Introduction
In the previous sections of this essay, a model incorporating network size, congestion, and demand for internet usage, has been developed. This model will now be utilised to examine alternate strategies for reducing the congestion externality as it applies to internet traffic. The first alternative to be evaluated involves evaluating the impact of expanding network capacity as a means of alleviating congestion.
Figure 8 (below) presents the typical situation of a network connection of fixed capacity Q# with an initial demand curve D0. Again, assuming that price is set equal to MPC, a price of P2 results in quantity Q0 of traffic being demanded. A congestion externality of P1-P2 exists at this quantity of traffic.
Now let us consider the impact of a decision to reduce the congestion externality by increasing the capacity of the network. In Figure 8 this is represented by extending the horizontal portion of the marginal cost curves from Q# to Q##. The MSC and MPC curves are thus shifted outwards to the right by the amount of the capacity expansion. The new marginal cost curves MSC1 and MPC1 are drawn parallel to the original curves.1 With the new curves MSC1 and MPC1 , a higher quantity of traffic Q0 is demanded at the lower price P3. The congestion externality at this quantity of traffic is now a lower amount, P2-P3.
However it has been argued previously in this essay that because lowering congestion raises the marginal benefit of usage at that network size, that lowering congestion would lead to an increase in demand for usage. Another explanation which supports this hypothosis is based on the argument that time is complimentary to internet usage and that congestion reduces the quality of service of the internet (Villasis (1996:72)). Thus a reduction in congestion will increase demand because it reduces the time taken to complete a given task and is seen by most internet users as a quality improvement.
By reducing the congestion at the network size associated with demand curve D0, this will result in an increase in the marginal benefit of network size at that particular network size. As a consequence of this, assuming no other changes, demand will increase until a price/quantity equilibrium is reached which restores the original congestion externality. 2 The new demand curve is given by D1 with a new price/quantity equilibrium at price P2 and quantity Q1. At this combination the congestion externality is again equal to P1-P2. Thus, whilst a temporary reduction in the congestion externality was achieved by expansion of the physical capacity of the network, the original congestion externality will be restored ceteris paribus.
This situation, where a lowering of congestion leads to an increase of demand and a return to the original problem of congestion, is similar to the situation with respect to highways and motor vehicle traffic. Adding more roads to alleviate traffic congestion may not provide relief from congestion as suppressed demand soon soaks up the expanded capacity. (Balchin & Kieve (1977:161))
However let us now consider the impact of the policy if the reduction in congestion leads to more companies and individuals deciding to connect to the internet, increasing the network size3. In Figure 8, the capacity expansion initially reduced the congestion externality however the externality was eventually restored at the original size. Now, if the reduction in congestion produces an increase in network size, what will be the impact on the congestion externality?
In Figure 9, demand curve D1 represents the demand curve for the original network size with expanded capacity. (i.e. The final position of the demand curve after the expansion and subsequent increase in demand.) However if the reduction in congestion increases the network size, then so long as the marginal private benefit of increased network size is positive, the demand curve for the larger network size will sit to the right of the original network size. In Figure 9 this is represented by the demand curve D2. The higher level of demand D2 produces a new equilibrium at a higher quantity of traffic Q2 and higher price P4. However what is more significant is that P5-P4 > P1-P2 . That is, the congestion externality is larger than under the original network size. Thus capacity expansion, where it leads to increased network size, may increase the size of the congestion externality.
Expansion of capacity is closely related to the concept of overprovisioning. "Overprovisioning means maintaining sufficient network capacity to support the peak demands without noticeable service degradation" (MacKie-Mason & Varian (1997:43)). Overprovisioning is not without its problems. Firstly, it results in excess capacity during non-peak times. The extent of excess capacity required depends on the nature of the traffic. For traffic which flows at a constant rate or can tolerate delay, an average utlisation rate of 90% may provide enough excess capacity to accommodate peak periods. For traffic or tasks which are 'bursty' or are intolerant of delay, the average utilisation may need to be kept below 10% (Kelly (1997:253)). This means that during off-peak times there can be large amounts of capacity not being utilised.
Keeping ahead of the growth in internet traffic is also a problem for strategies based around expansion of capacity. Although the cost of expanding capacity is falling by 30% per year , increased use of multimedia applications and the rapid growth of the internet means this capacity is quickly soaked up (Hazlett (1994:2)). Indeed, "experience suggests that application developers will have little difficulty in designing new services that use up all available resources."(MacKie-Mason, Murphy & Murphy (1997:284)).
Conclusion
Expansion of capacity, by itself, thus does not appear to be a sustainable strategy for dealing with congestion. Expansion of capacity can produce an increase in demand for usage which results in a restoration of the original congestion. If the capacity expansion leads to an increase in network size, capacity expansion may, paradoxically, increase the congestion externality. Attempting to reduce congestion through overprovisioning is also hampered by the nature and rate of growth of internet traffic. Alternate strategies for alleviating internet congestion must therefore be evaluated.
Endnotes
1. That is to say that the slopes of the curves have not been altered since we have assumed at this stage that the only change is an expansion of capacity and that ADU sizes and application mixes remain unchanged.
2. In order to understand the adjustment process consider the following. The consumer starts from the position where the marginal cost of usage equals the marginal benefit of usage. Expanding capacity reduces congestion, which raises the marginal benefit of usage, so that the marginal benefit of usage is greater than the marginal cost. The consumer increases their demand since the marginal benefit of additional usage is greater than the marginal cost. As the consumer increases their demand and the quantity consumed, the price (the cost faced by the consumer) increases, as does congestion. Eventually, the increasing congestion will reduce the marginal benefit such that it equals the increased cost faced by the consumers. Assuming that expansion of capacity produces a constant and proportionate increase in marginal benefit (at any network size) , equilibrium will be restored at the original size of congestion externality, but at a higher quantity of traffic.
3. This relationship between reduced congestion and increases in network size is not covered in this essay and is simply being assumed so as to examine the impact of increased network size on the congestion externality. However since a reduction in congestion increases demand for usage it may be reasonable to assume that some of this increased usage demand comes form new users and therefore that reduced congestion will increase the network size.
Last update 12-Jun-98. To visit my personal web site click here.