There are a number of things that Oftel intends to do to ensure that Internet users are offered tariff packages for accessing the Internet which meet their needs. Some of the policy initiatives are already underway, others are the development of ideas that have already been floated, while other require further policy initiatives.

There appear to be two key problems relating to accessing the Internet:

• Usage prices are above costs because of the unbalanced tariff for telephony;
• It may be better to recover the costs of serving high users through an unmetered tariff whereby all capacity related costs are recovered by a fixed charge.

This paper addresses each of those issues, and suggest ways in which interconnection charging arrangements could be adjusted to give operators and service providers more flexibility to offer tariff structures that more closely meet the needs of their customers. The paper also addresses some other issues, such as the reasons for not pricing calls at marginal cost and the appropriate relationship between charges in peak and off peak periods.

However, the industry also has an important role to play. Operators are free to introduce new retail tariff structures, but in some circumstances this may be difficult with the current structure of interconnect charges. If this is of concern, it is up to operators to discuss their needs with BT. If negotiations fail, and Oftel considers the request to be reasonable, Oftel can intervene and determine a tariff on the basis of the new tariff structure.

Current telecoms pricing arrangements are not very cost reflective. Line charges are below cost and the loss is made up from profits on calls. In the main this does not matter greatly for telecoms users, since most users pay total charges which broadly relate to the total costs they impose on the system, and very high users are usually able to obtain services from competing operators. Internet usage creates a new problem. The structure of telecoms tariffs penalises those who use the Internet a lot by making them pay total charges which are in excess of the total costs of providing service to them. Competition does not help to solve this problem for many Internet users, as they tend to have no option other than to use the BT network for call origination.

In terms of assessing the effect of telecoms tariffing as it relates specifically to Internet users it is possible to focus on Number Translation Services (NTS) as almost all Internet access is via this type of call. In the past the NTS system tied the charge for a call to an ISP to an operator’s local call charge. It also limited BT’s retention for originating an NTS call to an ISP to a wholesale charge of 1-1.2ppm. The difference between the local call charge to users and BT’s retention has been shared between terminating operators and ISPs. This is the basis for the "Freeserve model" by which all the ISP’s costs are recovered through telephone charges and no subscription charge is made to the customer. It creates a "pay as you go" service for Internet access.

While the "pay as you go" model is attractive for many customers, particularly those who are experimenting with Internet usage and are unwilling to commit to subscription charges, it is not a suitable charging model for heavy usage of the Internet. The call charges levied on customers who make heavy use of the Internet might exceed, perhaps by a considerable amount, the actual costs of providing them with access to the Internet. High usage customers would therefore prefer some form of alternative arrangements, whereby the usage element was rather lower, and where they perhaps paid a subscription charge to the service provider.

The proposed changes to the NTS arrangements that we have set out in our consultation documents on NTS should go some way to addressing this concern. The proposed revised arrangements break the link between NTS services and local call charges, while maintaining the restriction on BT’s origination retention. They allow terminating operators and service providers much more control over the setting of the price charged to the end customer. This should make it possible for a number of alternative pricing structures to develop. There should also

be considerable scope for the offering of tariff packages with lower per minute charges than the current charge for Internet access.

A further development that would be possible would be to separate BT’s retention for NTS calls into two parts reflecting the two key cost drivers for call origination: the cost of setting up a call and the cost of maintaining the call. This is known as two-part charging. Currently, the set-up and duration related costs are averaged together to form a single pence per minute charge. This is appropriate if the main concern is to ensure overall cost recovery, but it has the effect that long duration calls are over-charged and short duration calls are under-charged. This clearly works to the disbenefit of those making long duration calls, such as users of the Internet. Two-part charging would address this problem.

Telecoms systems have high fixed costs which results in marginal costs being considerably below average costs. In general, there is merit in the argument that prices should be related to marginal cost, since economic efficiency is maximised when prices are set equal to marginal costs. However, if prices for all telecoms services were set equal to marginal costs, there would be a very large shortfall of revenue, so that total costs would not be covered. Oftel does not question the entitlement of BT or any other originating operator to recover all of its costs, and accepts that mark ups above marginal cost are required to achieve this. Pricing on the basis of marginal cost is not sustainable What is required is some mark up above marginal cost for various services to recover the fixed cost.

Currently, call origination charges are uniform and equal to average cost. Users contribute to the recovery of fixed costs in direct proportion to the use they make of call origination services. This may not be the most appropriate pricing structure. One alternative is to encourage greater use by offering discounts off BT’s retention for NTS calls to high volume users. Increased use would enable fixed costs to be recovered even though charges are lower for some groups of customers.

Oftel believes it would therefore be worth exploring whether it would be possible to offer discounts on the call origination charge to high volume users of the network. It should be recognised that there might be some billing difficulties with such an arrangement. However, in principle it would be a sound approach to pricing, as it would have the effect of bringing marginal prices more into line with marginal costs, which is likely to promote efficient use of the system.

BT’s origination retention includes an uplift on network costs to recover some of its retail costs. There is a question of how the retail uplift should be adjusted to reflect rapid volume growth. The economies of scale in retail activities are likely to be extremely high, so growth in volumes reduces unit retail costs substantially. Oftel has proposed in its consultation on reform of the NTS regime that the retail uplift in BT’s retention should be subject to a volume adjustment. This would serve to reduce the origination charge for NTS services.

Issue 2: The Costs of Providing Services to High Users Would be Better Recovered Under an Unmetered Option

While reform of the NTS regime should bring benefits to many users of the Internet, there is a group of particularly high users who use the Internet to such an extent that they would be better off paying for dedicated capacity, rather than paying for usage on a pence per minute basis.

It is quite possible to express the marginal costs associated with usage as a flat rate sum. This would represent the costs of the capacity required to handle an "always on" connection to the Internet. Costs of usage in switched telecommunications systems arise because additional usage in the busy hour reduces the grade of service available to other users, unless additional capacity is installed to handle the extra traffic in the busy hour. The marginal costs of additional usage can be assessed by estimating the amount of capacity that is required to be installed to offset the reduction in grade of service that would arise if extra traffic were carried. This calculation would yield an estimate of the marginal cost of an always on connection.

A further question which would have to be addressed is the extent to which customers using such a tariff should contribute towards the recovery of fixed costs which is discussed above.

Leaving aside the question of fixed cost recovery, it is quite possible to derive an estimate of the marginal cost of an always on connection. There is a well established statistical method for assessing the relationship between the number of circuits available, the volume of traffic in the busy hour and the required grade of service. Use of these statistical relationships enables the cost of conveying additional traffic in the busy hour to be computed. This approach may be used to establish the marginal cost of an always on connection. In essence, this would be the cost associated with providing the capacity to carry an additional hour of traffic in the busy hour. In principle, a tariff for unmetered usage could be calculated on this basis.

Currently, interconnection charges are set on the basis that all users contribute to fixed costs in proportion to the extent to which they use the network. No account is taken of the end use to which the service is being put. In principle, it would be possible to adopt a principle of "end use charging", by which the contributions to fixed costs vary according to end usage.

Economic efficiency would be served if the contributions were higher for those services where the elasticity of demand was lower, and lower for those where the elasticity was higher. Currently there is no reliable empirical evidence available to determine whether calls to the Internet have a higher or lower price elasticity than ordinary telephone services. In the absence of such information, it would seem unwise to propose that the contributions to fixed costs should be varied from the current approach.

2. Prices at Different Times of Day

A number of propositions have been put forward about the role of time of day pricing in pricing calls to the Internet.

One proposition that has been made is that the marginal cost of calls in the off peak period is zero, so prices in these periods should also be zero. It is true that the marginal costs of off peak calls could be considered to be zero, in the sense that it is busy hour traffic which determines required capacity and consequently marginal costs. However, it does not follow from this that the appropriate pricing level for calls in off peak periods is zero.

Whether this is the case depends crucially on whether the off peak period in question has the potential to become a peak period if the price were reduced by a significant amount. If this is the case, then it would not be sensible to set a price of zero in this period. In those periods where demand is so low or so unresponsive to price changes that it would be impossible to generate a new peak of traffic, it would be sensible to set a price so that none of the capacity costs were recovered in that period.

A system of peak load pricing is usually proposed as the suitable approach to this problem. This approach to pricing should lead to a flattening out of the peaks and achievement of as even a profile of traffic as is possible, subject to the constraint that demand in some periods might be so low that those periods would always constitute an off peak period.

An alternative argument which has been made in respect of BT’s prices is that the charges for usage in the peak are excessive relative to those in the off peak. This is the precise opposite of the claim underlying the first proposition examined above. Acceptance of the first proposition would imply that the optimum balance of prices would be such that all capacity costs were recovered through peak prices, with none being recovered from charges for usage outside the peak. The first proposition implies that the ratio of peak to off peak prices would be infinite. By contrast, the second proposition would suggest that the ratio should be quite low (at present it is about 4:1 for BT).

There is no relationship between peak and off peak charges which is generally correct. The appropriate relativity will depend upon a number of factors, including the cost of capacity, demand in different periods, and the propensity of consumers to shift demand across time periods in response to price charges. It is therefore not possible, ex ante, to establish whether the price for peak period services is excessive in relation to that of off peak services. The only sensible way to examine this proposition is to consider the relationship between the traffic levels in the various periods of the day. If a reasonably optimal approach to pricing throughout the day is being followed, then the load curve would tend to be reasonably flat.

On average, taking the country as a whole, the BT load curve exhibits some characteristics which would be associated with sensible peak load pricing. In particular, the three peaks which appear during week days, one in the morning, one in the afternoon and one in the evening, are of more or less identical size. This suggests that the load of traffic is being distributed with a reasonable degree of efficiency throughout the day, and would suggest that the relationship between peak and off peak charges is not unreasonable.

There is some debate about the level of marginal cost of usage in modern telecommunication systems. It has been asserted that the marginal cost of additional usage is either zero or very close to zero. In circuit switched networks, this is unlikely to be true, since additional traffic creates a demand for additional circuits, ports and switching capacity.

If the additional traffic is assumed to have the same distribution as existing traffic over time of day and day of week, it is possible to estimate the marginal cost of additional traffic. The average cost of a call minute across the BT network to an Internet service provider is just over 1p per minute. Work performed for Oftel by consultants suggested that the relationship between the marginal cost and the average cost of PSTN conveyance was around 1 to 3. This would imply that the true marginal cost of long distance conveyance through the BT network would be 0.3-0.4p per minute.

The calculations set out above are based on the assumption that additional traffic has the same distribution by time of day and week as existing traffic. This might of course not be the case. If the additional traffic had a peakier distribution, the cost per minute would be higher than in the calculations set out above, while if it had a less peaky distribution it would be lower.

Oftel believes that a number of initiatives, some of which it has taken and some of which it proposes to take, have the potential to increase the flexibility available to operators and service providers in devising tariff structures to meet their customers’ needs. The proposals also have the potential to match Internet tariff structures more closely with the underlying costs of providing service.

In summary, Oftel’s proposals are as follows:

1. The revised arrangements for NTS will break the almost automatic link that currently exists between local call charges and Internet access prices. Under the revised arrangements, one would expect to see a proliferation of charging structures which should better meet consumer demands.
2. The introduction of two-part charging for interconnection of Internet calls should end the current situation whereby calls of longer than average duration are over charged. The effect of this should be to reduce the cost of Internet access.
3. Oftel will consider the question of whether volume based discounts should be available for the interconnection service provided by BT in respect of calls made by high volume customers. The practicalities of such arrangements would need to be investigated.

Oftel acknowledges that, for heavy Internet users, a more appropriate tariff structure would be one that had unmetered usage. In order for this to become a viable proposition, it is probable that an interconnection service on an unmetered basis would need to be made available by BT.

It is for operators to negotiate such arrangements with BT, but Oftel would note that it sees no objection in principle to such arrangements becoming established. It would be willing to intervene and make a determination on an unmetered basis if a request for such an interconnection charging arrangement were referred to it.

Oftel would hope that progress on all of these fronts could be made quickly, so that benefits available from use of the Internet generally, and from the development of e-commerce in particular, are not denied to the UK because of inappropriate tariff structures for Internet access.