Further failures to respect ‘adding-up’ constraints have arisen from substitution between declared services caused by inconsistencies in their relative prices.
The approach the ACCC has adopted to declaration has resulted in access charges needing to be determined for a wide range of services that are close, and in some cases very close, substitutes. Errors in setting relative prices for these services will have two consequences:
They will distort production choices, as access seekers replace inputs that are ‘over-priced’ with those that are ‘under-priced’. This is a first-order welfare loss (meaning, in this case, a distortion that potentially affects all of output, rather than merely having an effect at the margin); and
When the ‘under-priced’ services are priced below costs (or make a contribution to common costs that is insufficient to allow all of those costs to be covered), the shift in access-seeker demand to these services will prevent full cost recovery (or worsen what would in any event have been losses).
Thus, access seekers can provide local carriage service over Telstra’s network either by using the Local Carriage Service, or by using the PSTN OTA service. The latter involves ‘over-riding’ the default routing of the call by inserting a long-distance code in front of the called number and hence is referred to as ‘Local Call Over-ride’ (LCO).
From a technical perspective, LCO is a highly inefficient way of providing local carriage service as it requires between 1.5 and 2 times the resources actually needed to complete a local call.[15] The additional amount is a pure waste of society’s resources.
That waste notwithstanding, LCO accounts for a non-negligible share of total local calling minutes. This is because the ACCC prices LCS on an untimed basis, while pricing PSTN OTA (and hence LCO) on a timed basis (that is, access seekers are charged per minute of use). As a result, it is highly profitable for access seekers to use LCO, rather than LCS, for short-held calls. As local calls generated by businesses tend to be of below-average duration, this can be done by modifying the software businesses have on their PABXs (the private exchanges that manage their traffic) to insert an over-ride code into local calls.
LCO has two effects on the economics of the local carriage service. The first is that it results in the technical inefficiency discussed above. The second is that it aggravates the shortfall Telstra incurs in the supply of local calls.
That shortfall, discussed above, is aggravated because removing short-held local calls from the stock of local calls in and of itself saves very few costs. In effect, the costs of the inter-exchange network (the part of the network that goes between, and includes, local exchanges) tend to be almost completely insensitive to even substantial variations in traffic volumes.[16] Moreover, in reality, LCO does not remove those calls but merely re-routes them through the inter-exchange network, potentially increasing the actual costs that Telstra bears (though Telstra’s loss is partially offset by the LCO payment). On balance, revenues are reduced while no costs are avoided, reducing the extent of cost recovery even further.
Even more serious distortions have arisen as between the ULLS, WLR and LSS services, which are all access line services. Simplifying somewhat, these services are broadly alternative ways of providing the same set of end-user services, including telephony and ADSL.
The ACCC has set prices for these services in a manner that creates enormous scope for inefficient substitution. Underpinning these distortions is the fact that (1) retail prices for line-rental services are capped and (2) Telstra is required to offer specified line-rental services at the same price in all parts of Australia (see Australian Competition Tribunal in Telstra Corporation Ltd (No 3) [2007] ACompT 3 (17 May 2007) at paras. 218–224). These regulatory constraints, applied at the retail level, then create issues for the setting of access charges.
Whether charges for line-rental service cover costs is controversial. It is likely that non-traffic-sensitive costs account for 80 per cent or more of total PSTN costs. Were those costs to be covered by line-rental charges, those charges would need to be substantially higher than they currently are, at least going by Telstra’s estimates of network costs. However, what is clear is that the geographically averaged line-rental charge that is applied in non-metropolitan areas is significantly below the costs of providing line-rental service in those areas, with the gap being especially large in areas of sparse population settlement.
As with LCS, the ACCC has therefore faced a choice in terms of how it set access charges for WLR:
It could set WLR charges on the basis of cost, which would have resulted in those charges exceeding retail line-rental charges in rural and remote areas; or alternatively,
It could use a ‘retail-minus’ methodology, which would mean that charges would be well below costs, most clearly so in rural and remote areas.
The ACCC has chosen the latter approach.
At the same time, the ACCC has set charges for LSS on the basis of assuming that WLR charges, at least on average, entirely cover network costs, so that all that LSS charges need to cover are some incremental costs.[17] Since these costs are assumed to be the same in all areas, the resulting LSS charge is geographically uniform.
Finally, the ULLS charge has been set so as to recover network costs in each area. Specifically, ULLS charges have been set by ‘Bands’, with Band 1 covering the CBDs, Band 2 being the main metropolitan areas (excluding the CBDs), Band 3 being outer metropolitan areas and non-metropolitan population centres (such as regional centres), and Band 4 covering rural and remote areas.
As a result:
WLR and LSS charges are set on a geographically averaged basis, with the sum of these charges being well below costs in non-metropolitan areas;
ULLS charges are set on a geographically de-averaged basis, and hence tend to be well above charges for WLR and LSS in non-metropolitan areas.
This ‘mix and match’ pricing methodology invites ‘cherry picking’, in which demand swings to the inputs that are relatively under-priced.
The ‘cherry-picking’ opportunities arise from the juxtaposition of geographically de-averaged charges for some declared services with geographically averaged charges for others that are substitutes for those that are geographically de-averaged. This juxtaposition makes it possible to use the declared services for which charges are geographically de-averaged in those areas where costs are low, while using the declared services for which charges are geographically averaged in those areas where costs are high.
Thus, for ULLS, which allows an access seeker to provide both voice and data services, the geographically de-averaged prices mean that (at the time of writing) access seekers can use Telstra’s CAN at an access price of $6 per month in CBD areas and $14 per month in other metropolitan areas (2007/08 prices). The ACCC has not made any determination for ULLS prices in rural areas, and has not disclosed any recent estimate of ULLS costs in those areas. However, it has previously suggested that the price of ULLS in rural areas would be $100 per month (see, for example, ACCC 2003: 84, Table 10.4) and research undertaken by the Productivity Commission suggests that costs in those areas are substantially higher.[18]
Whether the ULLS price in rural areas should be $100 or $200 per month is, however, in many respects irrelevant. In effect, the ACCC has set the charges for Telstra’s CAN in rural areas on a geographically averaged basis. Thus, at the time of writing, access seekers can provide voice services in rural areas using the wholesale line-rental service for $23.12 per month for residential customers and $25.84 for business customers.[19] At the same time, they can provide data service to those same customers using the (also geographically averaged) LSS service. The ACCC’s most recent decision on LSS sets a geographically averaged price of just $2.50 per month.[20] The effective input price access seekers face for the CAN in non-metropolitan areas is therefore capped at some $26 per month for residential customers and $29 per month for business customers. These charges are obviously far below TSLRIC-based costs in those areas, even on the ACCC’s estimates.
As a result, the ACCC’s wholesale pricing allows the access provider to recover from access seekers no more than the cost of the CAN in low-cost areas and well below the cost of the CAN from access seekers in rural areas. No other access prices (not even those for services such as terminating access that face highly inelastic demand, so that they could be marked up to recover the resulting shortfall) have been increased to compensate. Total recovery must therefore fall short of total costs, as the multi-service adding-up constraint is breached.
[15] An LCO call must use a local switch at least twice (although a quarter or so of local calls would otherwise only use a local switch once), and additionally requires transmission resources to and from the access seeker’s Point of Interconnection. The Point of Interconnection is the point at which the call exits the Telstra network and is handed over to the access seeker’s network, where it is ‘turned around’ and handed straight back to the Telstra network for termination.
[16] For example, using the Hatfield model for seven U.S. States, the percentage of the dominant local exchange carrier’s total costs that is traffic-sensitive can be calculated as ranging from 15.0 per cent (for Georgia) to 21.1 per cent (for New York). However, increasing traffic by 30 per cent above the default level specified in the model increases total costs in each of the States by less than 1 per cent. The cost savings associated with a similar reduction in traffic levels would be no greater.
[17] Originally, those incremental costs were the ‘LSS-specific’ costs — that is, the fixed set-up costs of providing and managing LSS. However, the ACCC subsequently changed its view on this and shifted to only allowing the LSS to recover a contribution to the averaged costs of managing all wholesale ADSL services — averaged costs that (given the number of ADSL services now in use) are very small.
[18] See Productivity Commission 2000. This research indicates that line costs in low-density areas of Australia are six to 14 times higher than in the rest of Australia. Given the line density in rural areas of Australia and based on the ACCC’s own cost estimates in other areas of Australia this suggests that monthly CAN costs in rural areas are in the order of $140 to $209 per line.
[19] See ACCC 2006.
[20] See ACCC 2007c.