select from a menu on a display.

4) A recent study by PriMetrica, Inc. revealed that about half of all households would drop their fixed-network service in favor of wireless at prices that are comparable to the most recent pricing plans. Leap Wireless just released a study that found 37% of their subscribers do not have a fixed-line service in their home today.

5) Traffic can be differentiated by TCP/IP port used – for example, outbound email normally uses TCP port 25, FTP file transfers use port 21, POP3 mail retrieval uses 110, DNS lookups use TCP and UDP port 53. There are a number of “well-known” ports used by TCP/IP but an application can easily alter the port used if need in order to avoid detection. For example, HTTP (web) uses port 80 by default, but security-sensitive web-based applications that want an extra level of protection will require you to put a “:81” after the URL to specify that you want your browser to connect on port 81 instead. In this way, there can be generalized differentiation of the way in which something’s transmitted, which is often correlated to what is being transmitted. However, if this were applied to any sort of billing scheme, people would work around this in a heartbeat to reduce charges – i.e. a website would automatically redirect web visitors over to their non-standard “port xx” web site that may have lower fees.

6) The same conclusion applies to a constant price per minute for a fixed bit rate or any other equivalent constant price.

7) IPv6 is the “Internet Protocol version 6” and IPng is the” Internet Protocol next generation”.

8) The last of these, location, is not directly evident in the header packet header. The location of the origination and destination of the communications must be captured with reference to a database that links an IP address to a location or its surrogate.

9) In a future article, I will address the potential financial crisis that could destroy the traditional PSTN once and for all.

Dr. Richard Emmerson, Founding Director PriMetrica